national centre for vocational education research - annual …€¦  · web view2016-03-29 · in...

68
Annual transitions between labour market states for young Australians Hielke Buddelmeyer MELBOURNE INSTITUTE OF APPLIED ECONOMIC AND SOCIAL RESEARCH Gary Marks AUSTRALIAN COUNCIL FOR EDUCATIONAL RESEARCH The views and opinions expressed in this document are those of the author/project team and do not necessarily reflect the views of the Australian Government, state and territory governments or NCVER. Any interpretation of data is the responsibility of the author/project team.

Upload: others

Post on 26-Jul-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Annual transitions between labour market states for

young Australians

Hielke BuddelmeyerMELBOURNE INSTITUTE OF APPLIED ECONOMIC

AND SOCIAL RESEARCH

Gary MarksAUSTRALIAN COUNCIL FOR EDUCATIONAL RESEARCH

The views and opinions expressed in this document are those of the authorproject team and do not necessarily reflect the views of

the Australian Government state and territory governments or NCVER

Any interpretation of data is the responsibility of the authorproject team

Publisherrsquos noteTo find other material of interest search VOCED (the UNESCONCVER international database lthttpwwwvocededuaugt) using the following keywords youth employment unemployment outcome of education post-compulsory education statistical analysis vocational education

copy Commonwealth of Australia 2010

This work has been produced by the National Centre for Vocational Education Research (NCVER) under the National Vocational Education and Training Research and Evaluation (NVETRE) Program which is coordinated and managed by NCVER on behalf of the Australian Government and state and territory governments Funding is provided through the Department of Education Employment and Workplace Relations Apart from any use permitted under the Copyright Act 1968 no part of this publication may be reproduced by any process without written permission Requests should be made to NCVER

The NVETRE program is based upon priorities approved by ministers with responsibility for vocational education and training (VET) This research aims to improve policy and practice in the VET sector For further information about the program go to the NCVER website lthttpwwwncvereduaugt The authorproject team was funded to undertake this research via a grant under the NVETRE program These grants are awarded to organisations through a competitive process in which NCVER does not participate

The views and opinions expressed in this document are those of the authorproject team and do not necessarily reflect the views of the Australian Government state and territory governments or NCVER

ISBN 978 1 921413 90 2 web edition978 1 921413 91 9 print edition

TDTNC 10002

Published by NCVERABN 87 007 967 311

Level 11 33 King William Street Adelaide SA 5000PO Box 8288 Station Arcade Adelaide SA 5000 Australia

ph +61 8 8230 8400 fax +61 8 8212 3436email ncverncvereduaulthttpwwwncvereduaugtlthttpwwwncvereduaupublications2247htmlgt

NCVERAbout the research

Annual transitions between labour market states for young AustraliansHielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research

Much analysis of youth transitions focuses on the first year after education or outcomes at a specific age Such work looks for example at the effect of education on the likelihood of being employed or unemployed

This study takes a different angle by considering the effect of education on the persistence of labour market outcomes For example leaving school before Year 12 may be associated with high levels of unemployment but the question is whether such a person is less likely to remain in employment once he or she has a job compared with people with better educational qualifications

Specifically this study examines the role that post-school qualifications play in the annual transitions between labour market states for young Australians and is based on the 1995 cohort of the Longitudinal Surveys of Australian Youth (LSAY) The labour states that were examined were permanent employment casual employment unemployment and not in the labour force The effect of personality traits and ability on labour market transitions was also examined

We know that having post-school qualifications particularly higher-level qualifications increases the chances of permanent employment and the study confirms this However by focusing also on the occurrence of persistent labour market states this work generates new insights The study finds that the most persistent labour market states are casual employment for men while for women it is being out of the labour force Having at least a certificate IV for women or a bachelor degree or higher for men provides a buffer against undesirable labour market states such as unemployment or being out of the labour force becoming persistent

Tom KarmelManaging Director NCVER

ContentsTables 6Executive summary 7Introduction 9

Objective of the research 9Previous studies 10

Methodology 12The data 13Defining mutually exclusive labour market states 13Factors that can help explain labour market status post-qualification 13The general structure of the model 14

Results 18Results from scenario analyses 18

Conclusion 33References 35Appendix 36

4 Annual transitions between labour market states for young Australians

Tables1a Males Results from what-if scenarios based on

parameter estimates ndash Personal characteristics 20

1b Females Results from what-if scenarios based on parameter estimates ndash Personal characteristics 21

2a Males Results from what-if scenarios based on parameter estimates ndash Personality 23

2b Females Results from what-if scenarios based on parameter estimates ndash Personality 24

3a Males Results from what-if scenarios based on parameter estimates ndash Qualifications and past labour market status 27

3b Females Results from what-if scenarios based on parameter estimates ndash Qualifications and past labour market status 28

4a Males Results from what-if scenarios based on parameter estimates ndash Qualifications and (select) past labour market status combined 31

4b Females Results from what-if scenarios based on parameter estimates ndash Qualifications and (select) past labour market status combined 32

A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects 36

A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects 38

A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects 40

A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics 42

A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality 43

NCVER 5

A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status 44

A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined 45

6 Annual transitions between labour market states for young Australians

Executive summaryThis study uses an annual timeframe to evaluate the influence of labour market status in one period on status in the subsequent period Understanding the role of past labour market experiences is important when it is the objective of policy-makers to increase the proportion of time spent by young Australians in desirable labour market states such as full-time work and reduce the time they spend in marginalised activities such as unemployment These concerns are heightened during lean economic times but they never really go away A natural question that may arise especially in a weak labour market for youth is whether promoting casual employment today would lead to more people being employed permanently in the future or would simply result in more people working on a casual basis For such a question to be answered it is necessary to understand the role of previous labour market states and specifically in this study how this role differs by the level of education and qualifications obtained

Four labour market states are considered in this study permanent employment casual employment unemployment and not being in the labour force The analysis used differs from previous research in that it models year-on-year transitions and does not follow the more commonly used approach of taking a group of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later a lsquowhat did they do then and where are they nowrsquo approach This study is therefore a natural complement to this previous research

Using the 1995 cohort of the Longitudinal Surveys of Australian Youth (LSAY) this study finds that there is substantial predictive power in the previous yearrsquos labour market state when modelling the current labour market state In fact the previous yearrsquos state has the largest predictive power of all factors considered including post-school qualifications

Much of the persistence was shown to be due to an underlying process that drives labour market outcomes in all years particularly for casual employment in the case of men and for women being out of the labour market Ignoring this process will result in their effects being passed through the previous labour market status variables thus creating an illusion of severe persistence Controlling for the process shows that although persistence was reduced previous labour market experience has a genuine impact on current labour market status

The study shows that being in full-time permanent employment in the previous periodmdashcompared with the alternative of being out of the labour forcemdashis for women associated with a 194 percentage point increase in

NCVER 7

the probability of being permanently employed in the next period For men the increase is much smaller at 100 percentage points

Much smaller effects are found for factors other than previous labour market states In terms of the level of post-school qualifications higher levels of education are associated with increased probabilities of being permanently employed Although any post-school qualification is better than none the biggest boost is provided by certificate IV and bachelor degree or better for men and bachelor degree or higher for women Post-school qualifications also play a greater role for women in general

In addition to studying the effect of post-school qualifications and previous period labour market state in isolation they are also studied concurrently This addresses the effect of previous period labour market outcome for different levels of post-school qualifications

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men and -147 percentage points for women (off the base prediction of about 85 for both men and women) But when related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (and -273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications Having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

Similarly the effect of the previous period of unemployment on the probability of being employed permanently in the current period ranged from negligible when unemployment is combined with having a bachelor degree or better to -69 percentage points for men in the worst case when unemployment is combined with having no post-school qualifications and -146 percentage points for women (off the base prediction of about 85 permanently employed for both genders) In other words an event that would lead to all of Australiarsquos youth collectively becoming unemployed would almost be completely offset in terms of impact on next periodrsquos labour market outcome if this event resulted in everyonersquos post-school qualification levels being lifted to a bachelor degree or better or alternatively a certificate IV for women1

Although having much smaller an impact policies focusing on the social skills of young Australians could also improve their labour market outcomes Lifting young Australian womenrsquos confidence to a point where they all considered themselves as being very confident would increase the proportion of young women in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about 1 percentage point extra in casual employment while reducing unemployment and exits from the labour force

1 It is hard to think of an event that would result in all young people losing their job The point made here however is about the comparative strength of the post-school qualification variables

8 Annual transitions between labour market states for young Australians

IntroductionThis study examines the dynamics of the labour market particularly in terms of flows between employment unemployment and non-participation in the labour force Of particular interest is the role of post-school qualifications on these labour market transitions It thus naturally fits in with existing studies on labour market dynamics in general What sets this study apart is that the population of interest is very narrowly defined young Australians between the ages of roughly 22 and 25 who have completed their education and training2 To paraphrase we study labour market dynamics for young Australians lsquoafter the dust has settledrsquo

Objective of the researchThe main objective of this project is to investigate the role of post-school qualifications on the transitions between employment and non-employment Specifically the following three questions are addressed1 What are the transitions between the following labour market states for

young Australians permanent employment casual employment unemployment and not being in the labour force

2 How persistent are the following labour market states which can be classified as less desirable for different levels of post-school qualifications casual employment unemployment and not being in the labour force

3 How much of any observed persistence is lsquogenuinersquo and how much of the observed persistence is lsquospuriousrsquo3

The policy narrative of these three research questions runs from getting a good perspective on how socioeconomic and personal characteristics are correlated with labour market choices and what role previous experiences play (question 1) to investigating the role of education and training in escaping less desirable or bad labour market states (question 2) and finally the mechanism at work behind the persistence in labour market states as observed in the data

2 There has been an increasing awareness and emphasis on lifelong learning to ensure that individuals maintain or acquire the necessary skills to participate in society throughout their lives Having completed education and training as used here means that individuals are not observed to be enrolled in study or training leading to a qualification for the duration of the period that is being modelled that is roughly between the ages of 22 and 25 They may take up such study further into their careerlife

3 The concepts of genuine and spurious persistence will be discussed in more detail in the methodology section but here it is sufficient to state that any persistence observed in the data can have two different mechanisms that would give rise to this persistence

NCVER 9

There exists a body of research on the factors associated with particular labour market outcomes and some of the findings from these research efforts are discussed later However much less is known about the process of switching back and forth between labour market states Understanding these dynamics is crucial in order to develop initiatives that would for instance increase transitions out of unemployment and into permanent employment For instance it is a long-standing desire of successive Australian governments to minimise the amount of time that Australian youths spend in so-called marginal activities In this context knowing what factors are associated with being unemployed is not enough It is also necessary to understand the transition from employment into unemployment and the role education and other personal characteristics play in these transitions4

To frame our research objectives we first briefly discuss Australian evidence with respect to the role of education and training in young peoplersquos post-school outcomes A broader literature on labour market dynamics does exist but is not discussed in detail

Previous studiesParticipation in vocational education and training (VET) is an important pathway in the school-to-work transitions for many young people in Australia VET offers opportunities for young people to develop skills that employers value VET comprises apprenticeships and traineeships (which involve employment) and non-apprenticeship courses offered at publicly funded technical and further (TAFE) institutions and by private VET providers VET is particularly useful for young men who did not complete Year 12 of whom just over a half completes an apprenticeship or traineeship (Curtis amp McMillan 2008)

Apprentices are more likely to be male have a low-to-medium socioeconomic background have a parent in a technical or trade occupation have attended a government school and have relatively lower test scores in reading and mathematics while at school They are also likely to have a father with a trade qualification and have studied VET subjects at school (Ainley amp Corrigan 2005 Curtis 2008) Trainees are more likely to be female less likely to come from a professional background and have relatively high scores in reading and mathematics (Ainley amp Corrigan 2005) Participation in non-apprenticeship VET study is generally evenly distributed across social groups although there are some differences by VET course level (McMillan Rothman amp Wernert 2005)

In general participation in VET tends to be associated with subsequent higher rates of full-time employment by comparison with those who undertake no post-school study Curtis (2008) reports that of those who had participated in a non-apprenticeship VET course 66 of non-completers and 78 of completers were working full-time The comparable figure for those with no post-school qualifications was 69 For apprenticeships the level of full-time employment is higher at around 93 for completers and 80 for non-completers For traineeships the level of

4 To give an example we find that obtaining a high enough level of qualification can mitigate the effect of becoming unemployed but these findings and others will be discussed in the results section

10 Annual transitions between labour market states for young Australians

full-time employment was in the middle at around 82 for completers and 79 for non-completers Completion of an apprenticeship has the strongest positive impact on obtaining full-time work which is not surprising as apprenticeships are themselves considered a form of full-time work

However earlier research has suggested that full-time vocational study is not particularly beneficial in terms of labour market outcomes for at least some types of courses Analyses of the three older cohorts from the Youth in Transition (YIT) cohorts born in 1961 1965 and 1970 the youngest YIT cohort born in 1975 and the 1995 Longitudinal Surveys of Australian Youth Year 9 cohort have not found strong positive effects for VET on labour market outcomes (Long McKenzie amp Sturman 1996 Marks Hillman amp Beavis 2003 McMillan amp Marks 2003 but see Ryan 2002) The exception is apprenticeships which do substantially improve labour market outcomes By contrast according to Long (2004 pp20ndash1) the benefits of TAFE qualifications are at best equivocal

In the year after completing their qualification 25 of TAFE graduates were not in full-time work and were not enrolled in further study For those TAFE graduates not studying (47 of all graduates) some form of marginal attachment or no attachment to the labour force is a more likely outcome in the short term than is full employment(Long 2004 p6)

These findings for vocational educationmdashthat apprenticeships improve employment prospects but that other forms of vocational education in general do not substantially improve labour market outcomesmdashis consistent with other work in Australia conducted by economists (Dockery amp Norris 1996 Nevile amp Saunders 1998) and is similar to international research (Ryan 2001)

More recently similar conclusions were reached by Marks (2006) comparing full-time vocational study in the first year after leaving school with full-time work (full-time vocational study includes study at a TAFE or private institution but excludes apprentices whom are classified as full-time workers) He found that full-time vocational study increases the chances of being in full-time study or part-time work in the fourth year after leaving school It does not however increase the chances of full-time work Full-time vocational study in the first year increases the odds of being in full-time study rather than full-time work three years later about three times for men and twice as much for women It also increases the odds of being in part-time rather than full-time work in the fourth year Among men full-time vocational study in the first year (relative to full-time work) increases the likelihood of unemployment and lsquootherrsquo activities in the fourth year The lack of positive effects for full-time study on full-time work several years later is an important finding

It is not clear whether the differences between the conclusions of the earlier and later studies on the impact of VET reflect differences in emphasis placed on estimation results methodological differences the choice of comparison group or that VET is more useful for employment outcomes now than in the 1990s

The approach taken in this study is different from previous studies in one major respect Most of the previous research can be characterised as taking a lsquowhat did they do then and where are they now approachrsquo that is comparing outcomes for those with a VET qualification with those without a VET qualification This is eminently sensible for studying the outcome for

NCVER 11

individuals in a given year (for example what are the most likely labour market states for individuals given their post-school qualifications six years after leaving school) but it is not very well suited to study labour market dynamics Instead we seek to model year-on-year transitions between labour market states and investigate the role education and training has on the probability of making these transitions

12 Annual transitions between labour market states for young Australians

MethodologyAs individuals are interviewed annually information is available on their labour market status on a year-by-year basis In answering the first research question we therefore use an annual timeframe to evaluate the influence of labour market status in one period on labour market status in the subsequent period Note that this implies that we not only capture persistence in particular labour market states but also all transitions from one labour market state to another The different labour market states defined are full-time permanent employment part-time permanent employment casual employment5 unemployment and not being in the labour force

Many factors influence labour market outcomes and it is common not to have indicators for some of them When such unobserved variables are not controlled for in the analysis their effects may be attributed erroneously to observed variables This is what triggers the third research question on the nature of the observed persistence of labour market states

The labelling of genuine and spurious persistence6 is widely used in studies of labour market dynamics and dates back to Heckman (1978) The idea paraphrased here is that there are potentially two mechanisms at work that will lead to the same empirical observation that people unemployed in the previous period are more likely to be unemployed in the current period compared with individuals who were not7 The first mechanism genuine persistence captures that there is something intrinsic about unemployment that has a causal effect on the probability of being unemployed next period It may for instance act to diminish the job-ready skills of an individual or give an individual a stigma that makes them less likely to be hired by employers

The second mechanism spurious persistence captures the fact that there are underlying non-observed factors that drive the probability of being unemployed now and in the future Because these underlying factors affect the probability of being unemployed in every period it only appears that previous unemployment leads to future unemployment In actual effect being unemployed in both periods is driven by the same underlying (unobserved) process This underlying process is typically labelled lsquounobserved heterogeneityrsquo indicating that different people are unique in their own special way and that this uniqueness is not observable to the researcher who is trying to model the individualrsquos labour market dynamics

5 Includes both full-time and part-time casual employment6 Persistence is often referred to as lsquostate dependencersquo7 Unemployment is chosen here to illustrate the point One can readily insert any other

labour market outcome instead

NCVER 13

The methods used to control for the influence of unobserved variables are described later In controlling for these influences using a random effects specification we make two assumptions namely that the unobserved variables are constant over time and secondly that unobserved characteristics are not related to those that are observed As these assumptions are not only necessary but also contentious they will also be discussed later in the methodology section Further in light of the assumptions we make much of what would typically be regarded as unobserved heterogeneity explicit where possible (for example personality traits and ability) and we limit the analysis to a reasonably short four-year window when individuals have completed their post-school education and training making the assumption that the unobserved heterogeneity does not vary over time more palatable

The dataThe data used for this report are based on responses from a nationally representative sample of the cohort of young people who were in Year 9 in 1995 (the Y95 cohort) This cohort sample forms part of the LSAY program The young people in this sample responded to a printed questionnaire and to reading comprehension and numeracy tests conducted in their schools in 1995 to a mailed questionnaire administered in 1996 and to telephone interviews conducted annually since then

The initial sample included 13 613 respondents from approximately 300 government Catholic and independent schools from all Australian states and territories In the 2001 survey responses were received from 6876 individuals and by 2006 3914 individuals provided useable responses The modal age of respondents in 2006 was 25 years Because attrition was not uniform sample weights were used to reflect the original population characteristics

Defining mutually exclusive labour market statesIn order to study the annual transitions between different labour market states it is necessary to identify mutually exclusive labour market states We use the time-consistent version of the LSAY Y95 cohort data as provided by NCVER8 and create mutually exclusive labour market states based on this (1) full-time permanently employed (2) part-time permanently employed (3) casually employed (4) unemployed and (5) not in the labour force

Before describing the model used for the statistical analysis we discuss those factors that are included as explanatory variables for the labour market states we observe

8 This is the version of LSAY95 that is publicly available at the Australian Social Science Data Archive (ASSDA) subject to approval It contains the original data elements in the previous release of the LSAY95 data but also includes a series of derived variables in relation to labour market status education etc that have been made consistent over all 12 waves of the LSAY Y95

14 Annual transitions between labour market states for young Australians

Factors that can help explain labour market status post-qualificationPrevious period labour market stateThe research questions we seek to answer immediately lead to one set of factors that need to be included as explanatory variables lagged versions of our dependent variable Without the lags we cannot say anything about transitions between labour market states

Details of post-school qualificationsIn addition to the lagged labour market states that are included as explanatory factors we also include details on the highest level of post-school qualifications obtained distinguishing between certificates I or II certificate III certificate IV a certificate but at an unknown level an advanced diplomadiploma (including associate degree) and bachelor degree or higher

Personal characteristics of the respondentA small number of personal characteristics of the respondent are included They are indicators for being male and indicators for being married or being in a de facto relationship Those individuals not born in Australia are classified as having been born in either a mainly English-speaking country or a non-English speaking country Furthermore we use a categorised parental occupational class indicator distinguishing between upper upper-middle lower-middle and lower

Reading and maths abilityStudentsrsquo reading and maths ability was tested in wave 1 of LSAY Y95 that is at age 15 These are expressed as achievement scores on a scale from 0 to 20 Self-rated proficiencies are also available but these are not used in favour of the objectively measured achievement levels

Personality traitsStudents were asked in wave 3 (that is at approximately age 17) to rate themselves on a 1 to 4 scale according to specific personality traits with 1 corresponding to lsquoveryrsquo and 4 relating to lsquonot at allrsquo The traits distinguished were being agreeable open popular intellectual calm hardworking outgoing and confident

The general structure of the modelThe method used in the statistical analysis to model the labour market dynamics is a multinomial logit model (MNL) The inclusion of the respondentsrsquo previous labour market states makes it a dynamic multinomial logit model The role of unobserved heterogeneity is accounted for by the inclusion of random effects An alternative way to interpret random effects is to think of them as the constant terms in the multinomial logit model being random The resulting model with random alternative specific constants is known as a form of the lsquomixed multinomial logitrsquo (MMNL) or

NCVER 15

lsquorandom parameter logitrsquo (RPL) model9 The random parameters in this case are the constants in the model This model has considerable advantages when analysing state dependence in the presence of unobserved heterogeneity and will be briefly discussed here

To discuss the modelrsquos structure and to illustrate why this specification is the best choice we first introduce some notation Let the dependent variable Yit denote the labour market state at time t for individual i Factors that influence the choice for Yit are denoted by Xit and lagged values of the dependent variable Yit-1 The explanatory variables in Xit as outlined previously imply a clear direction of the association between Xit and Yit That is Xit influences Yit but the reverse cannot hold The unobserved heterogeneitymdashin this study captured by an individual specific random effectmdashis denoted by μi

Why a logit specificationWhen it comes to modelling discrete choice variables the two main types of models are the logit and the probit models Usually the inclusion of lagged dependent variables creates an endogeneity problem but not so in a mixed logit framework Train (2003 p118) explains the issue in plain language Using our notation the Yit depends on Xit and the error term εit If thatrsquos the case then Yit-1 depends on Xit-1 and the error term εit-1 In the presence of unobserved heterogeneity the error term εit includes the unobserved heterogeneity component μi This μi component in the error terms makes these error terms correlated over time (Train 2003 p115) When one includes the lagged dependent variable Yt-1 on the right-hand side as an explanatory variable it will thus violate the independence assumption between the right-hand side variables and the error term in the equation Yt = γYt-1 + β Xt + εt This is easy to see when one realises that Yt-1 depends on εt-1 and εt and εt-1 are correlated because both include μi If a probit specification were to be used the error structure used in the estimation would have to be corrected to account for this Standard statistical software packages such as Stata now have routines that make this correction for binary probits that include lagged values of the dependent variable in the presence of unobserved heterogeneity10 However these routines have not yet been extended to multinomial probits

Train (2003 p150) explains why choosing a mixed logit set-up including lagged dependent variables as explanatory variables on the right-hand side does not pose a problem in the presence of unobserved heterogeneity unlike the case of a probit specification The crux of it is that conditional on the unobserved heterogeneity μi the estimation boils down to estimating a standard multinomial logit modelmdashwith a single complication the unobserved heterogeneity μi on which it was conditioned needs to be lsquointegrated outrsquo Fortunately this can be done using standard numerical simulation making the mixed logit approach (in our case multinomial mixed logit due to a multiple of choices) an ideal candidate to model state dependence in the presence of unobserved heterogeneity In the next subsections we provide more detail on how the estimation works

9 Any parameter in a MMNL can be modelled as a random variable not just the constants10Note that when it is assumed that there is no correlation over time in the error terms eg

in the absence of unobserved heterogeneity including lagged dependent variables Yt-1 does not pose a problem

16 Annual transitions between labour market states for young Australians

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 2: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Publisherrsquos noteTo find other material of interest search VOCED (the UNESCONCVER international database lthttpwwwvocededuaugt) using the following keywords youth employment unemployment outcome of education post-compulsory education statistical analysis vocational education

copy Commonwealth of Australia 2010

This work has been produced by the National Centre for Vocational Education Research (NCVER) under the National Vocational Education and Training Research and Evaluation (NVETRE) Program which is coordinated and managed by NCVER on behalf of the Australian Government and state and territory governments Funding is provided through the Department of Education Employment and Workplace Relations Apart from any use permitted under the Copyright Act 1968 no part of this publication may be reproduced by any process without written permission Requests should be made to NCVER

The NVETRE program is based upon priorities approved by ministers with responsibility for vocational education and training (VET) This research aims to improve policy and practice in the VET sector For further information about the program go to the NCVER website lthttpwwwncvereduaugt The authorproject team was funded to undertake this research via a grant under the NVETRE program These grants are awarded to organisations through a competitive process in which NCVER does not participate

The views and opinions expressed in this document are those of the authorproject team and do not necessarily reflect the views of the Australian Government state and territory governments or NCVER

ISBN 978 1 921413 90 2 web edition978 1 921413 91 9 print edition

TDTNC 10002

Published by NCVERABN 87 007 967 311

Level 11 33 King William Street Adelaide SA 5000PO Box 8288 Station Arcade Adelaide SA 5000 Australia

ph +61 8 8230 8400 fax +61 8 8212 3436email ncverncvereduaulthttpwwwncvereduaugtlthttpwwwncvereduaupublications2247htmlgt

NCVERAbout the research

Annual transitions between labour market states for young AustraliansHielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research

Much analysis of youth transitions focuses on the first year after education or outcomes at a specific age Such work looks for example at the effect of education on the likelihood of being employed or unemployed

This study takes a different angle by considering the effect of education on the persistence of labour market outcomes For example leaving school before Year 12 may be associated with high levels of unemployment but the question is whether such a person is less likely to remain in employment once he or she has a job compared with people with better educational qualifications

Specifically this study examines the role that post-school qualifications play in the annual transitions between labour market states for young Australians and is based on the 1995 cohort of the Longitudinal Surveys of Australian Youth (LSAY) The labour states that were examined were permanent employment casual employment unemployment and not in the labour force The effect of personality traits and ability on labour market transitions was also examined

We know that having post-school qualifications particularly higher-level qualifications increases the chances of permanent employment and the study confirms this However by focusing also on the occurrence of persistent labour market states this work generates new insights The study finds that the most persistent labour market states are casual employment for men while for women it is being out of the labour force Having at least a certificate IV for women or a bachelor degree or higher for men provides a buffer against undesirable labour market states such as unemployment or being out of the labour force becoming persistent

Tom KarmelManaging Director NCVER

ContentsTables 6Executive summary 7Introduction 9

Objective of the research 9Previous studies 10

Methodology 12The data 13Defining mutually exclusive labour market states 13Factors that can help explain labour market status post-qualification 13The general structure of the model 14

Results 18Results from scenario analyses 18

Conclusion 33References 35Appendix 36

4 Annual transitions between labour market states for young Australians

Tables1a Males Results from what-if scenarios based on

parameter estimates ndash Personal characteristics 20

1b Females Results from what-if scenarios based on parameter estimates ndash Personal characteristics 21

2a Males Results from what-if scenarios based on parameter estimates ndash Personality 23

2b Females Results from what-if scenarios based on parameter estimates ndash Personality 24

3a Males Results from what-if scenarios based on parameter estimates ndash Qualifications and past labour market status 27

3b Females Results from what-if scenarios based on parameter estimates ndash Qualifications and past labour market status 28

4a Males Results from what-if scenarios based on parameter estimates ndash Qualifications and (select) past labour market status combined 31

4b Females Results from what-if scenarios based on parameter estimates ndash Qualifications and (select) past labour market status combined 32

A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects 36

A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects 38

A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects 40

A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics 42

A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality 43

NCVER 5

A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status 44

A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined 45

6 Annual transitions between labour market states for young Australians

Executive summaryThis study uses an annual timeframe to evaluate the influence of labour market status in one period on status in the subsequent period Understanding the role of past labour market experiences is important when it is the objective of policy-makers to increase the proportion of time spent by young Australians in desirable labour market states such as full-time work and reduce the time they spend in marginalised activities such as unemployment These concerns are heightened during lean economic times but they never really go away A natural question that may arise especially in a weak labour market for youth is whether promoting casual employment today would lead to more people being employed permanently in the future or would simply result in more people working on a casual basis For such a question to be answered it is necessary to understand the role of previous labour market states and specifically in this study how this role differs by the level of education and qualifications obtained

Four labour market states are considered in this study permanent employment casual employment unemployment and not being in the labour force The analysis used differs from previous research in that it models year-on-year transitions and does not follow the more commonly used approach of taking a group of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later a lsquowhat did they do then and where are they nowrsquo approach This study is therefore a natural complement to this previous research

Using the 1995 cohort of the Longitudinal Surveys of Australian Youth (LSAY) this study finds that there is substantial predictive power in the previous yearrsquos labour market state when modelling the current labour market state In fact the previous yearrsquos state has the largest predictive power of all factors considered including post-school qualifications

Much of the persistence was shown to be due to an underlying process that drives labour market outcomes in all years particularly for casual employment in the case of men and for women being out of the labour market Ignoring this process will result in their effects being passed through the previous labour market status variables thus creating an illusion of severe persistence Controlling for the process shows that although persistence was reduced previous labour market experience has a genuine impact on current labour market status

The study shows that being in full-time permanent employment in the previous periodmdashcompared with the alternative of being out of the labour forcemdashis for women associated with a 194 percentage point increase in

NCVER 7

the probability of being permanently employed in the next period For men the increase is much smaller at 100 percentage points

Much smaller effects are found for factors other than previous labour market states In terms of the level of post-school qualifications higher levels of education are associated with increased probabilities of being permanently employed Although any post-school qualification is better than none the biggest boost is provided by certificate IV and bachelor degree or better for men and bachelor degree or higher for women Post-school qualifications also play a greater role for women in general

In addition to studying the effect of post-school qualifications and previous period labour market state in isolation they are also studied concurrently This addresses the effect of previous period labour market outcome for different levels of post-school qualifications

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men and -147 percentage points for women (off the base prediction of about 85 for both men and women) But when related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (and -273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications Having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

Similarly the effect of the previous period of unemployment on the probability of being employed permanently in the current period ranged from negligible when unemployment is combined with having a bachelor degree or better to -69 percentage points for men in the worst case when unemployment is combined with having no post-school qualifications and -146 percentage points for women (off the base prediction of about 85 permanently employed for both genders) In other words an event that would lead to all of Australiarsquos youth collectively becoming unemployed would almost be completely offset in terms of impact on next periodrsquos labour market outcome if this event resulted in everyonersquos post-school qualification levels being lifted to a bachelor degree or better or alternatively a certificate IV for women1

Although having much smaller an impact policies focusing on the social skills of young Australians could also improve their labour market outcomes Lifting young Australian womenrsquos confidence to a point where they all considered themselves as being very confident would increase the proportion of young women in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about 1 percentage point extra in casual employment while reducing unemployment and exits from the labour force

1 It is hard to think of an event that would result in all young people losing their job The point made here however is about the comparative strength of the post-school qualification variables

8 Annual transitions between labour market states for young Australians

IntroductionThis study examines the dynamics of the labour market particularly in terms of flows between employment unemployment and non-participation in the labour force Of particular interest is the role of post-school qualifications on these labour market transitions It thus naturally fits in with existing studies on labour market dynamics in general What sets this study apart is that the population of interest is very narrowly defined young Australians between the ages of roughly 22 and 25 who have completed their education and training2 To paraphrase we study labour market dynamics for young Australians lsquoafter the dust has settledrsquo

Objective of the researchThe main objective of this project is to investigate the role of post-school qualifications on the transitions between employment and non-employment Specifically the following three questions are addressed1 What are the transitions between the following labour market states for

young Australians permanent employment casual employment unemployment and not being in the labour force

2 How persistent are the following labour market states which can be classified as less desirable for different levels of post-school qualifications casual employment unemployment and not being in the labour force

3 How much of any observed persistence is lsquogenuinersquo and how much of the observed persistence is lsquospuriousrsquo3

The policy narrative of these three research questions runs from getting a good perspective on how socioeconomic and personal characteristics are correlated with labour market choices and what role previous experiences play (question 1) to investigating the role of education and training in escaping less desirable or bad labour market states (question 2) and finally the mechanism at work behind the persistence in labour market states as observed in the data

2 There has been an increasing awareness and emphasis on lifelong learning to ensure that individuals maintain or acquire the necessary skills to participate in society throughout their lives Having completed education and training as used here means that individuals are not observed to be enrolled in study or training leading to a qualification for the duration of the period that is being modelled that is roughly between the ages of 22 and 25 They may take up such study further into their careerlife

3 The concepts of genuine and spurious persistence will be discussed in more detail in the methodology section but here it is sufficient to state that any persistence observed in the data can have two different mechanisms that would give rise to this persistence

NCVER 9

There exists a body of research on the factors associated with particular labour market outcomes and some of the findings from these research efforts are discussed later However much less is known about the process of switching back and forth between labour market states Understanding these dynamics is crucial in order to develop initiatives that would for instance increase transitions out of unemployment and into permanent employment For instance it is a long-standing desire of successive Australian governments to minimise the amount of time that Australian youths spend in so-called marginal activities In this context knowing what factors are associated with being unemployed is not enough It is also necessary to understand the transition from employment into unemployment and the role education and other personal characteristics play in these transitions4

To frame our research objectives we first briefly discuss Australian evidence with respect to the role of education and training in young peoplersquos post-school outcomes A broader literature on labour market dynamics does exist but is not discussed in detail

Previous studiesParticipation in vocational education and training (VET) is an important pathway in the school-to-work transitions for many young people in Australia VET offers opportunities for young people to develop skills that employers value VET comprises apprenticeships and traineeships (which involve employment) and non-apprenticeship courses offered at publicly funded technical and further (TAFE) institutions and by private VET providers VET is particularly useful for young men who did not complete Year 12 of whom just over a half completes an apprenticeship or traineeship (Curtis amp McMillan 2008)

Apprentices are more likely to be male have a low-to-medium socioeconomic background have a parent in a technical or trade occupation have attended a government school and have relatively lower test scores in reading and mathematics while at school They are also likely to have a father with a trade qualification and have studied VET subjects at school (Ainley amp Corrigan 2005 Curtis 2008) Trainees are more likely to be female less likely to come from a professional background and have relatively high scores in reading and mathematics (Ainley amp Corrigan 2005) Participation in non-apprenticeship VET study is generally evenly distributed across social groups although there are some differences by VET course level (McMillan Rothman amp Wernert 2005)

In general participation in VET tends to be associated with subsequent higher rates of full-time employment by comparison with those who undertake no post-school study Curtis (2008) reports that of those who had participated in a non-apprenticeship VET course 66 of non-completers and 78 of completers were working full-time The comparable figure for those with no post-school qualifications was 69 For apprenticeships the level of full-time employment is higher at around 93 for completers and 80 for non-completers For traineeships the level of

4 To give an example we find that obtaining a high enough level of qualification can mitigate the effect of becoming unemployed but these findings and others will be discussed in the results section

10 Annual transitions between labour market states for young Australians

full-time employment was in the middle at around 82 for completers and 79 for non-completers Completion of an apprenticeship has the strongest positive impact on obtaining full-time work which is not surprising as apprenticeships are themselves considered a form of full-time work

However earlier research has suggested that full-time vocational study is not particularly beneficial in terms of labour market outcomes for at least some types of courses Analyses of the three older cohorts from the Youth in Transition (YIT) cohorts born in 1961 1965 and 1970 the youngest YIT cohort born in 1975 and the 1995 Longitudinal Surveys of Australian Youth Year 9 cohort have not found strong positive effects for VET on labour market outcomes (Long McKenzie amp Sturman 1996 Marks Hillman amp Beavis 2003 McMillan amp Marks 2003 but see Ryan 2002) The exception is apprenticeships which do substantially improve labour market outcomes By contrast according to Long (2004 pp20ndash1) the benefits of TAFE qualifications are at best equivocal

In the year after completing their qualification 25 of TAFE graduates were not in full-time work and were not enrolled in further study For those TAFE graduates not studying (47 of all graduates) some form of marginal attachment or no attachment to the labour force is a more likely outcome in the short term than is full employment(Long 2004 p6)

These findings for vocational educationmdashthat apprenticeships improve employment prospects but that other forms of vocational education in general do not substantially improve labour market outcomesmdashis consistent with other work in Australia conducted by economists (Dockery amp Norris 1996 Nevile amp Saunders 1998) and is similar to international research (Ryan 2001)

More recently similar conclusions were reached by Marks (2006) comparing full-time vocational study in the first year after leaving school with full-time work (full-time vocational study includes study at a TAFE or private institution but excludes apprentices whom are classified as full-time workers) He found that full-time vocational study increases the chances of being in full-time study or part-time work in the fourth year after leaving school It does not however increase the chances of full-time work Full-time vocational study in the first year increases the odds of being in full-time study rather than full-time work three years later about three times for men and twice as much for women It also increases the odds of being in part-time rather than full-time work in the fourth year Among men full-time vocational study in the first year (relative to full-time work) increases the likelihood of unemployment and lsquootherrsquo activities in the fourth year The lack of positive effects for full-time study on full-time work several years later is an important finding

It is not clear whether the differences between the conclusions of the earlier and later studies on the impact of VET reflect differences in emphasis placed on estimation results methodological differences the choice of comparison group or that VET is more useful for employment outcomes now than in the 1990s

The approach taken in this study is different from previous studies in one major respect Most of the previous research can be characterised as taking a lsquowhat did they do then and where are they now approachrsquo that is comparing outcomes for those with a VET qualification with those without a VET qualification This is eminently sensible for studying the outcome for

NCVER 11

individuals in a given year (for example what are the most likely labour market states for individuals given their post-school qualifications six years after leaving school) but it is not very well suited to study labour market dynamics Instead we seek to model year-on-year transitions between labour market states and investigate the role education and training has on the probability of making these transitions

12 Annual transitions between labour market states for young Australians

MethodologyAs individuals are interviewed annually information is available on their labour market status on a year-by-year basis In answering the first research question we therefore use an annual timeframe to evaluate the influence of labour market status in one period on labour market status in the subsequent period Note that this implies that we not only capture persistence in particular labour market states but also all transitions from one labour market state to another The different labour market states defined are full-time permanent employment part-time permanent employment casual employment5 unemployment and not being in the labour force

Many factors influence labour market outcomes and it is common not to have indicators for some of them When such unobserved variables are not controlled for in the analysis their effects may be attributed erroneously to observed variables This is what triggers the third research question on the nature of the observed persistence of labour market states

The labelling of genuine and spurious persistence6 is widely used in studies of labour market dynamics and dates back to Heckman (1978) The idea paraphrased here is that there are potentially two mechanisms at work that will lead to the same empirical observation that people unemployed in the previous period are more likely to be unemployed in the current period compared with individuals who were not7 The first mechanism genuine persistence captures that there is something intrinsic about unemployment that has a causal effect on the probability of being unemployed next period It may for instance act to diminish the job-ready skills of an individual or give an individual a stigma that makes them less likely to be hired by employers

The second mechanism spurious persistence captures the fact that there are underlying non-observed factors that drive the probability of being unemployed now and in the future Because these underlying factors affect the probability of being unemployed in every period it only appears that previous unemployment leads to future unemployment In actual effect being unemployed in both periods is driven by the same underlying (unobserved) process This underlying process is typically labelled lsquounobserved heterogeneityrsquo indicating that different people are unique in their own special way and that this uniqueness is not observable to the researcher who is trying to model the individualrsquos labour market dynamics

5 Includes both full-time and part-time casual employment6 Persistence is often referred to as lsquostate dependencersquo7 Unemployment is chosen here to illustrate the point One can readily insert any other

labour market outcome instead

NCVER 13

The methods used to control for the influence of unobserved variables are described later In controlling for these influences using a random effects specification we make two assumptions namely that the unobserved variables are constant over time and secondly that unobserved characteristics are not related to those that are observed As these assumptions are not only necessary but also contentious they will also be discussed later in the methodology section Further in light of the assumptions we make much of what would typically be regarded as unobserved heterogeneity explicit where possible (for example personality traits and ability) and we limit the analysis to a reasonably short four-year window when individuals have completed their post-school education and training making the assumption that the unobserved heterogeneity does not vary over time more palatable

The dataThe data used for this report are based on responses from a nationally representative sample of the cohort of young people who were in Year 9 in 1995 (the Y95 cohort) This cohort sample forms part of the LSAY program The young people in this sample responded to a printed questionnaire and to reading comprehension and numeracy tests conducted in their schools in 1995 to a mailed questionnaire administered in 1996 and to telephone interviews conducted annually since then

The initial sample included 13 613 respondents from approximately 300 government Catholic and independent schools from all Australian states and territories In the 2001 survey responses were received from 6876 individuals and by 2006 3914 individuals provided useable responses The modal age of respondents in 2006 was 25 years Because attrition was not uniform sample weights were used to reflect the original population characteristics

Defining mutually exclusive labour market statesIn order to study the annual transitions between different labour market states it is necessary to identify mutually exclusive labour market states We use the time-consistent version of the LSAY Y95 cohort data as provided by NCVER8 and create mutually exclusive labour market states based on this (1) full-time permanently employed (2) part-time permanently employed (3) casually employed (4) unemployed and (5) not in the labour force

Before describing the model used for the statistical analysis we discuss those factors that are included as explanatory variables for the labour market states we observe

8 This is the version of LSAY95 that is publicly available at the Australian Social Science Data Archive (ASSDA) subject to approval It contains the original data elements in the previous release of the LSAY95 data but also includes a series of derived variables in relation to labour market status education etc that have been made consistent over all 12 waves of the LSAY Y95

14 Annual transitions between labour market states for young Australians

Factors that can help explain labour market status post-qualificationPrevious period labour market stateThe research questions we seek to answer immediately lead to one set of factors that need to be included as explanatory variables lagged versions of our dependent variable Without the lags we cannot say anything about transitions between labour market states

Details of post-school qualificationsIn addition to the lagged labour market states that are included as explanatory factors we also include details on the highest level of post-school qualifications obtained distinguishing between certificates I or II certificate III certificate IV a certificate but at an unknown level an advanced diplomadiploma (including associate degree) and bachelor degree or higher

Personal characteristics of the respondentA small number of personal characteristics of the respondent are included They are indicators for being male and indicators for being married or being in a de facto relationship Those individuals not born in Australia are classified as having been born in either a mainly English-speaking country or a non-English speaking country Furthermore we use a categorised parental occupational class indicator distinguishing between upper upper-middle lower-middle and lower

Reading and maths abilityStudentsrsquo reading and maths ability was tested in wave 1 of LSAY Y95 that is at age 15 These are expressed as achievement scores on a scale from 0 to 20 Self-rated proficiencies are also available but these are not used in favour of the objectively measured achievement levels

Personality traitsStudents were asked in wave 3 (that is at approximately age 17) to rate themselves on a 1 to 4 scale according to specific personality traits with 1 corresponding to lsquoveryrsquo and 4 relating to lsquonot at allrsquo The traits distinguished were being agreeable open popular intellectual calm hardworking outgoing and confident

The general structure of the modelThe method used in the statistical analysis to model the labour market dynamics is a multinomial logit model (MNL) The inclusion of the respondentsrsquo previous labour market states makes it a dynamic multinomial logit model The role of unobserved heterogeneity is accounted for by the inclusion of random effects An alternative way to interpret random effects is to think of them as the constant terms in the multinomial logit model being random The resulting model with random alternative specific constants is known as a form of the lsquomixed multinomial logitrsquo (MMNL) or

NCVER 15

lsquorandom parameter logitrsquo (RPL) model9 The random parameters in this case are the constants in the model This model has considerable advantages when analysing state dependence in the presence of unobserved heterogeneity and will be briefly discussed here

To discuss the modelrsquos structure and to illustrate why this specification is the best choice we first introduce some notation Let the dependent variable Yit denote the labour market state at time t for individual i Factors that influence the choice for Yit are denoted by Xit and lagged values of the dependent variable Yit-1 The explanatory variables in Xit as outlined previously imply a clear direction of the association between Xit and Yit That is Xit influences Yit but the reverse cannot hold The unobserved heterogeneitymdashin this study captured by an individual specific random effectmdashis denoted by μi

Why a logit specificationWhen it comes to modelling discrete choice variables the two main types of models are the logit and the probit models Usually the inclusion of lagged dependent variables creates an endogeneity problem but not so in a mixed logit framework Train (2003 p118) explains the issue in plain language Using our notation the Yit depends on Xit and the error term εit If thatrsquos the case then Yit-1 depends on Xit-1 and the error term εit-1 In the presence of unobserved heterogeneity the error term εit includes the unobserved heterogeneity component μi This μi component in the error terms makes these error terms correlated over time (Train 2003 p115) When one includes the lagged dependent variable Yt-1 on the right-hand side as an explanatory variable it will thus violate the independence assumption between the right-hand side variables and the error term in the equation Yt = γYt-1 + β Xt + εt This is easy to see when one realises that Yt-1 depends on εt-1 and εt and εt-1 are correlated because both include μi If a probit specification were to be used the error structure used in the estimation would have to be corrected to account for this Standard statistical software packages such as Stata now have routines that make this correction for binary probits that include lagged values of the dependent variable in the presence of unobserved heterogeneity10 However these routines have not yet been extended to multinomial probits

Train (2003 p150) explains why choosing a mixed logit set-up including lagged dependent variables as explanatory variables on the right-hand side does not pose a problem in the presence of unobserved heterogeneity unlike the case of a probit specification The crux of it is that conditional on the unobserved heterogeneity μi the estimation boils down to estimating a standard multinomial logit modelmdashwith a single complication the unobserved heterogeneity μi on which it was conditioned needs to be lsquointegrated outrsquo Fortunately this can be done using standard numerical simulation making the mixed logit approach (in our case multinomial mixed logit due to a multiple of choices) an ideal candidate to model state dependence in the presence of unobserved heterogeneity In the next subsections we provide more detail on how the estimation works

9 Any parameter in a MMNL can be modelled as a random variable not just the constants10Note that when it is assumed that there is no correlation over time in the error terms eg

in the absence of unobserved heterogeneity including lagged dependent variables Yt-1 does not pose a problem

16 Annual transitions between labour market states for young Australians

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 3: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

NCVERAbout the research

Annual transitions between labour market states for young AustraliansHielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research

Much analysis of youth transitions focuses on the first year after education or outcomes at a specific age Such work looks for example at the effect of education on the likelihood of being employed or unemployed

This study takes a different angle by considering the effect of education on the persistence of labour market outcomes For example leaving school before Year 12 may be associated with high levels of unemployment but the question is whether such a person is less likely to remain in employment once he or she has a job compared with people with better educational qualifications

Specifically this study examines the role that post-school qualifications play in the annual transitions between labour market states for young Australians and is based on the 1995 cohort of the Longitudinal Surveys of Australian Youth (LSAY) The labour states that were examined were permanent employment casual employment unemployment and not in the labour force The effect of personality traits and ability on labour market transitions was also examined

We know that having post-school qualifications particularly higher-level qualifications increases the chances of permanent employment and the study confirms this However by focusing also on the occurrence of persistent labour market states this work generates new insights The study finds that the most persistent labour market states are casual employment for men while for women it is being out of the labour force Having at least a certificate IV for women or a bachelor degree or higher for men provides a buffer against undesirable labour market states such as unemployment or being out of the labour force becoming persistent

Tom KarmelManaging Director NCVER

ContentsTables 6Executive summary 7Introduction 9

Objective of the research 9Previous studies 10

Methodology 12The data 13Defining mutually exclusive labour market states 13Factors that can help explain labour market status post-qualification 13The general structure of the model 14

Results 18Results from scenario analyses 18

Conclusion 33References 35Appendix 36

4 Annual transitions between labour market states for young Australians

Tables1a Males Results from what-if scenarios based on

parameter estimates ndash Personal characteristics 20

1b Females Results from what-if scenarios based on parameter estimates ndash Personal characteristics 21

2a Males Results from what-if scenarios based on parameter estimates ndash Personality 23

2b Females Results from what-if scenarios based on parameter estimates ndash Personality 24

3a Males Results from what-if scenarios based on parameter estimates ndash Qualifications and past labour market status 27

3b Females Results from what-if scenarios based on parameter estimates ndash Qualifications and past labour market status 28

4a Males Results from what-if scenarios based on parameter estimates ndash Qualifications and (select) past labour market status combined 31

4b Females Results from what-if scenarios based on parameter estimates ndash Qualifications and (select) past labour market status combined 32

A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects 36

A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects 38

A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects 40

A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics 42

A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality 43

NCVER 5

A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status 44

A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined 45

6 Annual transitions between labour market states for young Australians

Executive summaryThis study uses an annual timeframe to evaluate the influence of labour market status in one period on status in the subsequent period Understanding the role of past labour market experiences is important when it is the objective of policy-makers to increase the proportion of time spent by young Australians in desirable labour market states such as full-time work and reduce the time they spend in marginalised activities such as unemployment These concerns are heightened during lean economic times but they never really go away A natural question that may arise especially in a weak labour market for youth is whether promoting casual employment today would lead to more people being employed permanently in the future or would simply result in more people working on a casual basis For such a question to be answered it is necessary to understand the role of previous labour market states and specifically in this study how this role differs by the level of education and qualifications obtained

Four labour market states are considered in this study permanent employment casual employment unemployment and not being in the labour force The analysis used differs from previous research in that it models year-on-year transitions and does not follow the more commonly used approach of taking a group of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later a lsquowhat did they do then and where are they nowrsquo approach This study is therefore a natural complement to this previous research

Using the 1995 cohort of the Longitudinal Surveys of Australian Youth (LSAY) this study finds that there is substantial predictive power in the previous yearrsquos labour market state when modelling the current labour market state In fact the previous yearrsquos state has the largest predictive power of all factors considered including post-school qualifications

Much of the persistence was shown to be due to an underlying process that drives labour market outcomes in all years particularly for casual employment in the case of men and for women being out of the labour market Ignoring this process will result in their effects being passed through the previous labour market status variables thus creating an illusion of severe persistence Controlling for the process shows that although persistence was reduced previous labour market experience has a genuine impact on current labour market status

The study shows that being in full-time permanent employment in the previous periodmdashcompared with the alternative of being out of the labour forcemdashis for women associated with a 194 percentage point increase in

NCVER 7

the probability of being permanently employed in the next period For men the increase is much smaller at 100 percentage points

Much smaller effects are found for factors other than previous labour market states In terms of the level of post-school qualifications higher levels of education are associated with increased probabilities of being permanently employed Although any post-school qualification is better than none the biggest boost is provided by certificate IV and bachelor degree or better for men and bachelor degree or higher for women Post-school qualifications also play a greater role for women in general

In addition to studying the effect of post-school qualifications and previous period labour market state in isolation they are also studied concurrently This addresses the effect of previous period labour market outcome for different levels of post-school qualifications

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men and -147 percentage points for women (off the base prediction of about 85 for both men and women) But when related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (and -273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications Having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

Similarly the effect of the previous period of unemployment on the probability of being employed permanently in the current period ranged from negligible when unemployment is combined with having a bachelor degree or better to -69 percentage points for men in the worst case when unemployment is combined with having no post-school qualifications and -146 percentage points for women (off the base prediction of about 85 permanently employed for both genders) In other words an event that would lead to all of Australiarsquos youth collectively becoming unemployed would almost be completely offset in terms of impact on next periodrsquos labour market outcome if this event resulted in everyonersquos post-school qualification levels being lifted to a bachelor degree or better or alternatively a certificate IV for women1

Although having much smaller an impact policies focusing on the social skills of young Australians could also improve their labour market outcomes Lifting young Australian womenrsquos confidence to a point where they all considered themselves as being very confident would increase the proportion of young women in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about 1 percentage point extra in casual employment while reducing unemployment and exits from the labour force

1 It is hard to think of an event that would result in all young people losing their job The point made here however is about the comparative strength of the post-school qualification variables

8 Annual transitions between labour market states for young Australians

IntroductionThis study examines the dynamics of the labour market particularly in terms of flows between employment unemployment and non-participation in the labour force Of particular interest is the role of post-school qualifications on these labour market transitions It thus naturally fits in with existing studies on labour market dynamics in general What sets this study apart is that the population of interest is very narrowly defined young Australians between the ages of roughly 22 and 25 who have completed their education and training2 To paraphrase we study labour market dynamics for young Australians lsquoafter the dust has settledrsquo

Objective of the researchThe main objective of this project is to investigate the role of post-school qualifications on the transitions between employment and non-employment Specifically the following three questions are addressed1 What are the transitions between the following labour market states for

young Australians permanent employment casual employment unemployment and not being in the labour force

2 How persistent are the following labour market states which can be classified as less desirable for different levels of post-school qualifications casual employment unemployment and not being in the labour force

3 How much of any observed persistence is lsquogenuinersquo and how much of the observed persistence is lsquospuriousrsquo3

The policy narrative of these three research questions runs from getting a good perspective on how socioeconomic and personal characteristics are correlated with labour market choices and what role previous experiences play (question 1) to investigating the role of education and training in escaping less desirable or bad labour market states (question 2) and finally the mechanism at work behind the persistence in labour market states as observed in the data

2 There has been an increasing awareness and emphasis on lifelong learning to ensure that individuals maintain or acquire the necessary skills to participate in society throughout their lives Having completed education and training as used here means that individuals are not observed to be enrolled in study or training leading to a qualification for the duration of the period that is being modelled that is roughly between the ages of 22 and 25 They may take up such study further into their careerlife

3 The concepts of genuine and spurious persistence will be discussed in more detail in the methodology section but here it is sufficient to state that any persistence observed in the data can have two different mechanisms that would give rise to this persistence

NCVER 9

There exists a body of research on the factors associated with particular labour market outcomes and some of the findings from these research efforts are discussed later However much less is known about the process of switching back and forth between labour market states Understanding these dynamics is crucial in order to develop initiatives that would for instance increase transitions out of unemployment and into permanent employment For instance it is a long-standing desire of successive Australian governments to minimise the amount of time that Australian youths spend in so-called marginal activities In this context knowing what factors are associated with being unemployed is not enough It is also necessary to understand the transition from employment into unemployment and the role education and other personal characteristics play in these transitions4

To frame our research objectives we first briefly discuss Australian evidence with respect to the role of education and training in young peoplersquos post-school outcomes A broader literature on labour market dynamics does exist but is not discussed in detail

Previous studiesParticipation in vocational education and training (VET) is an important pathway in the school-to-work transitions for many young people in Australia VET offers opportunities for young people to develop skills that employers value VET comprises apprenticeships and traineeships (which involve employment) and non-apprenticeship courses offered at publicly funded technical and further (TAFE) institutions and by private VET providers VET is particularly useful for young men who did not complete Year 12 of whom just over a half completes an apprenticeship or traineeship (Curtis amp McMillan 2008)

Apprentices are more likely to be male have a low-to-medium socioeconomic background have a parent in a technical or trade occupation have attended a government school and have relatively lower test scores in reading and mathematics while at school They are also likely to have a father with a trade qualification and have studied VET subjects at school (Ainley amp Corrigan 2005 Curtis 2008) Trainees are more likely to be female less likely to come from a professional background and have relatively high scores in reading and mathematics (Ainley amp Corrigan 2005) Participation in non-apprenticeship VET study is generally evenly distributed across social groups although there are some differences by VET course level (McMillan Rothman amp Wernert 2005)

In general participation in VET tends to be associated with subsequent higher rates of full-time employment by comparison with those who undertake no post-school study Curtis (2008) reports that of those who had participated in a non-apprenticeship VET course 66 of non-completers and 78 of completers were working full-time The comparable figure for those with no post-school qualifications was 69 For apprenticeships the level of full-time employment is higher at around 93 for completers and 80 for non-completers For traineeships the level of

4 To give an example we find that obtaining a high enough level of qualification can mitigate the effect of becoming unemployed but these findings and others will be discussed in the results section

10 Annual transitions between labour market states for young Australians

full-time employment was in the middle at around 82 for completers and 79 for non-completers Completion of an apprenticeship has the strongest positive impact on obtaining full-time work which is not surprising as apprenticeships are themselves considered a form of full-time work

However earlier research has suggested that full-time vocational study is not particularly beneficial in terms of labour market outcomes for at least some types of courses Analyses of the three older cohorts from the Youth in Transition (YIT) cohorts born in 1961 1965 and 1970 the youngest YIT cohort born in 1975 and the 1995 Longitudinal Surveys of Australian Youth Year 9 cohort have not found strong positive effects for VET on labour market outcomes (Long McKenzie amp Sturman 1996 Marks Hillman amp Beavis 2003 McMillan amp Marks 2003 but see Ryan 2002) The exception is apprenticeships which do substantially improve labour market outcomes By contrast according to Long (2004 pp20ndash1) the benefits of TAFE qualifications are at best equivocal

In the year after completing their qualification 25 of TAFE graduates were not in full-time work and were not enrolled in further study For those TAFE graduates not studying (47 of all graduates) some form of marginal attachment or no attachment to the labour force is a more likely outcome in the short term than is full employment(Long 2004 p6)

These findings for vocational educationmdashthat apprenticeships improve employment prospects but that other forms of vocational education in general do not substantially improve labour market outcomesmdashis consistent with other work in Australia conducted by economists (Dockery amp Norris 1996 Nevile amp Saunders 1998) and is similar to international research (Ryan 2001)

More recently similar conclusions were reached by Marks (2006) comparing full-time vocational study in the first year after leaving school with full-time work (full-time vocational study includes study at a TAFE or private institution but excludes apprentices whom are classified as full-time workers) He found that full-time vocational study increases the chances of being in full-time study or part-time work in the fourth year after leaving school It does not however increase the chances of full-time work Full-time vocational study in the first year increases the odds of being in full-time study rather than full-time work three years later about three times for men and twice as much for women It also increases the odds of being in part-time rather than full-time work in the fourth year Among men full-time vocational study in the first year (relative to full-time work) increases the likelihood of unemployment and lsquootherrsquo activities in the fourth year The lack of positive effects for full-time study on full-time work several years later is an important finding

It is not clear whether the differences between the conclusions of the earlier and later studies on the impact of VET reflect differences in emphasis placed on estimation results methodological differences the choice of comparison group or that VET is more useful for employment outcomes now than in the 1990s

The approach taken in this study is different from previous studies in one major respect Most of the previous research can be characterised as taking a lsquowhat did they do then and where are they now approachrsquo that is comparing outcomes for those with a VET qualification with those without a VET qualification This is eminently sensible for studying the outcome for

NCVER 11

individuals in a given year (for example what are the most likely labour market states for individuals given their post-school qualifications six years after leaving school) but it is not very well suited to study labour market dynamics Instead we seek to model year-on-year transitions between labour market states and investigate the role education and training has on the probability of making these transitions

12 Annual transitions between labour market states for young Australians

MethodologyAs individuals are interviewed annually information is available on their labour market status on a year-by-year basis In answering the first research question we therefore use an annual timeframe to evaluate the influence of labour market status in one period on labour market status in the subsequent period Note that this implies that we not only capture persistence in particular labour market states but also all transitions from one labour market state to another The different labour market states defined are full-time permanent employment part-time permanent employment casual employment5 unemployment and not being in the labour force

Many factors influence labour market outcomes and it is common not to have indicators for some of them When such unobserved variables are not controlled for in the analysis their effects may be attributed erroneously to observed variables This is what triggers the third research question on the nature of the observed persistence of labour market states

The labelling of genuine and spurious persistence6 is widely used in studies of labour market dynamics and dates back to Heckman (1978) The idea paraphrased here is that there are potentially two mechanisms at work that will lead to the same empirical observation that people unemployed in the previous period are more likely to be unemployed in the current period compared with individuals who were not7 The first mechanism genuine persistence captures that there is something intrinsic about unemployment that has a causal effect on the probability of being unemployed next period It may for instance act to diminish the job-ready skills of an individual or give an individual a stigma that makes them less likely to be hired by employers

The second mechanism spurious persistence captures the fact that there are underlying non-observed factors that drive the probability of being unemployed now and in the future Because these underlying factors affect the probability of being unemployed in every period it only appears that previous unemployment leads to future unemployment In actual effect being unemployed in both periods is driven by the same underlying (unobserved) process This underlying process is typically labelled lsquounobserved heterogeneityrsquo indicating that different people are unique in their own special way and that this uniqueness is not observable to the researcher who is trying to model the individualrsquos labour market dynamics

5 Includes both full-time and part-time casual employment6 Persistence is often referred to as lsquostate dependencersquo7 Unemployment is chosen here to illustrate the point One can readily insert any other

labour market outcome instead

NCVER 13

The methods used to control for the influence of unobserved variables are described later In controlling for these influences using a random effects specification we make two assumptions namely that the unobserved variables are constant over time and secondly that unobserved characteristics are not related to those that are observed As these assumptions are not only necessary but also contentious they will also be discussed later in the methodology section Further in light of the assumptions we make much of what would typically be regarded as unobserved heterogeneity explicit where possible (for example personality traits and ability) and we limit the analysis to a reasonably short four-year window when individuals have completed their post-school education and training making the assumption that the unobserved heterogeneity does not vary over time more palatable

The dataThe data used for this report are based on responses from a nationally representative sample of the cohort of young people who were in Year 9 in 1995 (the Y95 cohort) This cohort sample forms part of the LSAY program The young people in this sample responded to a printed questionnaire and to reading comprehension and numeracy tests conducted in their schools in 1995 to a mailed questionnaire administered in 1996 and to telephone interviews conducted annually since then

The initial sample included 13 613 respondents from approximately 300 government Catholic and independent schools from all Australian states and territories In the 2001 survey responses were received from 6876 individuals and by 2006 3914 individuals provided useable responses The modal age of respondents in 2006 was 25 years Because attrition was not uniform sample weights were used to reflect the original population characteristics

Defining mutually exclusive labour market statesIn order to study the annual transitions between different labour market states it is necessary to identify mutually exclusive labour market states We use the time-consistent version of the LSAY Y95 cohort data as provided by NCVER8 and create mutually exclusive labour market states based on this (1) full-time permanently employed (2) part-time permanently employed (3) casually employed (4) unemployed and (5) not in the labour force

Before describing the model used for the statistical analysis we discuss those factors that are included as explanatory variables for the labour market states we observe

8 This is the version of LSAY95 that is publicly available at the Australian Social Science Data Archive (ASSDA) subject to approval It contains the original data elements in the previous release of the LSAY95 data but also includes a series of derived variables in relation to labour market status education etc that have been made consistent over all 12 waves of the LSAY Y95

14 Annual transitions between labour market states for young Australians

Factors that can help explain labour market status post-qualificationPrevious period labour market stateThe research questions we seek to answer immediately lead to one set of factors that need to be included as explanatory variables lagged versions of our dependent variable Without the lags we cannot say anything about transitions between labour market states

Details of post-school qualificationsIn addition to the lagged labour market states that are included as explanatory factors we also include details on the highest level of post-school qualifications obtained distinguishing between certificates I or II certificate III certificate IV a certificate but at an unknown level an advanced diplomadiploma (including associate degree) and bachelor degree or higher

Personal characteristics of the respondentA small number of personal characteristics of the respondent are included They are indicators for being male and indicators for being married or being in a de facto relationship Those individuals not born in Australia are classified as having been born in either a mainly English-speaking country or a non-English speaking country Furthermore we use a categorised parental occupational class indicator distinguishing between upper upper-middle lower-middle and lower

Reading and maths abilityStudentsrsquo reading and maths ability was tested in wave 1 of LSAY Y95 that is at age 15 These are expressed as achievement scores on a scale from 0 to 20 Self-rated proficiencies are also available but these are not used in favour of the objectively measured achievement levels

Personality traitsStudents were asked in wave 3 (that is at approximately age 17) to rate themselves on a 1 to 4 scale according to specific personality traits with 1 corresponding to lsquoveryrsquo and 4 relating to lsquonot at allrsquo The traits distinguished were being agreeable open popular intellectual calm hardworking outgoing and confident

The general structure of the modelThe method used in the statistical analysis to model the labour market dynamics is a multinomial logit model (MNL) The inclusion of the respondentsrsquo previous labour market states makes it a dynamic multinomial logit model The role of unobserved heterogeneity is accounted for by the inclusion of random effects An alternative way to interpret random effects is to think of them as the constant terms in the multinomial logit model being random The resulting model with random alternative specific constants is known as a form of the lsquomixed multinomial logitrsquo (MMNL) or

NCVER 15

lsquorandom parameter logitrsquo (RPL) model9 The random parameters in this case are the constants in the model This model has considerable advantages when analysing state dependence in the presence of unobserved heterogeneity and will be briefly discussed here

To discuss the modelrsquos structure and to illustrate why this specification is the best choice we first introduce some notation Let the dependent variable Yit denote the labour market state at time t for individual i Factors that influence the choice for Yit are denoted by Xit and lagged values of the dependent variable Yit-1 The explanatory variables in Xit as outlined previously imply a clear direction of the association between Xit and Yit That is Xit influences Yit but the reverse cannot hold The unobserved heterogeneitymdashin this study captured by an individual specific random effectmdashis denoted by μi

Why a logit specificationWhen it comes to modelling discrete choice variables the two main types of models are the logit and the probit models Usually the inclusion of lagged dependent variables creates an endogeneity problem but not so in a mixed logit framework Train (2003 p118) explains the issue in plain language Using our notation the Yit depends on Xit and the error term εit If thatrsquos the case then Yit-1 depends on Xit-1 and the error term εit-1 In the presence of unobserved heterogeneity the error term εit includes the unobserved heterogeneity component μi This μi component in the error terms makes these error terms correlated over time (Train 2003 p115) When one includes the lagged dependent variable Yt-1 on the right-hand side as an explanatory variable it will thus violate the independence assumption between the right-hand side variables and the error term in the equation Yt = γYt-1 + β Xt + εt This is easy to see when one realises that Yt-1 depends on εt-1 and εt and εt-1 are correlated because both include μi If a probit specification were to be used the error structure used in the estimation would have to be corrected to account for this Standard statistical software packages such as Stata now have routines that make this correction for binary probits that include lagged values of the dependent variable in the presence of unobserved heterogeneity10 However these routines have not yet been extended to multinomial probits

Train (2003 p150) explains why choosing a mixed logit set-up including lagged dependent variables as explanatory variables on the right-hand side does not pose a problem in the presence of unobserved heterogeneity unlike the case of a probit specification The crux of it is that conditional on the unobserved heterogeneity μi the estimation boils down to estimating a standard multinomial logit modelmdashwith a single complication the unobserved heterogeneity μi on which it was conditioned needs to be lsquointegrated outrsquo Fortunately this can be done using standard numerical simulation making the mixed logit approach (in our case multinomial mixed logit due to a multiple of choices) an ideal candidate to model state dependence in the presence of unobserved heterogeneity In the next subsections we provide more detail on how the estimation works

9 Any parameter in a MMNL can be modelled as a random variable not just the constants10Note that when it is assumed that there is no correlation over time in the error terms eg

in the absence of unobserved heterogeneity including lagged dependent variables Yt-1 does not pose a problem

16 Annual transitions between labour market states for young Australians

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 4: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

ContentsTables 6Executive summary 7Introduction 9

Objective of the research 9Previous studies 10

Methodology 12The data 13Defining mutually exclusive labour market states 13Factors that can help explain labour market status post-qualification 13The general structure of the model 14

Results 18Results from scenario analyses 18

Conclusion 33References 35Appendix 36

4 Annual transitions between labour market states for young Australians

Tables1a Males Results from what-if scenarios based on

parameter estimates ndash Personal characteristics 20

1b Females Results from what-if scenarios based on parameter estimates ndash Personal characteristics 21

2a Males Results from what-if scenarios based on parameter estimates ndash Personality 23

2b Females Results from what-if scenarios based on parameter estimates ndash Personality 24

3a Males Results from what-if scenarios based on parameter estimates ndash Qualifications and past labour market status 27

3b Females Results from what-if scenarios based on parameter estimates ndash Qualifications and past labour market status 28

4a Males Results from what-if scenarios based on parameter estimates ndash Qualifications and (select) past labour market status combined 31

4b Females Results from what-if scenarios based on parameter estimates ndash Qualifications and (select) past labour market status combined 32

A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects 36

A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects 38

A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects 40

A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics 42

A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality 43

NCVER 5

A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status 44

A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined 45

6 Annual transitions between labour market states for young Australians

Executive summaryThis study uses an annual timeframe to evaluate the influence of labour market status in one period on status in the subsequent period Understanding the role of past labour market experiences is important when it is the objective of policy-makers to increase the proportion of time spent by young Australians in desirable labour market states such as full-time work and reduce the time they spend in marginalised activities such as unemployment These concerns are heightened during lean economic times but they never really go away A natural question that may arise especially in a weak labour market for youth is whether promoting casual employment today would lead to more people being employed permanently in the future or would simply result in more people working on a casual basis For such a question to be answered it is necessary to understand the role of previous labour market states and specifically in this study how this role differs by the level of education and qualifications obtained

Four labour market states are considered in this study permanent employment casual employment unemployment and not being in the labour force The analysis used differs from previous research in that it models year-on-year transitions and does not follow the more commonly used approach of taking a group of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later a lsquowhat did they do then and where are they nowrsquo approach This study is therefore a natural complement to this previous research

Using the 1995 cohort of the Longitudinal Surveys of Australian Youth (LSAY) this study finds that there is substantial predictive power in the previous yearrsquos labour market state when modelling the current labour market state In fact the previous yearrsquos state has the largest predictive power of all factors considered including post-school qualifications

Much of the persistence was shown to be due to an underlying process that drives labour market outcomes in all years particularly for casual employment in the case of men and for women being out of the labour market Ignoring this process will result in their effects being passed through the previous labour market status variables thus creating an illusion of severe persistence Controlling for the process shows that although persistence was reduced previous labour market experience has a genuine impact on current labour market status

The study shows that being in full-time permanent employment in the previous periodmdashcompared with the alternative of being out of the labour forcemdashis for women associated with a 194 percentage point increase in

NCVER 7

the probability of being permanently employed in the next period For men the increase is much smaller at 100 percentage points

Much smaller effects are found for factors other than previous labour market states In terms of the level of post-school qualifications higher levels of education are associated with increased probabilities of being permanently employed Although any post-school qualification is better than none the biggest boost is provided by certificate IV and bachelor degree or better for men and bachelor degree or higher for women Post-school qualifications also play a greater role for women in general

In addition to studying the effect of post-school qualifications and previous period labour market state in isolation they are also studied concurrently This addresses the effect of previous period labour market outcome for different levels of post-school qualifications

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men and -147 percentage points for women (off the base prediction of about 85 for both men and women) But when related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (and -273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications Having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

Similarly the effect of the previous period of unemployment on the probability of being employed permanently in the current period ranged from negligible when unemployment is combined with having a bachelor degree or better to -69 percentage points for men in the worst case when unemployment is combined with having no post-school qualifications and -146 percentage points for women (off the base prediction of about 85 permanently employed for both genders) In other words an event that would lead to all of Australiarsquos youth collectively becoming unemployed would almost be completely offset in terms of impact on next periodrsquos labour market outcome if this event resulted in everyonersquos post-school qualification levels being lifted to a bachelor degree or better or alternatively a certificate IV for women1

Although having much smaller an impact policies focusing on the social skills of young Australians could also improve their labour market outcomes Lifting young Australian womenrsquos confidence to a point where they all considered themselves as being very confident would increase the proportion of young women in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about 1 percentage point extra in casual employment while reducing unemployment and exits from the labour force

1 It is hard to think of an event that would result in all young people losing their job The point made here however is about the comparative strength of the post-school qualification variables

8 Annual transitions between labour market states for young Australians

IntroductionThis study examines the dynamics of the labour market particularly in terms of flows between employment unemployment and non-participation in the labour force Of particular interest is the role of post-school qualifications on these labour market transitions It thus naturally fits in with existing studies on labour market dynamics in general What sets this study apart is that the population of interest is very narrowly defined young Australians between the ages of roughly 22 and 25 who have completed their education and training2 To paraphrase we study labour market dynamics for young Australians lsquoafter the dust has settledrsquo

Objective of the researchThe main objective of this project is to investigate the role of post-school qualifications on the transitions between employment and non-employment Specifically the following three questions are addressed1 What are the transitions between the following labour market states for

young Australians permanent employment casual employment unemployment and not being in the labour force

2 How persistent are the following labour market states which can be classified as less desirable for different levels of post-school qualifications casual employment unemployment and not being in the labour force

3 How much of any observed persistence is lsquogenuinersquo and how much of the observed persistence is lsquospuriousrsquo3

The policy narrative of these three research questions runs from getting a good perspective on how socioeconomic and personal characteristics are correlated with labour market choices and what role previous experiences play (question 1) to investigating the role of education and training in escaping less desirable or bad labour market states (question 2) and finally the mechanism at work behind the persistence in labour market states as observed in the data

2 There has been an increasing awareness and emphasis on lifelong learning to ensure that individuals maintain or acquire the necessary skills to participate in society throughout their lives Having completed education and training as used here means that individuals are not observed to be enrolled in study or training leading to a qualification for the duration of the period that is being modelled that is roughly between the ages of 22 and 25 They may take up such study further into their careerlife

3 The concepts of genuine and spurious persistence will be discussed in more detail in the methodology section but here it is sufficient to state that any persistence observed in the data can have two different mechanisms that would give rise to this persistence

NCVER 9

There exists a body of research on the factors associated with particular labour market outcomes and some of the findings from these research efforts are discussed later However much less is known about the process of switching back and forth between labour market states Understanding these dynamics is crucial in order to develop initiatives that would for instance increase transitions out of unemployment and into permanent employment For instance it is a long-standing desire of successive Australian governments to minimise the amount of time that Australian youths spend in so-called marginal activities In this context knowing what factors are associated with being unemployed is not enough It is also necessary to understand the transition from employment into unemployment and the role education and other personal characteristics play in these transitions4

To frame our research objectives we first briefly discuss Australian evidence with respect to the role of education and training in young peoplersquos post-school outcomes A broader literature on labour market dynamics does exist but is not discussed in detail

Previous studiesParticipation in vocational education and training (VET) is an important pathway in the school-to-work transitions for many young people in Australia VET offers opportunities for young people to develop skills that employers value VET comprises apprenticeships and traineeships (which involve employment) and non-apprenticeship courses offered at publicly funded technical and further (TAFE) institutions and by private VET providers VET is particularly useful for young men who did not complete Year 12 of whom just over a half completes an apprenticeship or traineeship (Curtis amp McMillan 2008)

Apprentices are more likely to be male have a low-to-medium socioeconomic background have a parent in a technical or trade occupation have attended a government school and have relatively lower test scores in reading and mathematics while at school They are also likely to have a father with a trade qualification and have studied VET subjects at school (Ainley amp Corrigan 2005 Curtis 2008) Trainees are more likely to be female less likely to come from a professional background and have relatively high scores in reading and mathematics (Ainley amp Corrigan 2005) Participation in non-apprenticeship VET study is generally evenly distributed across social groups although there are some differences by VET course level (McMillan Rothman amp Wernert 2005)

In general participation in VET tends to be associated with subsequent higher rates of full-time employment by comparison with those who undertake no post-school study Curtis (2008) reports that of those who had participated in a non-apprenticeship VET course 66 of non-completers and 78 of completers were working full-time The comparable figure for those with no post-school qualifications was 69 For apprenticeships the level of full-time employment is higher at around 93 for completers and 80 for non-completers For traineeships the level of

4 To give an example we find that obtaining a high enough level of qualification can mitigate the effect of becoming unemployed but these findings and others will be discussed in the results section

10 Annual transitions between labour market states for young Australians

full-time employment was in the middle at around 82 for completers and 79 for non-completers Completion of an apprenticeship has the strongest positive impact on obtaining full-time work which is not surprising as apprenticeships are themselves considered a form of full-time work

However earlier research has suggested that full-time vocational study is not particularly beneficial in terms of labour market outcomes for at least some types of courses Analyses of the three older cohorts from the Youth in Transition (YIT) cohorts born in 1961 1965 and 1970 the youngest YIT cohort born in 1975 and the 1995 Longitudinal Surveys of Australian Youth Year 9 cohort have not found strong positive effects for VET on labour market outcomes (Long McKenzie amp Sturman 1996 Marks Hillman amp Beavis 2003 McMillan amp Marks 2003 but see Ryan 2002) The exception is apprenticeships which do substantially improve labour market outcomes By contrast according to Long (2004 pp20ndash1) the benefits of TAFE qualifications are at best equivocal

In the year after completing their qualification 25 of TAFE graduates were not in full-time work and were not enrolled in further study For those TAFE graduates not studying (47 of all graduates) some form of marginal attachment or no attachment to the labour force is a more likely outcome in the short term than is full employment(Long 2004 p6)

These findings for vocational educationmdashthat apprenticeships improve employment prospects but that other forms of vocational education in general do not substantially improve labour market outcomesmdashis consistent with other work in Australia conducted by economists (Dockery amp Norris 1996 Nevile amp Saunders 1998) and is similar to international research (Ryan 2001)

More recently similar conclusions were reached by Marks (2006) comparing full-time vocational study in the first year after leaving school with full-time work (full-time vocational study includes study at a TAFE or private institution but excludes apprentices whom are classified as full-time workers) He found that full-time vocational study increases the chances of being in full-time study or part-time work in the fourth year after leaving school It does not however increase the chances of full-time work Full-time vocational study in the first year increases the odds of being in full-time study rather than full-time work three years later about three times for men and twice as much for women It also increases the odds of being in part-time rather than full-time work in the fourth year Among men full-time vocational study in the first year (relative to full-time work) increases the likelihood of unemployment and lsquootherrsquo activities in the fourth year The lack of positive effects for full-time study on full-time work several years later is an important finding

It is not clear whether the differences between the conclusions of the earlier and later studies on the impact of VET reflect differences in emphasis placed on estimation results methodological differences the choice of comparison group or that VET is more useful for employment outcomes now than in the 1990s

The approach taken in this study is different from previous studies in one major respect Most of the previous research can be characterised as taking a lsquowhat did they do then and where are they now approachrsquo that is comparing outcomes for those with a VET qualification with those without a VET qualification This is eminently sensible for studying the outcome for

NCVER 11

individuals in a given year (for example what are the most likely labour market states for individuals given their post-school qualifications six years after leaving school) but it is not very well suited to study labour market dynamics Instead we seek to model year-on-year transitions between labour market states and investigate the role education and training has on the probability of making these transitions

12 Annual transitions between labour market states for young Australians

MethodologyAs individuals are interviewed annually information is available on their labour market status on a year-by-year basis In answering the first research question we therefore use an annual timeframe to evaluate the influence of labour market status in one period on labour market status in the subsequent period Note that this implies that we not only capture persistence in particular labour market states but also all transitions from one labour market state to another The different labour market states defined are full-time permanent employment part-time permanent employment casual employment5 unemployment and not being in the labour force

Many factors influence labour market outcomes and it is common not to have indicators for some of them When such unobserved variables are not controlled for in the analysis their effects may be attributed erroneously to observed variables This is what triggers the third research question on the nature of the observed persistence of labour market states

The labelling of genuine and spurious persistence6 is widely used in studies of labour market dynamics and dates back to Heckman (1978) The idea paraphrased here is that there are potentially two mechanisms at work that will lead to the same empirical observation that people unemployed in the previous period are more likely to be unemployed in the current period compared with individuals who were not7 The first mechanism genuine persistence captures that there is something intrinsic about unemployment that has a causal effect on the probability of being unemployed next period It may for instance act to diminish the job-ready skills of an individual or give an individual a stigma that makes them less likely to be hired by employers

The second mechanism spurious persistence captures the fact that there are underlying non-observed factors that drive the probability of being unemployed now and in the future Because these underlying factors affect the probability of being unemployed in every period it only appears that previous unemployment leads to future unemployment In actual effect being unemployed in both periods is driven by the same underlying (unobserved) process This underlying process is typically labelled lsquounobserved heterogeneityrsquo indicating that different people are unique in their own special way and that this uniqueness is not observable to the researcher who is trying to model the individualrsquos labour market dynamics

5 Includes both full-time and part-time casual employment6 Persistence is often referred to as lsquostate dependencersquo7 Unemployment is chosen here to illustrate the point One can readily insert any other

labour market outcome instead

NCVER 13

The methods used to control for the influence of unobserved variables are described later In controlling for these influences using a random effects specification we make two assumptions namely that the unobserved variables are constant over time and secondly that unobserved characteristics are not related to those that are observed As these assumptions are not only necessary but also contentious they will also be discussed later in the methodology section Further in light of the assumptions we make much of what would typically be regarded as unobserved heterogeneity explicit where possible (for example personality traits and ability) and we limit the analysis to a reasonably short four-year window when individuals have completed their post-school education and training making the assumption that the unobserved heterogeneity does not vary over time more palatable

The dataThe data used for this report are based on responses from a nationally representative sample of the cohort of young people who were in Year 9 in 1995 (the Y95 cohort) This cohort sample forms part of the LSAY program The young people in this sample responded to a printed questionnaire and to reading comprehension and numeracy tests conducted in their schools in 1995 to a mailed questionnaire administered in 1996 and to telephone interviews conducted annually since then

The initial sample included 13 613 respondents from approximately 300 government Catholic and independent schools from all Australian states and territories In the 2001 survey responses were received from 6876 individuals and by 2006 3914 individuals provided useable responses The modal age of respondents in 2006 was 25 years Because attrition was not uniform sample weights were used to reflect the original population characteristics

Defining mutually exclusive labour market statesIn order to study the annual transitions between different labour market states it is necessary to identify mutually exclusive labour market states We use the time-consistent version of the LSAY Y95 cohort data as provided by NCVER8 and create mutually exclusive labour market states based on this (1) full-time permanently employed (2) part-time permanently employed (3) casually employed (4) unemployed and (5) not in the labour force

Before describing the model used for the statistical analysis we discuss those factors that are included as explanatory variables for the labour market states we observe

8 This is the version of LSAY95 that is publicly available at the Australian Social Science Data Archive (ASSDA) subject to approval It contains the original data elements in the previous release of the LSAY95 data but also includes a series of derived variables in relation to labour market status education etc that have been made consistent over all 12 waves of the LSAY Y95

14 Annual transitions between labour market states for young Australians

Factors that can help explain labour market status post-qualificationPrevious period labour market stateThe research questions we seek to answer immediately lead to one set of factors that need to be included as explanatory variables lagged versions of our dependent variable Without the lags we cannot say anything about transitions between labour market states

Details of post-school qualificationsIn addition to the lagged labour market states that are included as explanatory factors we also include details on the highest level of post-school qualifications obtained distinguishing between certificates I or II certificate III certificate IV a certificate but at an unknown level an advanced diplomadiploma (including associate degree) and bachelor degree or higher

Personal characteristics of the respondentA small number of personal characteristics of the respondent are included They are indicators for being male and indicators for being married or being in a de facto relationship Those individuals not born in Australia are classified as having been born in either a mainly English-speaking country or a non-English speaking country Furthermore we use a categorised parental occupational class indicator distinguishing between upper upper-middle lower-middle and lower

Reading and maths abilityStudentsrsquo reading and maths ability was tested in wave 1 of LSAY Y95 that is at age 15 These are expressed as achievement scores on a scale from 0 to 20 Self-rated proficiencies are also available but these are not used in favour of the objectively measured achievement levels

Personality traitsStudents were asked in wave 3 (that is at approximately age 17) to rate themselves on a 1 to 4 scale according to specific personality traits with 1 corresponding to lsquoveryrsquo and 4 relating to lsquonot at allrsquo The traits distinguished were being agreeable open popular intellectual calm hardworking outgoing and confident

The general structure of the modelThe method used in the statistical analysis to model the labour market dynamics is a multinomial logit model (MNL) The inclusion of the respondentsrsquo previous labour market states makes it a dynamic multinomial logit model The role of unobserved heterogeneity is accounted for by the inclusion of random effects An alternative way to interpret random effects is to think of them as the constant terms in the multinomial logit model being random The resulting model with random alternative specific constants is known as a form of the lsquomixed multinomial logitrsquo (MMNL) or

NCVER 15

lsquorandom parameter logitrsquo (RPL) model9 The random parameters in this case are the constants in the model This model has considerable advantages when analysing state dependence in the presence of unobserved heterogeneity and will be briefly discussed here

To discuss the modelrsquos structure and to illustrate why this specification is the best choice we first introduce some notation Let the dependent variable Yit denote the labour market state at time t for individual i Factors that influence the choice for Yit are denoted by Xit and lagged values of the dependent variable Yit-1 The explanatory variables in Xit as outlined previously imply a clear direction of the association between Xit and Yit That is Xit influences Yit but the reverse cannot hold The unobserved heterogeneitymdashin this study captured by an individual specific random effectmdashis denoted by μi

Why a logit specificationWhen it comes to modelling discrete choice variables the two main types of models are the logit and the probit models Usually the inclusion of lagged dependent variables creates an endogeneity problem but not so in a mixed logit framework Train (2003 p118) explains the issue in plain language Using our notation the Yit depends on Xit and the error term εit If thatrsquos the case then Yit-1 depends on Xit-1 and the error term εit-1 In the presence of unobserved heterogeneity the error term εit includes the unobserved heterogeneity component μi This μi component in the error terms makes these error terms correlated over time (Train 2003 p115) When one includes the lagged dependent variable Yt-1 on the right-hand side as an explanatory variable it will thus violate the independence assumption between the right-hand side variables and the error term in the equation Yt = γYt-1 + β Xt + εt This is easy to see when one realises that Yt-1 depends on εt-1 and εt and εt-1 are correlated because both include μi If a probit specification were to be used the error structure used in the estimation would have to be corrected to account for this Standard statistical software packages such as Stata now have routines that make this correction for binary probits that include lagged values of the dependent variable in the presence of unobserved heterogeneity10 However these routines have not yet been extended to multinomial probits

Train (2003 p150) explains why choosing a mixed logit set-up including lagged dependent variables as explanatory variables on the right-hand side does not pose a problem in the presence of unobserved heterogeneity unlike the case of a probit specification The crux of it is that conditional on the unobserved heterogeneity μi the estimation boils down to estimating a standard multinomial logit modelmdashwith a single complication the unobserved heterogeneity μi on which it was conditioned needs to be lsquointegrated outrsquo Fortunately this can be done using standard numerical simulation making the mixed logit approach (in our case multinomial mixed logit due to a multiple of choices) an ideal candidate to model state dependence in the presence of unobserved heterogeneity In the next subsections we provide more detail on how the estimation works

9 Any parameter in a MMNL can be modelled as a random variable not just the constants10Note that when it is assumed that there is no correlation over time in the error terms eg

in the absence of unobserved heterogeneity including lagged dependent variables Yt-1 does not pose a problem

16 Annual transitions between labour market states for young Australians

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 5: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Tables1a Males Results from what-if scenarios based on

parameter estimates ndash Personal characteristics 20

1b Females Results from what-if scenarios based on parameter estimates ndash Personal characteristics 21

2a Males Results from what-if scenarios based on parameter estimates ndash Personality 23

2b Females Results from what-if scenarios based on parameter estimates ndash Personality 24

3a Males Results from what-if scenarios based on parameter estimates ndash Qualifications and past labour market status 27

3b Females Results from what-if scenarios based on parameter estimates ndash Qualifications and past labour market status 28

4a Males Results from what-if scenarios based on parameter estimates ndash Qualifications and (select) past labour market status combined 31

4b Females Results from what-if scenarios based on parameter estimates ndash Qualifications and (select) past labour market status combined 32

A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects 36

A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects 38

A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects 40

A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics 42

A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality 43

NCVER 5

A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status 44

A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined 45

6 Annual transitions between labour market states for young Australians

Executive summaryThis study uses an annual timeframe to evaluate the influence of labour market status in one period on status in the subsequent period Understanding the role of past labour market experiences is important when it is the objective of policy-makers to increase the proportion of time spent by young Australians in desirable labour market states such as full-time work and reduce the time they spend in marginalised activities such as unemployment These concerns are heightened during lean economic times but they never really go away A natural question that may arise especially in a weak labour market for youth is whether promoting casual employment today would lead to more people being employed permanently in the future or would simply result in more people working on a casual basis For such a question to be answered it is necessary to understand the role of previous labour market states and specifically in this study how this role differs by the level of education and qualifications obtained

Four labour market states are considered in this study permanent employment casual employment unemployment and not being in the labour force The analysis used differs from previous research in that it models year-on-year transitions and does not follow the more commonly used approach of taking a group of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later a lsquowhat did they do then and where are they nowrsquo approach This study is therefore a natural complement to this previous research

Using the 1995 cohort of the Longitudinal Surveys of Australian Youth (LSAY) this study finds that there is substantial predictive power in the previous yearrsquos labour market state when modelling the current labour market state In fact the previous yearrsquos state has the largest predictive power of all factors considered including post-school qualifications

Much of the persistence was shown to be due to an underlying process that drives labour market outcomes in all years particularly for casual employment in the case of men and for women being out of the labour market Ignoring this process will result in their effects being passed through the previous labour market status variables thus creating an illusion of severe persistence Controlling for the process shows that although persistence was reduced previous labour market experience has a genuine impact on current labour market status

The study shows that being in full-time permanent employment in the previous periodmdashcompared with the alternative of being out of the labour forcemdashis for women associated with a 194 percentage point increase in

NCVER 7

the probability of being permanently employed in the next period For men the increase is much smaller at 100 percentage points

Much smaller effects are found for factors other than previous labour market states In terms of the level of post-school qualifications higher levels of education are associated with increased probabilities of being permanently employed Although any post-school qualification is better than none the biggest boost is provided by certificate IV and bachelor degree or better for men and bachelor degree or higher for women Post-school qualifications also play a greater role for women in general

In addition to studying the effect of post-school qualifications and previous period labour market state in isolation they are also studied concurrently This addresses the effect of previous period labour market outcome for different levels of post-school qualifications

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men and -147 percentage points for women (off the base prediction of about 85 for both men and women) But when related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (and -273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications Having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

Similarly the effect of the previous period of unemployment on the probability of being employed permanently in the current period ranged from negligible when unemployment is combined with having a bachelor degree or better to -69 percentage points for men in the worst case when unemployment is combined with having no post-school qualifications and -146 percentage points for women (off the base prediction of about 85 permanently employed for both genders) In other words an event that would lead to all of Australiarsquos youth collectively becoming unemployed would almost be completely offset in terms of impact on next periodrsquos labour market outcome if this event resulted in everyonersquos post-school qualification levels being lifted to a bachelor degree or better or alternatively a certificate IV for women1

Although having much smaller an impact policies focusing on the social skills of young Australians could also improve their labour market outcomes Lifting young Australian womenrsquos confidence to a point where they all considered themselves as being very confident would increase the proportion of young women in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about 1 percentage point extra in casual employment while reducing unemployment and exits from the labour force

1 It is hard to think of an event that would result in all young people losing their job The point made here however is about the comparative strength of the post-school qualification variables

8 Annual transitions between labour market states for young Australians

IntroductionThis study examines the dynamics of the labour market particularly in terms of flows between employment unemployment and non-participation in the labour force Of particular interest is the role of post-school qualifications on these labour market transitions It thus naturally fits in with existing studies on labour market dynamics in general What sets this study apart is that the population of interest is very narrowly defined young Australians between the ages of roughly 22 and 25 who have completed their education and training2 To paraphrase we study labour market dynamics for young Australians lsquoafter the dust has settledrsquo

Objective of the researchThe main objective of this project is to investigate the role of post-school qualifications on the transitions between employment and non-employment Specifically the following three questions are addressed1 What are the transitions between the following labour market states for

young Australians permanent employment casual employment unemployment and not being in the labour force

2 How persistent are the following labour market states which can be classified as less desirable for different levels of post-school qualifications casual employment unemployment and not being in the labour force

3 How much of any observed persistence is lsquogenuinersquo and how much of the observed persistence is lsquospuriousrsquo3

The policy narrative of these three research questions runs from getting a good perspective on how socioeconomic and personal characteristics are correlated with labour market choices and what role previous experiences play (question 1) to investigating the role of education and training in escaping less desirable or bad labour market states (question 2) and finally the mechanism at work behind the persistence in labour market states as observed in the data

2 There has been an increasing awareness and emphasis on lifelong learning to ensure that individuals maintain or acquire the necessary skills to participate in society throughout their lives Having completed education and training as used here means that individuals are not observed to be enrolled in study or training leading to a qualification for the duration of the period that is being modelled that is roughly between the ages of 22 and 25 They may take up such study further into their careerlife

3 The concepts of genuine and spurious persistence will be discussed in more detail in the methodology section but here it is sufficient to state that any persistence observed in the data can have two different mechanisms that would give rise to this persistence

NCVER 9

There exists a body of research on the factors associated with particular labour market outcomes and some of the findings from these research efforts are discussed later However much less is known about the process of switching back and forth between labour market states Understanding these dynamics is crucial in order to develop initiatives that would for instance increase transitions out of unemployment and into permanent employment For instance it is a long-standing desire of successive Australian governments to minimise the amount of time that Australian youths spend in so-called marginal activities In this context knowing what factors are associated with being unemployed is not enough It is also necessary to understand the transition from employment into unemployment and the role education and other personal characteristics play in these transitions4

To frame our research objectives we first briefly discuss Australian evidence with respect to the role of education and training in young peoplersquos post-school outcomes A broader literature on labour market dynamics does exist but is not discussed in detail

Previous studiesParticipation in vocational education and training (VET) is an important pathway in the school-to-work transitions for many young people in Australia VET offers opportunities for young people to develop skills that employers value VET comprises apprenticeships and traineeships (which involve employment) and non-apprenticeship courses offered at publicly funded technical and further (TAFE) institutions and by private VET providers VET is particularly useful for young men who did not complete Year 12 of whom just over a half completes an apprenticeship or traineeship (Curtis amp McMillan 2008)

Apprentices are more likely to be male have a low-to-medium socioeconomic background have a parent in a technical or trade occupation have attended a government school and have relatively lower test scores in reading and mathematics while at school They are also likely to have a father with a trade qualification and have studied VET subjects at school (Ainley amp Corrigan 2005 Curtis 2008) Trainees are more likely to be female less likely to come from a professional background and have relatively high scores in reading and mathematics (Ainley amp Corrigan 2005) Participation in non-apprenticeship VET study is generally evenly distributed across social groups although there are some differences by VET course level (McMillan Rothman amp Wernert 2005)

In general participation in VET tends to be associated with subsequent higher rates of full-time employment by comparison with those who undertake no post-school study Curtis (2008) reports that of those who had participated in a non-apprenticeship VET course 66 of non-completers and 78 of completers were working full-time The comparable figure for those with no post-school qualifications was 69 For apprenticeships the level of full-time employment is higher at around 93 for completers and 80 for non-completers For traineeships the level of

4 To give an example we find that obtaining a high enough level of qualification can mitigate the effect of becoming unemployed but these findings and others will be discussed in the results section

10 Annual transitions between labour market states for young Australians

full-time employment was in the middle at around 82 for completers and 79 for non-completers Completion of an apprenticeship has the strongest positive impact on obtaining full-time work which is not surprising as apprenticeships are themselves considered a form of full-time work

However earlier research has suggested that full-time vocational study is not particularly beneficial in terms of labour market outcomes for at least some types of courses Analyses of the three older cohorts from the Youth in Transition (YIT) cohorts born in 1961 1965 and 1970 the youngest YIT cohort born in 1975 and the 1995 Longitudinal Surveys of Australian Youth Year 9 cohort have not found strong positive effects for VET on labour market outcomes (Long McKenzie amp Sturman 1996 Marks Hillman amp Beavis 2003 McMillan amp Marks 2003 but see Ryan 2002) The exception is apprenticeships which do substantially improve labour market outcomes By contrast according to Long (2004 pp20ndash1) the benefits of TAFE qualifications are at best equivocal

In the year after completing their qualification 25 of TAFE graduates were not in full-time work and were not enrolled in further study For those TAFE graduates not studying (47 of all graduates) some form of marginal attachment or no attachment to the labour force is a more likely outcome in the short term than is full employment(Long 2004 p6)

These findings for vocational educationmdashthat apprenticeships improve employment prospects but that other forms of vocational education in general do not substantially improve labour market outcomesmdashis consistent with other work in Australia conducted by economists (Dockery amp Norris 1996 Nevile amp Saunders 1998) and is similar to international research (Ryan 2001)

More recently similar conclusions were reached by Marks (2006) comparing full-time vocational study in the first year after leaving school with full-time work (full-time vocational study includes study at a TAFE or private institution but excludes apprentices whom are classified as full-time workers) He found that full-time vocational study increases the chances of being in full-time study or part-time work in the fourth year after leaving school It does not however increase the chances of full-time work Full-time vocational study in the first year increases the odds of being in full-time study rather than full-time work three years later about three times for men and twice as much for women It also increases the odds of being in part-time rather than full-time work in the fourth year Among men full-time vocational study in the first year (relative to full-time work) increases the likelihood of unemployment and lsquootherrsquo activities in the fourth year The lack of positive effects for full-time study on full-time work several years later is an important finding

It is not clear whether the differences between the conclusions of the earlier and later studies on the impact of VET reflect differences in emphasis placed on estimation results methodological differences the choice of comparison group or that VET is more useful for employment outcomes now than in the 1990s

The approach taken in this study is different from previous studies in one major respect Most of the previous research can be characterised as taking a lsquowhat did they do then and where are they now approachrsquo that is comparing outcomes for those with a VET qualification with those without a VET qualification This is eminently sensible for studying the outcome for

NCVER 11

individuals in a given year (for example what are the most likely labour market states for individuals given their post-school qualifications six years after leaving school) but it is not very well suited to study labour market dynamics Instead we seek to model year-on-year transitions between labour market states and investigate the role education and training has on the probability of making these transitions

12 Annual transitions between labour market states for young Australians

MethodologyAs individuals are interviewed annually information is available on their labour market status on a year-by-year basis In answering the first research question we therefore use an annual timeframe to evaluate the influence of labour market status in one period on labour market status in the subsequent period Note that this implies that we not only capture persistence in particular labour market states but also all transitions from one labour market state to another The different labour market states defined are full-time permanent employment part-time permanent employment casual employment5 unemployment and not being in the labour force

Many factors influence labour market outcomes and it is common not to have indicators for some of them When such unobserved variables are not controlled for in the analysis their effects may be attributed erroneously to observed variables This is what triggers the third research question on the nature of the observed persistence of labour market states

The labelling of genuine and spurious persistence6 is widely used in studies of labour market dynamics and dates back to Heckman (1978) The idea paraphrased here is that there are potentially two mechanisms at work that will lead to the same empirical observation that people unemployed in the previous period are more likely to be unemployed in the current period compared with individuals who were not7 The first mechanism genuine persistence captures that there is something intrinsic about unemployment that has a causal effect on the probability of being unemployed next period It may for instance act to diminish the job-ready skills of an individual or give an individual a stigma that makes them less likely to be hired by employers

The second mechanism spurious persistence captures the fact that there are underlying non-observed factors that drive the probability of being unemployed now and in the future Because these underlying factors affect the probability of being unemployed in every period it only appears that previous unemployment leads to future unemployment In actual effect being unemployed in both periods is driven by the same underlying (unobserved) process This underlying process is typically labelled lsquounobserved heterogeneityrsquo indicating that different people are unique in their own special way and that this uniqueness is not observable to the researcher who is trying to model the individualrsquos labour market dynamics

5 Includes both full-time and part-time casual employment6 Persistence is often referred to as lsquostate dependencersquo7 Unemployment is chosen here to illustrate the point One can readily insert any other

labour market outcome instead

NCVER 13

The methods used to control for the influence of unobserved variables are described later In controlling for these influences using a random effects specification we make two assumptions namely that the unobserved variables are constant over time and secondly that unobserved characteristics are not related to those that are observed As these assumptions are not only necessary but also contentious they will also be discussed later in the methodology section Further in light of the assumptions we make much of what would typically be regarded as unobserved heterogeneity explicit where possible (for example personality traits and ability) and we limit the analysis to a reasonably short four-year window when individuals have completed their post-school education and training making the assumption that the unobserved heterogeneity does not vary over time more palatable

The dataThe data used for this report are based on responses from a nationally representative sample of the cohort of young people who were in Year 9 in 1995 (the Y95 cohort) This cohort sample forms part of the LSAY program The young people in this sample responded to a printed questionnaire and to reading comprehension and numeracy tests conducted in their schools in 1995 to a mailed questionnaire administered in 1996 and to telephone interviews conducted annually since then

The initial sample included 13 613 respondents from approximately 300 government Catholic and independent schools from all Australian states and territories In the 2001 survey responses were received from 6876 individuals and by 2006 3914 individuals provided useable responses The modal age of respondents in 2006 was 25 years Because attrition was not uniform sample weights were used to reflect the original population characteristics

Defining mutually exclusive labour market statesIn order to study the annual transitions between different labour market states it is necessary to identify mutually exclusive labour market states We use the time-consistent version of the LSAY Y95 cohort data as provided by NCVER8 and create mutually exclusive labour market states based on this (1) full-time permanently employed (2) part-time permanently employed (3) casually employed (4) unemployed and (5) not in the labour force

Before describing the model used for the statistical analysis we discuss those factors that are included as explanatory variables for the labour market states we observe

8 This is the version of LSAY95 that is publicly available at the Australian Social Science Data Archive (ASSDA) subject to approval It contains the original data elements in the previous release of the LSAY95 data but also includes a series of derived variables in relation to labour market status education etc that have been made consistent over all 12 waves of the LSAY Y95

14 Annual transitions between labour market states for young Australians

Factors that can help explain labour market status post-qualificationPrevious period labour market stateThe research questions we seek to answer immediately lead to one set of factors that need to be included as explanatory variables lagged versions of our dependent variable Without the lags we cannot say anything about transitions between labour market states

Details of post-school qualificationsIn addition to the lagged labour market states that are included as explanatory factors we also include details on the highest level of post-school qualifications obtained distinguishing between certificates I or II certificate III certificate IV a certificate but at an unknown level an advanced diplomadiploma (including associate degree) and bachelor degree or higher

Personal characteristics of the respondentA small number of personal characteristics of the respondent are included They are indicators for being male and indicators for being married or being in a de facto relationship Those individuals not born in Australia are classified as having been born in either a mainly English-speaking country or a non-English speaking country Furthermore we use a categorised parental occupational class indicator distinguishing between upper upper-middle lower-middle and lower

Reading and maths abilityStudentsrsquo reading and maths ability was tested in wave 1 of LSAY Y95 that is at age 15 These are expressed as achievement scores on a scale from 0 to 20 Self-rated proficiencies are also available but these are not used in favour of the objectively measured achievement levels

Personality traitsStudents were asked in wave 3 (that is at approximately age 17) to rate themselves on a 1 to 4 scale according to specific personality traits with 1 corresponding to lsquoveryrsquo and 4 relating to lsquonot at allrsquo The traits distinguished were being agreeable open popular intellectual calm hardworking outgoing and confident

The general structure of the modelThe method used in the statistical analysis to model the labour market dynamics is a multinomial logit model (MNL) The inclusion of the respondentsrsquo previous labour market states makes it a dynamic multinomial logit model The role of unobserved heterogeneity is accounted for by the inclusion of random effects An alternative way to interpret random effects is to think of them as the constant terms in the multinomial logit model being random The resulting model with random alternative specific constants is known as a form of the lsquomixed multinomial logitrsquo (MMNL) or

NCVER 15

lsquorandom parameter logitrsquo (RPL) model9 The random parameters in this case are the constants in the model This model has considerable advantages when analysing state dependence in the presence of unobserved heterogeneity and will be briefly discussed here

To discuss the modelrsquos structure and to illustrate why this specification is the best choice we first introduce some notation Let the dependent variable Yit denote the labour market state at time t for individual i Factors that influence the choice for Yit are denoted by Xit and lagged values of the dependent variable Yit-1 The explanatory variables in Xit as outlined previously imply a clear direction of the association between Xit and Yit That is Xit influences Yit but the reverse cannot hold The unobserved heterogeneitymdashin this study captured by an individual specific random effectmdashis denoted by μi

Why a logit specificationWhen it comes to modelling discrete choice variables the two main types of models are the logit and the probit models Usually the inclusion of lagged dependent variables creates an endogeneity problem but not so in a mixed logit framework Train (2003 p118) explains the issue in plain language Using our notation the Yit depends on Xit and the error term εit If thatrsquos the case then Yit-1 depends on Xit-1 and the error term εit-1 In the presence of unobserved heterogeneity the error term εit includes the unobserved heterogeneity component μi This μi component in the error terms makes these error terms correlated over time (Train 2003 p115) When one includes the lagged dependent variable Yt-1 on the right-hand side as an explanatory variable it will thus violate the independence assumption between the right-hand side variables and the error term in the equation Yt = γYt-1 + β Xt + εt This is easy to see when one realises that Yt-1 depends on εt-1 and εt and εt-1 are correlated because both include μi If a probit specification were to be used the error structure used in the estimation would have to be corrected to account for this Standard statistical software packages such as Stata now have routines that make this correction for binary probits that include lagged values of the dependent variable in the presence of unobserved heterogeneity10 However these routines have not yet been extended to multinomial probits

Train (2003 p150) explains why choosing a mixed logit set-up including lagged dependent variables as explanatory variables on the right-hand side does not pose a problem in the presence of unobserved heterogeneity unlike the case of a probit specification The crux of it is that conditional on the unobserved heterogeneity μi the estimation boils down to estimating a standard multinomial logit modelmdashwith a single complication the unobserved heterogeneity μi on which it was conditioned needs to be lsquointegrated outrsquo Fortunately this can be done using standard numerical simulation making the mixed logit approach (in our case multinomial mixed logit due to a multiple of choices) an ideal candidate to model state dependence in the presence of unobserved heterogeneity In the next subsections we provide more detail on how the estimation works

9 Any parameter in a MMNL can be modelled as a random variable not just the constants10Note that when it is assumed that there is no correlation over time in the error terms eg

in the absence of unobserved heterogeneity including lagged dependent variables Yt-1 does not pose a problem

16 Annual transitions between labour market states for young Australians

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 6: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status 44

A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined 45

6 Annual transitions between labour market states for young Australians

Executive summaryThis study uses an annual timeframe to evaluate the influence of labour market status in one period on status in the subsequent period Understanding the role of past labour market experiences is important when it is the objective of policy-makers to increase the proportion of time spent by young Australians in desirable labour market states such as full-time work and reduce the time they spend in marginalised activities such as unemployment These concerns are heightened during lean economic times but they never really go away A natural question that may arise especially in a weak labour market for youth is whether promoting casual employment today would lead to more people being employed permanently in the future or would simply result in more people working on a casual basis For such a question to be answered it is necessary to understand the role of previous labour market states and specifically in this study how this role differs by the level of education and qualifications obtained

Four labour market states are considered in this study permanent employment casual employment unemployment and not being in the labour force The analysis used differs from previous research in that it models year-on-year transitions and does not follow the more commonly used approach of taking a group of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later a lsquowhat did they do then and where are they nowrsquo approach This study is therefore a natural complement to this previous research

Using the 1995 cohort of the Longitudinal Surveys of Australian Youth (LSAY) this study finds that there is substantial predictive power in the previous yearrsquos labour market state when modelling the current labour market state In fact the previous yearrsquos state has the largest predictive power of all factors considered including post-school qualifications

Much of the persistence was shown to be due to an underlying process that drives labour market outcomes in all years particularly for casual employment in the case of men and for women being out of the labour market Ignoring this process will result in their effects being passed through the previous labour market status variables thus creating an illusion of severe persistence Controlling for the process shows that although persistence was reduced previous labour market experience has a genuine impact on current labour market status

The study shows that being in full-time permanent employment in the previous periodmdashcompared with the alternative of being out of the labour forcemdashis for women associated with a 194 percentage point increase in

NCVER 7

the probability of being permanently employed in the next period For men the increase is much smaller at 100 percentage points

Much smaller effects are found for factors other than previous labour market states In terms of the level of post-school qualifications higher levels of education are associated with increased probabilities of being permanently employed Although any post-school qualification is better than none the biggest boost is provided by certificate IV and bachelor degree or better for men and bachelor degree or higher for women Post-school qualifications also play a greater role for women in general

In addition to studying the effect of post-school qualifications and previous period labour market state in isolation they are also studied concurrently This addresses the effect of previous period labour market outcome for different levels of post-school qualifications

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men and -147 percentage points for women (off the base prediction of about 85 for both men and women) But when related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (and -273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications Having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

Similarly the effect of the previous period of unemployment on the probability of being employed permanently in the current period ranged from negligible when unemployment is combined with having a bachelor degree or better to -69 percentage points for men in the worst case when unemployment is combined with having no post-school qualifications and -146 percentage points for women (off the base prediction of about 85 permanently employed for both genders) In other words an event that would lead to all of Australiarsquos youth collectively becoming unemployed would almost be completely offset in terms of impact on next periodrsquos labour market outcome if this event resulted in everyonersquos post-school qualification levels being lifted to a bachelor degree or better or alternatively a certificate IV for women1

Although having much smaller an impact policies focusing on the social skills of young Australians could also improve their labour market outcomes Lifting young Australian womenrsquos confidence to a point where they all considered themselves as being very confident would increase the proportion of young women in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about 1 percentage point extra in casual employment while reducing unemployment and exits from the labour force

1 It is hard to think of an event that would result in all young people losing their job The point made here however is about the comparative strength of the post-school qualification variables

8 Annual transitions between labour market states for young Australians

IntroductionThis study examines the dynamics of the labour market particularly in terms of flows between employment unemployment and non-participation in the labour force Of particular interest is the role of post-school qualifications on these labour market transitions It thus naturally fits in with existing studies on labour market dynamics in general What sets this study apart is that the population of interest is very narrowly defined young Australians between the ages of roughly 22 and 25 who have completed their education and training2 To paraphrase we study labour market dynamics for young Australians lsquoafter the dust has settledrsquo

Objective of the researchThe main objective of this project is to investigate the role of post-school qualifications on the transitions between employment and non-employment Specifically the following three questions are addressed1 What are the transitions between the following labour market states for

young Australians permanent employment casual employment unemployment and not being in the labour force

2 How persistent are the following labour market states which can be classified as less desirable for different levels of post-school qualifications casual employment unemployment and not being in the labour force

3 How much of any observed persistence is lsquogenuinersquo and how much of the observed persistence is lsquospuriousrsquo3

The policy narrative of these three research questions runs from getting a good perspective on how socioeconomic and personal characteristics are correlated with labour market choices and what role previous experiences play (question 1) to investigating the role of education and training in escaping less desirable or bad labour market states (question 2) and finally the mechanism at work behind the persistence in labour market states as observed in the data

2 There has been an increasing awareness and emphasis on lifelong learning to ensure that individuals maintain or acquire the necessary skills to participate in society throughout their lives Having completed education and training as used here means that individuals are not observed to be enrolled in study or training leading to a qualification for the duration of the period that is being modelled that is roughly between the ages of 22 and 25 They may take up such study further into their careerlife

3 The concepts of genuine and spurious persistence will be discussed in more detail in the methodology section but here it is sufficient to state that any persistence observed in the data can have two different mechanisms that would give rise to this persistence

NCVER 9

There exists a body of research on the factors associated with particular labour market outcomes and some of the findings from these research efforts are discussed later However much less is known about the process of switching back and forth between labour market states Understanding these dynamics is crucial in order to develop initiatives that would for instance increase transitions out of unemployment and into permanent employment For instance it is a long-standing desire of successive Australian governments to minimise the amount of time that Australian youths spend in so-called marginal activities In this context knowing what factors are associated with being unemployed is not enough It is also necessary to understand the transition from employment into unemployment and the role education and other personal characteristics play in these transitions4

To frame our research objectives we first briefly discuss Australian evidence with respect to the role of education and training in young peoplersquos post-school outcomes A broader literature on labour market dynamics does exist but is not discussed in detail

Previous studiesParticipation in vocational education and training (VET) is an important pathway in the school-to-work transitions for many young people in Australia VET offers opportunities for young people to develop skills that employers value VET comprises apprenticeships and traineeships (which involve employment) and non-apprenticeship courses offered at publicly funded technical and further (TAFE) institutions and by private VET providers VET is particularly useful for young men who did not complete Year 12 of whom just over a half completes an apprenticeship or traineeship (Curtis amp McMillan 2008)

Apprentices are more likely to be male have a low-to-medium socioeconomic background have a parent in a technical or trade occupation have attended a government school and have relatively lower test scores in reading and mathematics while at school They are also likely to have a father with a trade qualification and have studied VET subjects at school (Ainley amp Corrigan 2005 Curtis 2008) Trainees are more likely to be female less likely to come from a professional background and have relatively high scores in reading and mathematics (Ainley amp Corrigan 2005) Participation in non-apprenticeship VET study is generally evenly distributed across social groups although there are some differences by VET course level (McMillan Rothman amp Wernert 2005)

In general participation in VET tends to be associated with subsequent higher rates of full-time employment by comparison with those who undertake no post-school study Curtis (2008) reports that of those who had participated in a non-apprenticeship VET course 66 of non-completers and 78 of completers were working full-time The comparable figure for those with no post-school qualifications was 69 For apprenticeships the level of full-time employment is higher at around 93 for completers and 80 for non-completers For traineeships the level of

4 To give an example we find that obtaining a high enough level of qualification can mitigate the effect of becoming unemployed but these findings and others will be discussed in the results section

10 Annual transitions between labour market states for young Australians

full-time employment was in the middle at around 82 for completers and 79 for non-completers Completion of an apprenticeship has the strongest positive impact on obtaining full-time work which is not surprising as apprenticeships are themselves considered a form of full-time work

However earlier research has suggested that full-time vocational study is not particularly beneficial in terms of labour market outcomes for at least some types of courses Analyses of the three older cohorts from the Youth in Transition (YIT) cohorts born in 1961 1965 and 1970 the youngest YIT cohort born in 1975 and the 1995 Longitudinal Surveys of Australian Youth Year 9 cohort have not found strong positive effects for VET on labour market outcomes (Long McKenzie amp Sturman 1996 Marks Hillman amp Beavis 2003 McMillan amp Marks 2003 but see Ryan 2002) The exception is apprenticeships which do substantially improve labour market outcomes By contrast according to Long (2004 pp20ndash1) the benefits of TAFE qualifications are at best equivocal

In the year after completing their qualification 25 of TAFE graduates were not in full-time work and were not enrolled in further study For those TAFE graduates not studying (47 of all graduates) some form of marginal attachment or no attachment to the labour force is a more likely outcome in the short term than is full employment(Long 2004 p6)

These findings for vocational educationmdashthat apprenticeships improve employment prospects but that other forms of vocational education in general do not substantially improve labour market outcomesmdashis consistent with other work in Australia conducted by economists (Dockery amp Norris 1996 Nevile amp Saunders 1998) and is similar to international research (Ryan 2001)

More recently similar conclusions were reached by Marks (2006) comparing full-time vocational study in the first year after leaving school with full-time work (full-time vocational study includes study at a TAFE or private institution but excludes apprentices whom are classified as full-time workers) He found that full-time vocational study increases the chances of being in full-time study or part-time work in the fourth year after leaving school It does not however increase the chances of full-time work Full-time vocational study in the first year increases the odds of being in full-time study rather than full-time work three years later about three times for men and twice as much for women It also increases the odds of being in part-time rather than full-time work in the fourth year Among men full-time vocational study in the first year (relative to full-time work) increases the likelihood of unemployment and lsquootherrsquo activities in the fourth year The lack of positive effects for full-time study on full-time work several years later is an important finding

It is not clear whether the differences between the conclusions of the earlier and later studies on the impact of VET reflect differences in emphasis placed on estimation results methodological differences the choice of comparison group or that VET is more useful for employment outcomes now than in the 1990s

The approach taken in this study is different from previous studies in one major respect Most of the previous research can be characterised as taking a lsquowhat did they do then and where are they now approachrsquo that is comparing outcomes for those with a VET qualification with those without a VET qualification This is eminently sensible for studying the outcome for

NCVER 11

individuals in a given year (for example what are the most likely labour market states for individuals given their post-school qualifications six years after leaving school) but it is not very well suited to study labour market dynamics Instead we seek to model year-on-year transitions between labour market states and investigate the role education and training has on the probability of making these transitions

12 Annual transitions between labour market states for young Australians

MethodologyAs individuals are interviewed annually information is available on their labour market status on a year-by-year basis In answering the first research question we therefore use an annual timeframe to evaluate the influence of labour market status in one period on labour market status in the subsequent period Note that this implies that we not only capture persistence in particular labour market states but also all transitions from one labour market state to another The different labour market states defined are full-time permanent employment part-time permanent employment casual employment5 unemployment and not being in the labour force

Many factors influence labour market outcomes and it is common not to have indicators for some of them When such unobserved variables are not controlled for in the analysis their effects may be attributed erroneously to observed variables This is what triggers the third research question on the nature of the observed persistence of labour market states

The labelling of genuine and spurious persistence6 is widely used in studies of labour market dynamics and dates back to Heckman (1978) The idea paraphrased here is that there are potentially two mechanisms at work that will lead to the same empirical observation that people unemployed in the previous period are more likely to be unemployed in the current period compared with individuals who were not7 The first mechanism genuine persistence captures that there is something intrinsic about unemployment that has a causal effect on the probability of being unemployed next period It may for instance act to diminish the job-ready skills of an individual or give an individual a stigma that makes them less likely to be hired by employers

The second mechanism spurious persistence captures the fact that there are underlying non-observed factors that drive the probability of being unemployed now and in the future Because these underlying factors affect the probability of being unemployed in every period it only appears that previous unemployment leads to future unemployment In actual effect being unemployed in both periods is driven by the same underlying (unobserved) process This underlying process is typically labelled lsquounobserved heterogeneityrsquo indicating that different people are unique in their own special way and that this uniqueness is not observable to the researcher who is trying to model the individualrsquos labour market dynamics

5 Includes both full-time and part-time casual employment6 Persistence is often referred to as lsquostate dependencersquo7 Unemployment is chosen here to illustrate the point One can readily insert any other

labour market outcome instead

NCVER 13

The methods used to control for the influence of unobserved variables are described later In controlling for these influences using a random effects specification we make two assumptions namely that the unobserved variables are constant over time and secondly that unobserved characteristics are not related to those that are observed As these assumptions are not only necessary but also contentious they will also be discussed later in the methodology section Further in light of the assumptions we make much of what would typically be regarded as unobserved heterogeneity explicit where possible (for example personality traits and ability) and we limit the analysis to a reasonably short four-year window when individuals have completed their post-school education and training making the assumption that the unobserved heterogeneity does not vary over time more palatable

The dataThe data used for this report are based on responses from a nationally representative sample of the cohort of young people who were in Year 9 in 1995 (the Y95 cohort) This cohort sample forms part of the LSAY program The young people in this sample responded to a printed questionnaire and to reading comprehension and numeracy tests conducted in their schools in 1995 to a mailed questionnaire administered in 1996 and to telephone interviews conducted annually since then

The initial sample included 13 613 respondents from approximately 300 government Catholic and independent schools from all Australian states and territories In the 2001 survey responses were received from 6876 individuals and by 2006 3914 individuals provided useable responses The modal age of respondents in 2006 was 25 years Because attrition was not uniform sample weights were used to reflect the original population characteristics

Defining mutually exclusive labour market statesIn order to study the annual transitions between different labour market states it is necessary to identify mutually exclusive labour market states We use the time-consistent version of the LSAY Y95 cohort data as provided by NCVER8 and create mutually exclusive labour market states based on this (1) full-time permanently employed (2) part-time permanently employed (3) casually employed (4) unemployed and (5) not in the labour force

Before describing the model used for the statistical analysis we discuss those factors that are included as explanatory variables for the labour market states we observe

8 This is the version of LSAY95 that is publicly available at the Australian Social Science Data Archive (ASSDA) subject to approval It contains the original data elements in the previous release of the LSAY95 data but also includes a series of derived variables in relation to labour market status education etc that have been made consistent over all 12 waves of the LSAY Y95

14 Annual transitions between labour market states for young Australians

Factors that can help explain labour market status post-qualificationPrevious period labour market stateThe research questions we seek to answer immediately lead to one set of factors that need to be included as explanatory variables lagged versions of our dependent variable Without the lags we cannot say anything about transitions between labour market states

Details of post-school qualificationsIn addition to the lagged labour market states that are included as explanatory factors we also include details on the highest level of post-school qualifications obtained distinguishing between certificates I or II certificate III certificate IV a certificate but at an unknown level an advanced diplomadiploma (including associate degree) and bachelor degree or higher

Personal characteristics of the respondentA small number of personal characteristics of the respondent are included They are indicators for being male and indicators for being married or being in a de facto relationship Those individuals not born in Australia are classified as having been born in either a mainly English-speaking country or a non-English speaking country Furthermore we use a categorised parental occupational class indicator distinguishing between upper upper-middle lower-middle and lower

Reading and maths abilityStudentsrsquo reading and maths ability was tested in wave 1 of LSAY Y95 that is at age 15 These are expressed as achievement scores on a scale from 0 to 20 Self-rated proficiencies are also available but these are not used in favour of the objectively measured achievement levels

Personality traitsStudents were asked in wave 3 (that is at approximately age 17) to rate themselves on a 1 to 4 scale according to specific personality traits with 1 corresponding to lsquoveryrsquo and 4 relating to lsquonot at allrsquo The traits distinguished were being agreeable open popular intellectual calm hardworking outgoing and confident

The general structure of the modelThe method used in the statistical analysis to model the labour market dynamics is a multinomial logit model (MNL) The inclusion of the respondentsrsquo previous labour market states makes it a dynamic multinomial logit model The role of unobserved heterogeneity is accounted for by the inclusion of random effects An alternative way to interpret random effects is to think of them as the constant terms in the multinomial logit model being random The resulting model with random alternative specific constants is known as a form of the lsquomixed multinomial logitrsquo (MMNL) or

NCVER 15

lsquorandom parameter logitrsquo (RPL) model9 The random parameters in this case are the constants in the model This model has considerable advantages when analysing state dependence in the presence of unobserved heterogeneity and will be briefly discussed here

To discuss the modelrsquos structure and to illustrate why this specification is the best choice we first introduce some notation Let the dependent variable Yit denote the labour market state at time t for individual i Factors that influence the choice for Yit are denoted by Xit and lagged values of the dependent variable Yit-1 The explanatory variables in Xit as outlined previously imply a clear direction of the association between Xit and Yit That is Xit influences Yit but the reverse cannot hold The unobserved heterogeneitymdashin this study captured by an individual specific random effectmdashis denoted by μi

Why a logit specificationWhen it comes to modelling discrete choice variables the two main types of models are the logit and the probit models Usually the inclusion of lagged dependent variables creates an endogeneity problem but not so in a mixed logit framework Train (2003 p118) explains the issue in plain language Using our notation the Yit depends on Xit and the error term εit If thatrsquos the case then Yit-1 depends on Xit-1 and the error term εit-1 In the presence of unobserved heterogeneity the error term εit includes the unobserved heterogeneity component μi This μi component in the error terms makes these error terms correlated over time (Train 2003 p115) When one includes the lagged dependent variable Yt-1 on the right-hand side as an explanatory variable it will thus violate the independence assumption between the right-hand side variables and the error term in the equation Yt = γYt-1 + β Xt + εt This is easy to see when one realises that Yt-1 depends on εt-1 and εt and εt-1 are correlated because both include μi If a probit specification were to be used the error structure used in the estimation would have to be corrected to account for this Standard statistical software packages such as Stata now have routines that make this correction for binary probits that include lagged values of the dependent variable in the presence of unobserved heterogeneity10 However these routines have not yet been extended to multinomial probits

Train (2003 p150) explains why choosing a mixed logit set-up including lagged dependent variables as explanatory variables on the right-hand side does not pose a problem in the presence of unobserved heterogeneity unlike the case of a probit specification The crux of it is that conditional on the unobserved heterogeneity μi the estimation boils down to estimating a standard multinomial logit modelmdashwith a single complication the unobserved heterogeneity μi on which it was conditioned needs to be lsquointegrated outrsquo Fortunately this can be done using standard numerical simulation making the mixed logit approach (in our case multinomial mixed logit due to a multiple of choices) an ideal candidate to model state dependence in the presence of unobserved heterogeneity In the next subsections we provide more detail on how the estimation works

9 Any parameter in a MMNL can be modelled as a random variable not just the constants10Note that when it is assumed that there is no correlation over time in the error terms eg

in the absence of unobserved heterogeneity including lagged dependent variables Yt-1 does not pose a problem

16 Annual transitions between labour market states for young Australians

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 7: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Executive summaryThis study uses an annual timeframe to evaluate the influence of labour market status in one period on status in the subsequent period Understanding the role of past labour market experiences is important when it is the objective of policy-makers to increase the proportion of time spent by young Australians in desirable labour market states such as full-time work and reduce the time they spend in marginalised activities such as unemployment These concerns are heightened during lean economic times but they never really go away A natural question that may arise especially in a weak labour market for youth is whether promoting casual employment today would lead to more people being employed permanently in the future or would simply result in more people working on a casual basis For such a question to be answered it is necessary to understand the role of previous labour market states and specifically in this study how this role differs by the level of education and qualifications obtained

Four labour market states are considered in this study permanent employment casual employment unemployment and not being in the labour force The analysis used differs from previous research in that it models year-on-year transitions and does not follow the more commonly used approach of taking a group of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later a lsquowhat did they do then and where are they nowrsquo approach This study is therefore a natural complement to this previous research

Using the 1995 cohort of the Longitudinal Surveys of Australian Youth (LSAY) this study finds that there is substantial predictive power in the previous yearrsquos labour market state when modelling the current labour market state In fact the previous yearrsquos state has the largest predictive power of all factors considered including post-school qualifications

Much of the persistence was shown to be due to an underlying process that drives labour market outcomes in all years particularly for casual employment in the case of men and for women being out of the labour market Ignoring this process will result in their effects being passed through the previous labour market status variables thus creating an illusion of severe persistence Controlling for the process shows that although persistence was reduced previous labour market experience has a genuine impact on current labour market status

The study shows that being in full-time permanent employment in the previous periodmdashcompared with the alternative of being out of the labour forcemdashis for women associated with a 194 percentage point increase in

NCVER 7

the probability of being permanently employed in the next period For men the increase is much smaller at 100 percentage points

Much smaller effects are found for factors other than previous labour market states In terms of the level of post-school qualifications higher levels of education are associated with increased probabilities of being permanently employed Although any post-school qualification is better than none the biggest boost is provided by certificate IV and bachelor degree or better for men and bachelor degree or higher for women Post-school qualifications also play a greater role for women in general

In addition to studying the effect of post-school qualifications and previous period labour market state in isolation they are also studied concurrently This addresses the effect of previous period labour market outcome for different levels of post-school qualifications

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men and -147 percentage points for women (off the base prediction of about 85 for both men and women) But when related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (and -273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications Having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

Similarly the effect of the previous period of unemployment on the probability of being employed permanently in the current period ranged from negligible when unemployment is combined with having a bachelor degree or better to -69 percentage points for men in the worst case when unemployment is combined with having no post-school qualifications and -146 percentage points for women (off the base prediction of about 85 permanently employed for both genders) In other words an event that would lead to all of Australiarsquos youth collectively becoming unemployed would almost be completely offset in terms of impact on next periodrsquos labour market outcome if this event resulted in everyonersquos post-school qualification levels being lifted to a bachelor degree or better or alternatively a certificate IV for women1

Although having much smaller an impact policies focusing on the social skills of young Australians could also improve their labour market outcomes Lifting young Australian womenrsquos confidence to a point where they all considered themselves as being very confident would increase the proportion of young women in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about 1 percentage point extra in casual employment while reducing unemployment and exits from the labour force

1 It is hard to think of an event that would result in all young people losing their job The point made here however is about the comparative strength of the post-school qualification variables

8 Annual transitions between labour market states for young Australians

IntroductionThis study examines the dynamics of the labour market particularly in terms of flows between employment unemployment and non-participation in the labour force Of particular interest is the role of post-school qualifications on these labour market transitions It thus naturally fits in with existing studies on labour market dynamics in general What sets this study apart is that the population of interest is very narrowly defined young Australians between the ages of roughly 22 and 25 who have completed their education and training2 To paraphrase we study labour market dynamics for young Australians lsquoafter the dust has settledrsquo

Objective of the researchThe main objective of this project is to investigate the role of post-school qualifications on the transitions between employment and non-employment Specifically the following three questions are addressed1 What are the transitions between the following labour market states for

young Australians permanent employment casual employment unemployment and not being in the labour force

2 How persistent are the following labour market states which can be classified as less desirable for different levels of post-school qualifications casual employment unemployment and not being in the labour force

3 How much of any observed persistence is lsquogenuinersquo and how much of the observed persistence is lsquospuriousrsquo3

The policy narrative of these three research questions runs from getting a good perspective on how socioeconomic and personal characteristics are correlated with labour market choices and what role previous experiences play (question 1) to investigating the role of education and training in escaping less desirable or bad labour market states (question 2) and finally the mechanism at work behind the persistence in labour market states as observed in the data

2 There has been an increasing awareness and emphasis on lifelong learning to ensure that individuals maintain or acquire the necessary skills to participate in society throughout their lives Having completed education and training as used here means that individuals are not observed to be enrolled in study or training leading to a qualification for the duration of the period that is being modelled that is roughly between the ages of 22 and 25 They may take up such study further into their careerlife

3 The concepts of genuine and spurious persistence will be discussed in more detail in the methodology section but here it is sufficient to state that any persistence observed in the data can have two different mechanisms that would give rise to this persistence

NCVER 9

There exists a body of research on the factors associated with particular labour market outcomes and some of the findings from these research efforts are discussed later However much less is known about the process of switching back and forth between labour market states Understanding these dynamics is crucial in order to develop initiatives that would for instance increase transitions out of unemployment and into permanent employment For instance it is a long-standing desire of successive Australian governments to minimise the amount of time that Australian youths spend in so-called marginal activities In this context knowing what factors are associated with being unemployed is not enough It is also necessary to understand the transition from employment into unemployment and the role education and other personal characteristics play in these transitions4

To frame our research objectives we first briefly discuss Australian evidence with respect to the role of education and training in young peoplersquos post-school outcomes A broader literature on labour market dynamics does exist but is not discussed in detail

Previous studiesParticipation in vocational education and training (VET) is an important pathway in the school-to-work transitions for many young people in Australia VET offers opportunities for young people to develop skills that employers value VET comprises apprenticeships and traineeships (which involve employment) and non-apprenticeship courses offered at publicly funded technical and further (TAFE) institutions and by private VET providers VET is particularly useful for young men who did not complete Year 12 of whom just over a half completes an apprenticeship or traineeship (Curtis amp McMillan 2008)

Apprentices are more likely to be male have a low-to-medium socioeconomic background have a parent in a technical or trade occupation have attended a government school and have relatively lower test scores in reading and mathematics while at school They are also likely to have a father with a trade qualification and have studied VET subjects at school (Ainley amp Corrigan 2005 Curtis 2008) Trainees are more likely to be female less likely to come from a professional background and have relatively high scores in reading and mathematics (Ainley amp Corrigan 2005) Participation in non-apprenticeship VET study is generally evenly distributed across social groups although there are some differences by VET course level (McMillan Rothman amp Wernert 2005)

In general participation in VET tends to be associated with subsequent higher rates of full-time employment by comparison with those who undertake no post-school study Curtis (2008) reports that of those who had participated in a non-apprenticeship VET course 66 of non-completers and 78 of completers were working full-time The comparable figure for those with no post-school qualifications was 69 For apprenticeships the level of full-time employment is higher at around 93 for completers and 80 for non-completers For traineeships the level of

4 To give an example we find that obtaining a high enough level of qualification can mitigate the effect of becoming unemployed but these findings and others will be discussed in the results section

10 Annual transitions between labour market states for young Australians

full-time employment was in the middle at around 82 for completers and 79 for non-completers Completion of an apprenticeship has the strongest positive impact on obtaining full-time work which is not surprising as apprenticeships are themselves considered a form of full-time work

However earlier research has suggested that full-time vocational study is not particularly beneficial in terms of labour market outcomes for at least some types of courses Analyses of the three older cohorts from the Youth in Transition (YIT) cohorts born in 1961 1965 and 1970 the youngest YIT cohort born in 1975 and the 1995 Longitudinal Surveys of Australian Youth Year 9 cohort have not found strong positive effects for VET on labour market outcomes (Long McKenzie amp Sturman 1996 Marks Hillman amp Beavis 2003 McMillan amp Marks 2003 but see Ryan 2002) The exception is apprenticeships which do substantially improve labour market outcomes By contrast according to Long (2004 pp20ndash1) the benefits of TAFE qualifications are at best equivocal

In the year after completing their qualification 25 of TAFE graduates were not in full-time work and were not enrolled in further study For those TAFE graduates not studying (47 of all graduates) some form of marginal attachment or no attachment to the labour force is a more likely outcome in the short term than is full employment(Long 2004 p6)

These findings for vocational educationmdashthat apprenticeships improve employment prospects but that other forms of vocational education in general do not substantially improve labour market outcomesmdashis consistent with other work in Australia conducted by economists (Dockery amp Norris 1996 Nevile amp Saunders 1998) and is similar to international research (Ryan 2001)

More recently similar conclusions were reached by Marks (2006) comparing full-time vocational study in the first year after leaving school with full-time work (full-time vocational study includes study at a TAFE or private institution but excludes apprentices whom are classified as full-time workers) He found that full-time vocational study increases the chances of being in full-time study or part-time work in the fourth year after leaving school It does not however increase the chances of full-time work Full-time vocational study in the first year increases the odds of being in full-time study rather than full-time work three years later about three times for men and twice as much for women It also increases the odds of being in part-time rather than full-time work in the fourth year Among men full-time vocational study in the first year (relative to full-time work) increases the likelihood of unemployment and lsquootherrsquo activities in the fourth year The lack of positive effects for full-time study on full-time work several years later is an important finding

It is not clear whether the differences between the conclusions of the earlier and later studies on the impact of VET reflect differences in emphasis placed on estimation results methodological differences the choice of comparison group or that VET is more useful for employment outcomes now than in the 1990s

The approach taken in this study is different from previous studies in one major respect Most of the previous research can be characterised as taking a lsquowhat did they do then and where are they now approachrsquo that is comparing outcomes for those with a VET qualification with those without a VET qualification This is eminently sensible for studying the outcome for

NCVER 11

individuals in a given year (for example what are the most likely labour market states for individuals given their post-school qualifications six years after leaving school) but it is not very well suited to study labour market dynamics Instead we seek to model year-on-year transitions between labour market states and investigate the role education and training has on the probability of making these transitions

12 Annual transitions between labour market states for young Australians

MethodologyAs individuals are interviewed annually information is available on their labour market status on a year-by-year basis In answering the first research question we therefore use an annual timeframe to evaluate the influence of labour market status in one period on labour market status in the subsequent period Note that this implies that we not only capture persistence in particular labour market states but also all transitions from one labour market state to another The different labour market states defined are full-time permanent employment part-time permanent employment casual employment5 unemployment and not being in the labour force

Many factors influence labour market outcomes and it is common not to have indicators for some of them When such unobserved variables are not controlled for in the analysis their effects may be attributed erroneously to observed variables This is what triggers the third research question on the nature of the observed persistence of labour market states

The labelling of genuine and spurious persistence6 is widely used in studies of labour market dynamics and dates back to Heckman (1978) The idea paraphrased here is that there are potentially two mechanisms at work that will lead to the same empirical observation that people unemployed in the previous period are more likely to be unemployed in the current period compared with individuals who were not7 The first mechanism genuine persistence captures that there is something intrinsic about unemployment that has a causal effect on the probability of being unemployed next period It may for instance act to diminish the job-ready skills of an individual or give an individual a stigma that makes them less likely to be hired by employers

The second mechanism spurious persistence captures the fact that there are underlying non-observed factors that drive the probability of being unemployed now and in the future Because these underlying factors affect the probability of being unemployed in every period it only appears that previous unemployment leads to future unemployment In actual effect being unemployed in both periods is driven by the same underlying (unobserved) process This underlying process is typically labelled lsquounobserved heterogeneityrsquo indicating that different people are unique in their own special way and that this uniqueness is not observable to the researcher who is trying to model the individualrsquos labour market dynamics

5 Includes both full-time and part-time casual employment6 Persistence is often referred to as lsquostate dependencersquo7 Unemployment is chosen here to illustrate the point One can readily insert any other

labour market outcome instead

NCVER 13

The methods used to control for the influence of unobserved variables are described later In controlling for these influences using a random effects specification we make two assumptions namely that the unobserved variables are constant over time and secondly that unobserved characteristics are not related to those that are observed As these assumptions are not only necessary but also contentious they will also be discussed later in the methodology section Further in light of the assumptions we make much of what would typically be regarded as unobserved heterogeneity explicit where possible (for example personality traits and ability) and we limit the analysis to a reasonably short four-year window when individuals have completed their post-school education and training making the assumption that the unobserved heterogeneity does not vary over time more palatable

The dataThe data used for this report are based on responses from a nationally representative sample of the cohort of young people who were in Year 9 in 1995 (the Y95 cohort) This cohort sample forms part of the LSAY program The young people in this sample responded to a printed questionnaire and to reading comprehension and numeracy tests conducted in their schools in 1995 to a mailed questionnaire administered in 1996 and to telephone interviews conducted annually since then

The initial sample included 13 613 respondents from approximately 300 government Catholic and independent schools from all Australian states and territories In the 2001 survey responses were received from 6876 individuals and by 2006 3914 individuals provided useable responses The modal age of respondents in 2006 was 25 years Because attrition was not uniform sample weights were used to reflect the original population characteristics

Defining mutually exclusive labour market statesIn order to study the annual transitions between different labour market states it is necessary to identify mutually exclusive labour market states We use the time-consistent version of the LSAY Y95 cohort data as provided by NCVER8 and create mutually exclusive labour market states based on this (1) full-time permanently employed (2) part-time permanently employed (3) casually employed (4) unemployed and (5) not in the labour force

Before describing the model used for the statistical analysis we discuss those factors that are included as explanatory variables for the labour market states we observe

8 This is the version of LSAY95 that is publicly available at the Australian Social Science Data Archive (ASSDA) subject to approval It contains the original data elements in the previous release of the LSAY95 data but also includes a series of derived variables in relation to labour market status education etc that have been made consistent over all 12 waves of the LSAY Y95

14 Annual transitions between labour market states for young Australians

Factors that can help explain labour market status post-qualificationPrevious period labour market stateThe research questions we seek to answer immediately lead to one set of factors that need to be included as explanatory variables lagged versions of our dependent variable Without the lags we cannot say anything about transitions between labour market states

Details of post-school qualificationsIn addition to the lagged labour market states that are included as explanatory factors we also include details on the highest level of post-school qualifications obtained distinguishing between certificates I or II certificate III certificate IV a certificate but at an unknown level an advanced diplomadiploma (including associate degree) and bachelor degree or higher

Personal characteristics of the respondentA small number of personal characteristics of the respondent are included They are indicators for being male and indicators for being married or being in a de facto relationship Those individuals not born in Australia are classified as having been born in either a mainly English-speaking country or a non-English speaking country Furthermore we use a categorised parental occupational class indicator distinguishing between upper upper-middle lower-middle and lower

Reading and maths abilityStudentsrsquo reading and maths ability was tested in wave 1 of LSAY Y95 that is at age 15 These are expressed as achievement scores on a scale from 0 to 20 Self-rated proficiencies are also available but these are not used in favour of the objectively measured achievement levels

Personality traitsStudents were asked in wave 3 (that is at approximately age 17) to rate themselves on a 1 to 4 scale according to specific personality traits with 1 corresponding to lsquoveryrsquo and 4 relating to lsquonot at allrsquo The traits distinguished were being agreeable open popular intellectual calm hardworking outgoing and confident

The general structure of the modelThe method used in the statistical analysis to model the labour market dynamics is a multinomial logit model (MNL) The inclusion of the respondentsrsquo previous labour market states makes it a dynamic multinomial logit model The role of unobserved heterogeneity is accounted for by the inclusion of random effects An alternative way to interpret random effects is to think of them as the constant terms in the multinomial logit model being random The resulting model with random alternative specific constants is known as a form of the lsquomixed multinomial logitrsquo (MMNL) or

NCVER 15

lsquorandom parameter logitrsquo (RPL) model9 The random parameters in this case are the constants in the model This model has considerable advantages when analysing state dependence in the presence of unobserved heterogeneity and will be briefly discussed here

To discuss the modelrsquos structure and to illustrate why this specification is the best choice we first introduce some notation Let the dependent variable Yit denote the labour market state at time t for individual i Factors that influence the choice for Yit are denoted by Xit and lagged values of the dependent variable Yit-1 The explanatory variables in Xit as outlined previously imply a clear direction of the association between Xit and Yit That is Xit influences Yit but the reverse cannot hold The unobserved heterogeneitymdashin this study captured by an individual specific random effectmdashis denoted by μi

Why a logit specificationWhen it comes to modelling discrete choice variables the two main types of models are the logit and the probit models Usually the inclusion of lagged dependent variables creates an endogeneity problem but not so in a mixed logit framework Train (2003 p118) explains the issue in plain language Using our notation the Yit depends on Xit and the error term εit If thatrsquos the case then Yit-1 depends on Xit-1 and the error term εit-1 In the presence of unobserved heterogeneity the error term εit includes the unobserved heterogeneity component μi This μi component in the error terms makes these error terms correlated over time (Train 2003 p115) When one includes the lagged dependent variable Yt-1 on the right-hand side as an explanatory variable it will thus violate the independence assumption between the right-hand side variables and the error term in the equation Yt = γYt-1 + β Xt + εt This is easy to see when one realises that Yt-1 depends on εt-1 and εt and εt-1 are correlated because both include μi If a probit specification were to be used the error structure used in the estimation would have to be corrected to account for this Standard statistical software packages such as Stata now have routines that make this correction for binary probits that include lagged values of the dependent variable in the presence of unobserved heterogeneity10 However these routines have not yet been extended to multinomial probits

Train (2003 p150) explains why choosing a mixed logit set-up including lagged dependent variables as explanatory variables on the right-hand side does not pose a problem in the presence of unobserved heterogeneity unlike the case of a probit specification The crux of it is that conditional on the unobserved heterogeneity μi the estimation boils down to estimating a standard multinomial logit modelmdashwith a single complication the unobserved heterogeneity μi on which it was conditioned needs to be lsquointegrated outrsquo Fortunately this can be done using standard numerical simulation making the mixed logit approach (in our case multinomial mixed logit due to a multiple of choices) an ideal candidate to model state dependence in the presence of unobserved heterogeneity In the next subsections we provide more detail on how the estimation works

9 Any parameter in a MMNL can be modelled as a random variable not just the constants10Note that when it is assumed that there is no correlation over time in the error terms eg

in the absence of unobserved heterogeneity including lagged dependent variables Yt-1 does not pose a problem

16 Annual transitions between labour market states for young Australians

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 8: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

the probability of being permanently employed in the next period For men the increase is much smaller at 100 percentage points

Much smaller effects are found for factors other than previous labour market states In terms of the level of post-school qualifications higher levels of education are associated with increased probabilities of being permanently employed Although any post-school qualification is better than none the biggest boost is provided by certificate IV and bachelor degree or better for men and bachelor degree or higher for women Post-school qualifications also play a greater role for women in general

In addition to studying the effect of post-school qualifications and previous period labour market state in isolation they are also studied concurrently This addresses the effect of previous period labour market outcome for different levels of post-school qualifications

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men and -147 percentage points for women (off the base prediction of about 85 for both men and women) But when related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (and -273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications Having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

Similarly the effect of the previous period of unemployment on the probability of being employed permanently in the current period ranged from negligible when unemployment is combined with having a bachelor degree or better to -69 percentage points for men in the worst case when unemployment is combined with having no post-school qualifications and -146 percentage points for women (off the base prediction of about 85 permanently employed for both genders) In other words an event that would lead to all of Australiarsquos youth collectively becoming unemployed would almost be completely offset in terms of impact on next periodrsquos labour market outcome if this event resulted in everyonersquos post-school qualification levels being lifted to a bachelor degree or better or alternatively a certificate IV for women1

Although having much smaller an impact policies focusing on the social skills of young Australians could also improve their labour market outcomes Lifting young Australian womenrsquos confidence to a point where they all considered themselves as being very confident would increase the proportion of young women in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about 1 percentage point extra in casual employment while reducing unemployment and exits from the labour force

1 It is hard to think of an event that would result in all young people losing their job The point made here however is about the comparative strength of the post-school qualification variables

8 Annual transitions between labour market states for young Australians

IntroductionThis study examines the dynamics of the labour market particularly in terms of flows between employment unemployment and non-participation in the labour force Of particular interest is the role of post-school qualifications on these labour market transitions It thus naturally fits in with existing studies on labour market dynamics in general What sets this study apart is that the population of interest is very narrowly defined young Australians between the ages of roughly 22 and 25 who have completed their education and training2 To paraphrase we study labour market dynamics for young Australians lsquoafter the dust has settledrsquo

Objective of the researchThe main objective of this project is to investigate the role of post-school qualifications on the transitions between employment and non-employment Specifically the following three questions are addressed1 What are the transitions between the following labour market states for

young Australians permanent employment casual employment unemployment and not being in the labour force

2 How persistent are the following labour market states which can be classified as less desirable for different levels of post-school qualifications casual employment unemployment and not being in the labour force

3 How much of any observed persistence is lsquogenuinersquo and how much of the observed persistence is lsquospuriousrsquo3

The policy narrative of these three research questions runs from getting a good perspective on how socioeconomic and personal characteristics are correlated with labour market choices and what role previous experiences play (question 1) to investigating the role of education and training in escaping less desirable or bad labour market states (question 2) and finally the mechanism at work behind the persistence in labour market states as observed in the data

2 There has been an increasing awareness and emphasis on lifelong learning to ensure that individuals maintain or acquire the necessary skills to participate in society throughout their lives Having completed education and training as used here means that individuals are not observed to be enrolled in study or training leading to a qualification for the duration of the period that is being modelled that is roughly between the ages of 22 and 25 They may take up such study further into their careerlife

3 The concepts of genuine and spurious persistence will be discussed in more detail in the methodology section but here it is sufficient to state that any persistence observed in the data can have two different mechanisms that would give rise to this persistence

NCVER 9

There exists a body of research on the factors associated with particular labour market outcomes and some of the findings from these research efforts are discussed later However much less is known about the process of switching back and forth between labour market states Understanding these dynamics is crucial in order to develop initiatives that would for instance increase transitions out of unemployment and into permanent employment For instance it is a long-standing desire of successive Australian governments to minimise the amount of time that Australian youths spend in so-called marginal activities In this context knowing what factors are associated with being unemployed is not enough It is also necessary to understand the transition from employment into unemployment and the role education and other personal characteristics play in these transitions4

To frame our research objectives we first briefly discuss Australian evidence with respect to the role of education and training in young peoplersquos post-school outcomes A broader literature on labour market dynamics does exist but is not discussed in detail

Previous studiesParticipation in vocational education and training (VET) is an important pathway in the school-to-work transitions for many young people in Australia VET offers opportunities for young people to develop skills that employers value VET comprises apprenticeships and traineeships (which involve employment) and non-apprenticeship courses offered at publicly funded technical and further (TAFE) institutions and by private VET providers VET is particularly useful for young men who did not complete Year 12 of whom just over a half completes an apprenticeship or traineeship (Curtis amp McMillan 2008)

Apprentices are more likely to be male have a low-to-medium socioeconomic background have a parent in a technical or trade occupation have attended a government school and have relatively lower test scores in reading and mathematics while at school They are also likely to have a father with a trade qualification and have studied VET subjects at school (Ainley amp Corrigan 2005 Curtis 2008) Trainees are more likely to be female less likely to come from a professional background and have relatively high scores in reading and mathematics (Ainley amp Corrigan 2005) Participation in non-apprenticeship VET study is generally evenly distributed across social groups although there are some differences by VET course level (McMillan Rothman amp Wernert 2005)

In general participation in VET tends to be associated with subsequent higher rates of full-time employment by comparison with those who undertake no post-school study Curtis (2008) reports that of those who had participated in a non-apprenticeship VET course 66 of non-completers and 78 of completers were working full-time The comparable figure for those with no post-school qualifications was 69 For apprenticeships the level of full-time employment is higher at around 93 for completers and 80 for non-completers For traineeships the level of

4 To give an example we find that obtaining a high enough level of qualification can mitigate the effect of becoming unemployed but these findings and others will be discussed in the results section

10 Annual transitions between labour market states for young Australians

full-time employment was in the middle at around 82 for completers and 79 for non-completers Completion of an apprenticeship has the strongest positive impact on obtaining full-time work which is not surprising as apprenticeships are themselves considered a form of full-time work

However earlier research has suggested that full-time vocational study is not particularly beneficial in terms of labour market outcomes for at least some types of courses Analyses of the three older cohorts from the Youth in Transition (YIT) cohorts born in 1961 1965 and 1970 the youngest YIT cohort born in 1975 and the 1995 Longitudinal Surveys of Australian Youth Year 9 cohort have not found strong positive effects for VET on labour market outcomes (Long McKenzie amp Sturman 1996 Marks Hillman amp Beavis 2003 McMillan amp Marks 2003 but see Ryan 2002) The exception is apprenticeships which do substantially improve labour market outcomes By contrast according to Long (2004 pp20ndash1) the benefits of TAFE qualifications are at best equivocal

In the year after completing their qualification 25 of TAFE graduates were not in full-time work and were not enrolled in further study For those TAFE graduates not studying (47 of all graduates) some form of marginal attachment or no attachment to the labour force is a more likely outcome in the short term than is full employment(Long 2004 p6)

These findings for vocational educationmdashthat apprenticeships improve employment prospects but that other forms of vocational education in general do not substantially improve labour market outcomesmdashis consistent with other work in Australia conducted by economists (Dockery amp Norris 1996 Nevile amp Saunders 1998) and is similar to international research (Ryan 2001)

More recently similar conclusions were reached by Marks (2006) comparing full-time vocational study in the first year after leaving school with full-time work (full-time vocational study includes study at a TAFE or private institution but excludes apprentices whom are classified as full-time workers) He found that full-time vocational study increases the chances of being in full-time study or part-time work in the fourth year after leaving school It does not however increase the chances of full-time work Full-time vocational study in the first year increases the odds of being in full-time study rather than full-time work three years later about three times for men and twice as much for women It also increases the odds of being in part-time rather than full-time work in the fourth year Among men full-time vocational study in the first year (relative to full-time work) increases the likelihood of unemployment and lsquootherrsquo activities in the fourth year The lack of positive effects for full-time study on full-time work several years later is an important finding

It is not clear whether the differences between the conclusions of the earlier and later studies on the impact of VET reflect differences in emphasis placed on estimation results methodological differences the choice of comparison group or that VET is more useful for employment outcomes now than in the 1990s

The approach taken in this study is different from previous studies in one major respect Most of the previous research can be characterised as taking a lsquowhat did they do then and where are they now approachrsquo that is comparing outcomes for those with a VET qualification with those without a VET qualification This is eminently sensible for studying the outcome for

NCVER 11

individuals in a given year (for example what are the most likely labour market states for individuals given their post-school qualifications six years after leaving school) but it is not very well suited to study labour market dynamics Instead we seek to model year-on-year transitions between labour market states and investigate the role education and training has on the probability of making these transitions

12 Annual transitions between labour market states for young Australians

MethodologyAs individuals are interviewed annually information is available on their labour market status on a year-by-year basis In answering the first research question we therefore use an annual timeframe to evaluate the influence of labour market status in one period on labour market status in the subsequent period Note that this implies that we not only capture persistence in particular labour market states but also all transitions from one labour market state to another The different labour market states defined are full-time permanent employment part-time permanent employment casual employment5 unemployment and not being in the labour force

Many factors influence labour market outcomes and it is common not to have indicators for some of them When such unobserved variables are not controlled for in the analysis their effects may be attributed erroneously to observed variables This is what triggers the third research question on the nature of the observed persistence of labour market states

The labelling of genuine and spurious persistence6 is widely used in studies of labour market dynamics and dates back to Heckman (1978) The idea paraphrased here is that there are potentially two mechanisms at work that will lead to the same empirical observation that people unemployed in the previous period are more likely to be unemployed in the current period compared with individuals who were not7 The first mechanism genuine persistence captures that there is something intrinsic about unemployment that has a causal effect on the probability of being unemployed next period It may for instance act to diminish the job-ready skills of an individual or give an individual a stigma that makes them less likely to be hired by employers

The second mechanism spurious persistence captures the fact that there are underlying non-observed factors that drive the probability of being unemployed now and in the future Because these underlying factors affect the probability of being unemployed in every period it only appears that previous unemployment leads to future unemployment In actual effect being unemployed in both periods is driven by the same underlying (unobserved) process This underlying process is typically labelled lsquounobserved heterogeneityrsquo indicating that different people are unique in their own special way and that this uniqueness is not observable to the researcher who is trying to model the individualrsquos labour market dynamics

5 Includes both full-time and part-time casual employment6 Persistence is often referred to as lsquostate dependencersquo7 Unemployment is chosen here to illustrate the point One can readily insert any other

labour market outcome instead

NCVER 13

The methods used to control for the influence of unobserved variables are described later In controlling for these influences using a random effects specification we make two assumptions namely that the unobserved variables are constant over time and secondly that unobserved characteristics are not related to those that are observed As these assumptions are not only necessary but also contentious they will also be discussed later in the methodology section Further in light of the assumptions we make much of what would typically be regarded as unobserved heterogeneity explicit where possible (for example personality traits and ability) and we limit the analysis to a reasonably short four-year window when individuals have completed their post-school education and training making the assumption that the unobserved heterogeneity does not vary over time more palatable

The dataThe data used for this report are based on responses from a nationally representative sample of the cohort of young people who were in Year 9 in 1995 (the Y95 cohort) This cohort sample forms part of the LSAY program The young people in this sample responded to a printed questionnaire and to reading comprehension and numeracy tests conducted in their schools in 1995 to a mailed questionnaire administered in 1996 and to telephone interviews conducted annually since then

The initial sample included 13 613 respondents from approximately 300 government Catholic and independent schools from all Australian states and territories In the 2001 survey responses were received from 6876 individuals and by 2006 3914 individuals provided useable responses The modal age of respondents in 2006 was 25 years Because attrition was not uniform sample weights were used to reflect the original population characteristics

Defining mutually exclusive labour market statesIn order to study the annual transitions between different labour market states it is necessary to identify mutually exclusive labour market states We use the time-consistent version of the LSAY Y95 cohort data as provided by NCVER8 and create mutually exclusive labour market states based on this (1) full-time permanently employed (2) part-time permanently employed (3) casually employed (4) unemployed and (5) not in the labour force

Before describing the model used for the statistical analysis we discuss those factors that are included as explanatory variables for the labour market states we observe

8 This is the version of LSAY95 that is publicly available at the Australian Social Science Data Archive (ASSDA) subject to approval It contains the original data elements in the previous release of the LSAY95 data but also includes a series of derived variables in relation to labour market status education etc that have been made consistent over all 12 waves of the LSAY Y95

14 Annual transitions between labour market states for young Australians

Factors that can help explain labour market status post-qualificationPrevious period labour market stateThe research questions we seek to answer immediately lead to one set of factors that need to be included as explanatory variables lagged versions of our dependent variable Without the lags we cannot say anything about transitions between labour market states

Details of post-school qualificationsIn addition to the lagged labour market states that are included as explanatory factors we also include details on the highest level of post-school qualifications obtained distinguishing between certificates I or II certificate III certificate IV a certificate but at an unknown level an advanced diplomadiploma (including associate degree) and bachelor degree or higher

Personal characteristics of the respondentA small number of personal characteristics of the respondent are included They are indicators for being male and indicators for being married or being in a de facto relationship Those individuals not born in Australia are classified as having been born in either a mainly English-speaking country or a non-English speaking country Furthermore we use a categorised parental occupational class indicator distinguishing between upper upper-middle lower-middle and lower

Reading and maths abilityStudentsrsquo reading and maths ability was tested in wave 1 of LSAY Y95 that is at age 15 These are expressed as achievement scores on a scale from 0 to 20 Self-rated proficiencies are also available but these are not used in favour of the objectively measured achievement levels

Personality traitsStudents were asked in wave 3 (that is at approximately age 17) to rate themselves on a 1 to 4 scale according to specific personality traits with 1 corresponding to lsquoveryrsquo and 4 relating to lsquonot at allrsquo The traits distinguished were being agreeable open popular intellectual calm hardworking outgoing and confident

The general structure of the modelThe method used in the statistical analysis to model the labour market dynamics is a multinomial logit model (MNL) The inclusion of the respondentsrsquo previous labour market states makes it a dynamic multinomial logit model The role of unobserved heterogeneity is accounted for by the inclusion of random effects An alternative way to interpret random effects is to think of them as the constant terms in the multinomial logit model being random The resulting model with random alternative specific constants is known as a form of the lsquomixed multinomial logitrsquo (MMNL) or

NCVER 15

lsquorandom parameter logitrsquo (RPL) model9 The random parameters in this case are the constants in the model This model has considerable advantages when analysing state dependence in the presence of unobserved heterogeneity and will be briefly discussed here

To discuss the modelrsquos structure and to illustrate why this specification is the best choice we first introduce some notation Let the dependent variable Yit denote the labour market state at time t for individual i Factors that influence the choice for Yit are denoted by Xit and lagged values of the dependent variable Yit-1 The explanatory variables in Xit as outlined previously imply a clear direction of the association between Xit and Yit That is Xit influences Yit but the reverse cannot hold The unobserved heterogeneitymdashin this study captured by an individual specific random effectmdashis denoted by μi

Why a logit specificationWhen it comes to modelling discrete choice variables the two main types of models are the logit and the probit models Usually the inclusion of lagged dependent variables creates an endogeneity problem but not so in a mixed logit framework Train (2003 p118) explains the issue in plain language Using our notation the Yit depends on Xit and the error term εit If thatrsquos the case then Yit-1 depends on Xit-1 and the error term εit-1 In the presence of unobserved heterogeneity the error term εit includes the unobserved heterogeneity component μi This μi component in the error terms makes these error terms correlated over time (Train 2003 p115) When one includes the lagged dependent variable Yt-1 on the right-hand side as an explanatory variable it will thus violate the independence assumption between the right-hand side variables and the error term in the equation Yt = γYt-1 + β Xt + εt This is easy to see when one realises that Yt-1 depends on εt-1 and εt and εt-1 are correlated because both include μi If a probit specification were to be used the error structure used in the estimation would have to be corrected to account for this Standard statistical software packages such as Stata now have routines that make this correction for binary probits that include lagged values of the dependent variable in the presence of unobserved heterogeneity10 However these routines have not yet been extended to multinomial probits

Train (2003 p150) explains why choosing a mixed logit set-up including lagged dependent variables as explanatory variables on the right-hand side does not pose a problem in the presence of unobserved heterogeneity unlike the case of a probit specification The crux of it is that conditional on the unobserved heterogeneity μi the estimation boils down to estimating a standard multinomial logit modelmdashwith a single complication the unobserved heterogeneity μi on which it was conditioned needs to be lsquointegrated outrsquo Fortunately this can be done using standard numerical simulation making the mixed logit approach (in our case multinomial mixed logit due to a multiple of choices) an ideal candidate to model state dependence in the presence of unobserved heterogeneity In the next subsections we provide more detail on how the estimation works

9 Any parameter in a MMNL can be modelled as a random variable not just the constants10Note that when it is assumed that there is no correlation over time in the error terms eg

in the absence of unobserved heterogeneity including lagged dependent variables Yt-1 does not pose a problem

16 Annual transitions between labour market states for young Australians

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 9: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

IntroductionThis study examines the dynamics of the labour market particularly in terms of flows between employment unemployment and non-participation in the labour force Of particular interest is the role of post-school qualifications on these labour market transitions It thus naturally fits in with existing studies on labour market dynamics in general What sets this study apart is that the population of interest is very narrowly defined young Australians between the ages of roughly 22 and 25 who have completed their education and training2 To paraphrase we study labour market dynamics for young Australians lsquoafter the dust has settledrsquo

Objective of the researchThe main objective of this project is to investigate the role of post-school qualifications on the transitions between employment and non-employment Specifically the following three questions are addressed1 What are the transitions between the following labour market states for

young Australians permanent employment casual employment unemployment and not being in the labour force

2 How persistent are the following labour market states which can be classified as less desirable for different levels of post-school qualifications casual employment unemployment and not being in the labour force

3 How much of any observed persistence is lsquogenuinersquo and how much of the observed persistence is lsquospuriousrsquo3

The policy narrative of these three research questions runs from getting a good perspective on how socioeconomic and personal characteristics are correlated with labour market choices and what role previous experiences play (question 1) to investigating the role of education and training in escaping less desirable or bad labour market states (question 2) and finally the mechanism at work behind the persistence in labour market states as observed in the data

2 There has been an increasing awareness and emphasis on lifelong learning to ensure that individuals maintain or acquire the necessary skills to participate in society throughout their lives Having completed education and training as used here means that individuals are not observed to be enrolled in study or training leading to a qualification for the duration of the period that is being modelled that is roughly between the ages of 22 and 25 They may take up such study further into their careerlife

3 The concepts of genuine and spurious persistence will be discussed in more detail in the methodology section but here it is sufficient to state that any persistence observed in the data can have two different mechanisms that would give rise to this persistence

NCVER 9

There exists a body of research on the factors associated with particular labour market outcomes and some of the findings from these research efforts are discussed later However much less is known about the process of switching back and forth between labour market states Understanding these dynamics is crucial in order to develop initiatives that would for instance increase transitions out of unemployment and into permanent employment For instance it is a long-standing desire of successive Australian governments to minimise the amount of time that Australian youths spend in so-called marginal activities In this context knowing what factors are associated with being unemployed is not enough It is also necessary to understand the transition from employment into unemployment and the role education and other personal characteristics play in these transitions4

To frame our research objectives we first briefly discuss Australian evidence with respect to the role of education and training in young peoplersquos post-school outcomes A broader literature on labour market dynamics does exist but is not discussed in detail

Previous studiesParticipation in vocational education and training (VET) is an important pathway in the school-to-work transitions for many young people in Australia VET offers opportunities for young people to develop skills that employers value VET comprises apprenticeships and traineeships (which involve employment) and non-apprenticeship courses offered at publicly funded technical and further (TAFE) institutions and by private VET providers VET is particularly useful for young men who did not complete Year 12 of whom just over a half completes an apprenticeship or traineeship (Curtis amp McMillan 2008)

Apprentices are more likely to be male have a low-to-medium socioeconomic background have a parent in a technical or trade occupation have attended a government school and have relatively lower test scores in reading and mathematics while at school They are also likely to have a father with a trade qualification and have studied VET subjects at school (Ainley amp Corrigan 2005 Curtis 2008) Trainees are more likely to be female less likely to come from a professional background and have relatively high scores in reading and mathematics (Ainley amp Corrigan 2005) Participation in non-apprenticeship VET study is generally evenly distributed across social groups although there are some differences by VET course level (McMillan Rothman amp Wernert 2005)

In general participation in VET tends to be associated with subsequent higher rates of full-time employment by comparison with those who undertake no post-school study Curtis (2008) reports that of those who had participated in a non-apprenticeship VET course 66 of non-completers and 78 of completers were working full-time The comparable figure for those with no post-school qualifications was 69 For apprenticeships the level of full-time employment is higher at around 93 for completers and 80 for non-completers For traineeships the level of

4 To give an example we find that obtaining a high enough level of qualification can mitigate the effect of becoming unemployed but these findings and others will be discussed in the results section

10 Annual transitions between labour market states for young Australians

full-time employment was in the middle at around 82 for completers and 79 for non-completers Completion of an apprenticeship has the strongest positive impact on obtaining full-time work which is not surprising as apprenticeships are themselves considered a form of full-time work

However earlier research has suggested that full-time vocational study is not particularly beneficial in terms of labour market outcomes for at least some types of courses Analyses of the three older cohorts from the Youth in Transition (YIT) cohorts born in 1961 1965 and 1970 the youngest YIT cohort born in 1975 and the 1995 Longitudinal Surveys of Australian Youth Year 9 cohort have not found strong positive effects for VET on labour market outcomes (Long McKenzie amp Sturman 1996 Marks Hillman amp Beavis 2003 McMillan amp Marks 2003 but see Ryan 2002) The exception is apprenticeships which do substantially improve labour market outcomes By contrast according to Long (2004 pp20ndash1) the benefits of TAFE qualifications are at best equivocal

In the year after completing their qualification 25 of TAFE graduates were not in full-time work and were not enrolled in further study For those TAFE graduates not studying (47 of all graduates) some form of marginal attachment or no attachment to the labour force is a more likely outcome in the short term than is full employment(Long 2004 p6)

These findings for vocational educationmdashthat apprenticeships improve employment prospects but that other forms of vocational education in general do not substantially improve labour market outcomesmdashis consistent with other work in Australia conducted by economists (Dockery amp Norris 1996 Nevile amp Saunders 1998) and is similar to international research (Ryan 2001)

More recently similar conclusions were reached by Marks (2006) comparing full-time vocational study in the first year after leaving school with full-time work (full-time vocational study includes study at a TAFE or private institution but excludes apprentices whom are classified as full-time workers) He found that full-time vocational study increases the chances of being in full-time study or part-time work in the fourth year after leaving school It does not however increase the chances of full-time work Full-time vocational study in the first year increases the odds of being in full-time study rather than full-time work three years later about three times for men and twice as much for women It also increases the odds of being in part-time rather than full-time work in the fourth year Among men full-time vocational study in the first year (relative to full-time work) increases the likelihood of unemployment and lsquootherrsquo activities in the fourth year The lack of positive effects for full-time study on full-time work several years later is an important finding

It is not clear whether the differences between the conclusions of the earlier and later studies on the impact of VET reflect differences in emphasis placed on estimation results methodological differences the choice of comparison group or that VET is more useful for employment outcomes now than in the 1990s

The approach taken in this study is different from previous studies in one major respect Most of the previous research can be characterised as taking a lsquowhat did they do then and where are they now approachrsquo that is comparing outcomes for those with a VET qualification with those without a VET qualification This is eminently sensible for studying the outcome for

NCVER 11

individuals in a given year (for example what are the most likely labour market states for individuals given their post-school qualifications six years after leaving school) but it is not very well suited to study labour market dynamics Instead we seek to model year-on-year transitions between labour market states and investigate the role education and training has on the probability of making these transitions

12 Annual transitions between labour market states for young Australians

MethodologyAs individuals are interviewed annually information is available on their labour market status on a year-by-year basis In answering the first research question we therefore use an annual timeframe to evaluate the influence of labour market status in one period on labour market status in the subsequent period Note that this implies that we not only capture persistence in particular labour market states but also all transitions from one labour market state to another The different labour market states defined are full-time permanent employment part-time permanent employment casual employment5 unemployment and not being in the labour force

Many factors influence labour market outcomes and it is common not to have indicators for some of them When such unobserved variables are not controlled for in the analysis their effects may be attributed erroneously to observed variables This is what triggers the third research question on the nature of the observed persistence of labour market states

The labelling of genuine and spurious persistence6 is widely used in studies of labour market dynamics and dates back to Heckman (1978) The idea paraphrased here is that there are potentially two mechanisms at work that will lead to the same empirical observation that people unemployed in the previous period are more likely to be unemployed in the current period compared with individuals who were not7 The first mechanism genuine persistence captures that there is something intrinsic about unemployment that has a causal effect on the probability of being unemployed next period It may for instance act to diminish the job-ready skills of an individual or give an individual a stigma that makes them less likely to be hired by employers

The second mechanism spurious persistence captures the fact that there are underlying non-observed factors that drive the probability of being unemployed now and in the future Because these underlying factors affect the probability of being unemployed in every period it only appears that previous unemployment leads to future unemployment In actual effect being unemployed in both periods is driven by the same underlying (unobserved) process This underlying process is typically labelled lsquounobserved heterogeneityrsquo indicating that different people are unique in their own special way and that this uniqueness is not observable to the researcher who is trying to model the individualrsquos labour market dynamics

5 Includes both full-time and part-time casual employment6 Persistence is often referred to as lsquostate dependencersquo7 Unemployment is chosen here to illustrate the point One can readily insert any other

labour market outcome instead

NCVER 13

The methods used to control for the influence of unobserved variables are described later In controlling for these influences using a random effects specification we make two assumptions namely that the unobserved variables are constant over time and secondly that unobserved characteristics are not related to those that are observed As these assumptions are not only necessary but also contentious they will also be discussed later in the methodology section Further in light of the assumptions we make much of what would typically be regarded as unobserved heterogeneity explicit where possible (for example personality traits and ability) and we limit the analysis to a reasonably short four-year window when individuals have completed their post-school education and training making the assumption that the unobserved heterogeneity does not vary over time more palatable

The dataThe data used for this report are based on responses from a nationally representative sample of the cohort of young people who were in Year 9 in 1995 (the Y95 cohort) This cohort sample forms part of the LSAY program The young people in this sample responded to a printed questionnaire and to reading comprehension and numeracy tests conducted in their schools in 1995 to a mailed questionnaire administered in 1996 and to telephone interviews conducted annually since then

The initial sample included 13 613 respondents from approximately 300 government Catholic and independent schools from all Australian states and territories In the 2001 survey responses were received from 6876 individuals and by 2006 3914 individuals provided useable responses The modal age of respondents in 2006 was 25 years Because attrition was not uniform sample weights were used to reflect the original population characteristics

Defining mutually exclusive labour market statesIn order to study the annual transitions between different labour market states it is necessary to identify mutually exclusive labour market states We use the time-consistent version of the LSAY Y95 cohort data as provided by NCVER8 and create mutually exclusive labour market states based on this (1) full-time permanently employed (2) part-time permanently employed (3) casually employed (4) unemployed and (5) not in the labour force

Before describing the model used for the statistical analysis we discuss those factors that are included as explanatory variables for the labour market states we observe

8 This is the version of LSAY95 that is publicly available at the Australian Social Science Data Archive (ASSDA) subject to approval It contains the original data elements in the previous release of the LSAY95 data but also includes a series of derived variables in relation to labour market status education etc that have been made consistent over all 12 waves of the LSAY Y95

14 Annual transitions between labour market states for young Australians

Factors that can help explain labour market status post-qualificationPrevious period labour market stateThe research questions we seek to answer immediately lead to one set of factors that need to be included as explanatory variables lagged versions of our dependent variable Without the lags we cannot say anything about transitions between labour market states

Details of post-school qualificationsIn addition to the lagged labour market states that are included as explanatory factors we also include details on the highest level of post-school qualifications obtained distinguishing between certificates I or II certificate III certificate IV a certificate but at an unknown level an advanced diplomadiploma (including associate degree) and bachelor degree or higher

Personal characteristics of the respondentA small number of personal characteristics of the respondent are included They are indicators for being male and indicators for being married or being in a de facto relationship Those individuals not born in Australia are classified as having been born in either a mainly English-speaking country or a non-English speaking country Furthermore we use a categorised parental occupational class indicator distinguishing between upper upper-middle lower-middle and lower

Reading and maths abilityStudentsrsquo reading and maths ability was tested in wave 1 of LSAY Y95 that is at age 15 These are expressed as achievement scores on a scale from 0 to 20 Self-rated proficiencies are also available but these are not used in favour of the objectively measured achievement levels

Personality traitsStudents were asked in wave 3 (that is at approximately age 17) to rate themselves on a 1 to 4 scale according to specific personality traits with 1 corresponding to lsquoveryrsquo and 4 relating to lsquonot at allrsquo The traits distinguished were being agreeable open popular intellectual calm hardworking outgoing and confident

The general structure of the modelThe method used in the statistical analysis to model the labour market dynamics is a multinomial logit model (MNL) The inclusion of the respondentsrsquo previous labour market states makes it a dynamic multinomial logit model The role of unobserved heterogeneity is accounted for by the inclusion of random effects An alternative way to interpret random effects is to think of them as the constant terms in the multinomial logit model being random The resulting model with random alternative specific constants is known as a form of the lsquomixed multinomial logitrsquo (MMNL) or

NCVER 15

lsquorandom parameter logitrsquo (RPL) model9 The random parameters in this case are the constants in the model This model has considerable advantages when analysing state dependence in the presence of unobserved heterogeneity and will be briefly discussed here

To discuss the modelrsquos structure and to illustrate why this specification is the best choice we first introduce some notation Let the dependent variable Yit denote the labour market state at time t for individual i Factors that influence the choice for Yit are denoted by Xit and lagged values of the dependent variable Yit-1 The explanatory variables in Xit as outlined previously imply a clear direction of the association between Xit and Yit That is Xit influences Yit but the reverse cannot hold The unobserved heterogeneitymdashin this study captured by an individual specific random effectmdashis denoted by μi

Why a logit specificationWhen it comes to modelling discrete choice variables the two main types of models are the logit and the probit models Usually the inclusion of lagged dependent variables creates an endogeneity problem but not so in a mixed logit framework Train (2003 p118) explains the issue in plain language Using our notation the Yit depends on Xit and the error term εit If thatrsquos the case then Yit-1 depends on Xit-1 and the error term εit-1 In the presence of unobserved heterogeneity the error term εit includes the unobserved heterogeneity component μi This μi component in the error terms makes these error terms correlated over time (Train 2003 p115) When one includes the lagged dependent variable Yt-1 on the right-hand side as an explanatory variable it will thus violate the independence assumption between the right-hand side variables and the error term in the equation Yt = γYt-1 + β Xt + εt This is easy to see when one realises that Yt-1 depends on εt-1 and εt and εt-1 are correlated because both include μi If a probit specification were to be used the error structure used in the estimation would have to be corrected to account for this Standard statistical software packages such as Stata now have routines that make this correction for binary probits that include lagged values of the dependent variable in the presence of unobserved heterogeneity10 However these routines have not yet been extended to multinomial probits

Train (2003 p150) explains why choosing a mixed logit set-up including lagged dependent variables as explanatory variables on the right-hand side does not pose a problem in the presence of unobserved heterogeneity unlike the case of a probit specification The crux of it is that conditional on the unobserved heterogeneity μi the estimation boils down to estimating a standard multinomial logit modelmdashwith a single complication the unobserved heterogeneity μi on which it was conditioned needs to be lsquointegrated outrsquo Fortunately this can be done using standard numerical simulation making the mixed logit approach (in our case multinomial mixed logit due to a multiple of choices) an ideal candidate to model state dependence in the presence of unobserved heterogeneity In the next subsections we provide more detail on how the estimation works

9 Any parameter in a MMNL can be modelled as a random variable not just the constants10Note that when it is assumed that there is no correlation over time in the error terms eg

in the absence of unobserved heterogeneity including lagged dependent variables Yt-1 does not pose a problem

16 Annual transitions between labour market states for young Australians

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 10: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

There exists a body of research on the factors associated with particular labour market outcomes and some of the findings from these research efforts are discussed later However much less is known about the process of switching back and forth between labour market states Understanding these dynamics is crucial in order to develop initiatives that would for instance increase transitions out of unemployment and into permanent employment For instance it is a long-standing desire of successive Australian governments to minimise the amount of time that Australian youths spend in so-called marginal activities In this context knowing what factors are associated with being unemployed is not enough It is also necessary to understand the transition from employment into unemployment and the role education and other personal characteristics play in these transitions4

To frame our research objectives we first briefly discuss Australian evidence with respect to the role of education and training in young peoplersquos post-school outcomes A broader literature on labour market dynamics does exist but is not discussed in detail

Previous studiesParticipation in vocational education and training (VET) is an important pathway in the school-to-work transitions for many young people in Australia VET offers opportunities for young people to develop skills that employers value VET comprises apprenticeships and traineeships (which involve employment) and non-apprenticeship courses offered at publicly funded technical and further (TAFE) institutions and by private VET providers VET is particularly useful for young men who did not complete Year 12 of whom just over a half completes an apprenticeship or traineeship (Curtis amp McMillan 2008)

Apprentices are more likely to be male have a low-to-medium socioeconomic background have a parent in a technical or trade occupation have attended a government school and have relatively lower test scores in reading and mathematics while at school They are also likely to have a father with a trade qualification and have studied VET subjects at school (Ainley amp Corrigan 2005 Curtis 2008) Trainees are more likely to be female less likely to come from a professional background and have relatively high scores in reading and mathematics (Ainley amp Corrigan 2005) Participation in non-apprenticeship VET study is generally evenly distributed across social groups although there are some differences by VET course level (McMillan Rothman amp Wernert 2005)

In general participation in VET tends to be associated with subsequent higher rates of full-time employment by comparison with those who undertake no post-school study Curtis (2008) reports that of those who had participated in a non-apprenticeship VET course 66 of non-completers and 78 of completers were working full-time The comparable figure for those with no post-school qualifications was 69 For apprenticeships the level of full-time employment is higher at around 93 for completers and 80 for non-completers For traineeships the level of

4 To give an example we find that obtaining a high enough level of qualification can mitigate the effect of becoming unemployed but these findings and others will be discussed in the results section

10 Annual transitions between labour market states for young Australians

full-time employment was in the middle at around 82 for completers and 79 for non-completers Completion of an apprenticeship has the strongest positive impact on obtaining full-time work which is not surprising as apprenticeships are themselves considered a form of full-time work

However earlier research has suggested that full-time vocational study is not particularly beneficial in terms of labour market outcomes for at least some types of courses Analyses of the three older cohorts from the Youth in Transition (YIT) cohorts born in 1961 1965 and 1970 the youngest YIT cohort born in 1975 and the 1995 Longitudinal Surveys of Australian Youth Year 9 cohort have not found strong positive effects for VET on labour market outcomes (Long McKenzie amp Sturman 1996 Marks Hillman amp Beavis 2003 McMillan amp Marks 2003 but see Ryan 2002) The exception is apprenticeships which do substantially improve labour market outcomes By contrast according to Long (2004 pp20ndash1) the benefits of TAFE qualifications are at best equivocal

In the year after completing their qualification 25 of TAFE graduates were not in full-time work and were not enrolled in further study For those TAFE graduates not studying (47 of all graduates) some form of marginal attachment or no attachment to the labour force is a more likely outcome in the short term than is full employment(Long 2004 p6)

These findings for vocational educationmdashthat apprenticeships improve employment prospects but that other forms of vocational education in general do not substantially improve labour market outcomesmdashis consistent with other work in Australia conducted by economists (Dockery amp Norris 1996 Nevile amp Saunders 1998) and is similar to international research (Ryan 2001)

More recently similar conclusions were reached by Marks (2006) comparing full-time vocational study in the first year after leaving school with full-time work (full-time vocational study includes study at a TAFE or private institution but excludes apprentices whom are classified as full-time workers) He found that full-time vocational study increases the chances of being in full-time study or part-time work in the fourth year after leaving school It does not however increase the chances of full-time work Full-time vocational study in the first year increases the odds of being in full-time study rather than full-time work three years later about three times for men and twice as much for women It also increases the odds of being in part-time rather than full-time work in the fourth year Among men full-time vocational study in the first year (relative to full-time work) increases the likelihood of unemployment and lsquootherrsquo activities in the fourth year The lack of positive effects for full-time study on full-time work several years later is an important finding

It is not clear whether the differences between the conclusions of the earlier and later studies on the impact of VET reflect differences in emphasis placed on estimation results methodological differences the choice of comparison group or that VET is more useful for employment outcomes now than in the 1990s

The approach taken in this study is different from previous studies in one major respect Most of the previous research can be characterised as taking a lsquowhat did they do then and where are they now approachrsquo that is comparing outcomes for those with a VET qualification with those without a VET qualification This is eminently sensible for studying the outcome for

NCVER 11

individuals in a given year (for example what are the most likely labour market states for individuals given their post-school qualifications six years after leaving school) but it is not very well suited to study labour market dynamics Instead we seek to model year-on-year transitions between labour market states and investigate the role education and training has on the probability of making these transitions

12 Annual transitions between labour market states for young Australians

MethodologyAs individuals are interviewed annually information is available on their labour market status on a year-by-year basis In answering the first research question we therefore use an annual timeframe to evaluate the influence of labour market status in one period on labour market status in the subsequent period Note that this implies that we not only capture persistence in particular labour market states but also all transitions from one labour market state to another The different labour market states defined are full-time permanent employment part-time permanent employment casual employment5 unemployment and not being in the labour force

Many factors influence labour market outcomes and it is common not to have indicators for some of them When such unobserved variables are not controlled for in the analysis their effects may be attributed erroneously to observed variables This is what triggers the third research question on the nature of the observed persistence of labour market states

The labelling of genuine and spurious persistence6 is widely used in studies of labour market dynamics and dates back to Heckman (1978) The idea paraphrased here is that there are potentially two mechanisms at work that will lead to the same empirical observation that people unemployed in the previous period are more likely to be unemployed in the current period compared with individuals who were not7 The first mechanism genuine persistence captures that there is something intrinsic about unemployment that has a causal effect on the probability of being unemployed next period It may for instance act to diminish the job-ready skills of an individual or give an individual a stigma that makes them less likely to be hired by employers

The second mechanism spurious persistence captures the fact that there are underlying non-observed factors that drive the probability of being unemployed now and in the future Because these underlying factors affect the probability of being unemployed in every period it only appears that previous unemployment leads to future unemployment In actual effect being unemployed in both periods is driven by the same underlying (unobserved) process This underlying process is typically labelled lsquounobserved heterogeneityrsquo indicating that different people are unique in their own special way and that this uniqueness is not observable to the researcher who is trying to model the individualrsquos labour market dynamics

5 Includes both full-time and part-time casual employment6 Persistence is often referred to as lsquostate dependencersquo7 Unemployment is chosen here to illustrate the point One can readily insert any other

labour market outcome instead

NCVER 13

The methods used to control for the influence of unobserved variables are described later In controlling for these influences using a random effects specification we make two assumptions namely that the unobserved variables are constant over time and secondly that unobserved characteristics are not related to those that are observed As these assumptions are not only necessary but also contentious they will also be discussed later in the methodology section Further in light of the assumptions we make much of what would typically be regarded as unobserved heterogeneity explicit where possible (for example personality traits and ability) and we limit the analysis to a reasonably short four-year window when individuals have completed their post-school education and training making the assumption that the unobserved heterogeneity does not vary over time more palatable

The dataThe data used for this report are based on responses from a nationally representative sample of the cohort of young people who were in Year 9 in 1995 (the Y95 cohort) This cohort sample forms part of the LSAY program The young people in this sample responded to a printed questionnaire and to reading comprehension and numeracy tests conducted in their schools in 1995 to a mailed questionnaire administered in 1996 and to telephone interviews conducted annually since then

The initial sample included 13 613 respondents from approximately 300 government Catholic and independent schools from all Australian states and territories In the 2001 survey responses were received from 6876 individuals and by 2006 3914 individuals provided useable responses The modal age of respondents in 2006 was 25 years Because attrition was not uniform sample weights were used to reflect the original population characteristics

Defining mutually exclusive labour market statesIn order to study the annual transitions between different labour market states it is necessary to identify mutually exclusive labour market states We use the time-consistent version of the LSAY Y95 cohort data as provided by NCVER8 and create mutually exclusive labour market states based on this (1) full-time permanently employed (2) part-time permanently employed (3) casually employed (4) unemployed and (5) not in the labour force

Before describing the model used for the statistical analysis we discuss those factors that are included as explanatory variables for the labour market states we observe

8 This is the version of LSAY95 that is publicly available at the Australian Social Science Data Archive (ASSDA) subject to approval It contains the original data elements in the previous release of the LSAY95 data but also includes a series of derived variables in relation to labour market status education etc that have been made consistent over all 12 waves of the LSAY Y95

14 Annual transitions between labour market states for young Australians

Factors that can help explain labour market status post-qualificationPrevious period labour market stateThe research questions we seek to answer immediately lead to one set of factors that need to be included as explanatory variables lagged versions of our dependent variable Without the lags we cannot say anything about transitions between labour market states

Details of post-school qualificationsIn addition to the lagged labour market states that are included as explanatory factors we also include details on the highest level of post-school qualifications obtained distinguishing between certificates I or II certificate III certificate IV a certificate but at an unknown level an advanced diplomadiploma (including associate degree) and bachelor degree or higher

Personal characteristics of the respondentA small number of personal characteristics of the respondent are included They are indicators for being male and indicators for being married or being in a de facto relationship Those individuals not born in Australia are classified as having been born in either a mainly English-speaking country or a non-English speaking country Furthermore we use a categorised parental occupational class indicator distinguishing between upper upper-middle lower-middle and lower

Reading and maths abilityStudentsrsquo reading and maths ability was tested in wave 1 of LSAY Y95 that is at age 15 These are expressed as achievement scores on a scale from 0 to 20 Self-rated proficiencies are also available but these are not used in favour of the objectively measured achievement levels

Personality traitsStudents were asked in wave 3 (that is at approximately age 17) to rate themselves on a 1 to 4 scale according to specific personality traits with 1 corresponding to lsquoveryrsquo and 4 relating to lsquonot at allrsquo The traits distinguished were being agreeable open popular intellectual calm hardworking outgoing and confident

The general structure of the modelThe method used in the statistical analysis to model the labour market dynamics is a multinomial logit model (MNL) The inclusion of the respondentsrsquo previous labour market states makes it a dynamic multinomial logit model The role of unobserved heterogeneity is accounted for by the inclusion of random effects An alternative way to interpret random effects is to think of them as the constant terms in the multinomial logit model being random The resulting model with random alternative specific constants is known as a form of the lsquomixed multinomial logitrsquo (MMNL) or

NCVER 15

lsquorandom parameter logitrsquo (RPL) model9 The random parameters in this case are the constants in the model This model has considerable advantages when analysing state dependence in the presence of unobserved heterogeneity and will be briefly discussed here

To discuss the modelrsquos structure and to illustrate why this specification is the best choice we first introduce some notation Let the dependent variable Yit denote the labour market state at time t for individual i Factors that influence the choice for Yit are denoted by Xit and lagged values of the dependent variable Yit-1 The explanatory variables in Xit as outlined previously imply a clear direction of the association between Xit and Yit That is Xit influences Yit but the reverse cannot hold The unobserved heterogeneitymdashin this study captured by an individual specific random effectmdashis denoted by μi

Why a logit specificationWhen it comes to modelling discrete choice variables the two main types of models are the logit and the probit models Usually the inclusion of lagged dependent variables creates an endogeneity problem but not so in a mixed logit framework Train (2003 p118) explains the issue in plain language Using our notation the Yit depends on Xit and the error term εit If thatrsquos the case then Yit-1 depends on Xit-1 and the error term εit-1 In the presence of unobserved heterogeneity the error term εit includes the unobserved heterogeneity component μi This μi component in the error terms makes these error terms correlated over time (Train 2003 p115) When one includes the lagged dependent variable Yt-1 on the right-hand side as an explanatory variable it will thus violate the independence assumption between the right-hand side variables and the error term in the equation Yt = γYt-1 + β Xt + εt This is easy to see when one realises that Yt-1 depends on εt-1 and εt and εt-1 are correlated because both include μi If a probit specification were to be used the error structure used in the estimation would have to be corrected to account for this Standard statistical software packages such as Stata now have routines that make this correction for binary probits that include lagged values of the dependent variable in the presence of unobserved heterogeneity10 However these routines have not yet been extended to multinomial probits

Train (2003 p150) explains why choosing a mixed logit set-up including lagged dependent variables as explanatory variables on the right-hand side does not pose a problem in the presence of unobserved heterogeneity unlike the case of a probit specification The crux of it is that conditional on the unobserved heterogeneity μi the estimation boils down to estimating a standard multinomial logit modelmdashwith a single complication the unobserved heterogeneity μi on which it was conditioned needs to be lsquointegrated outrsquo Fortunately this can be done using standard numerical simulation making the mixed logit approach (in our case multinomial mixed logit due to a multiple of choices) an ideal candidate to model state dependence in the presence of unobserved heterogeneity In the next subsections we provide more detail on how the estimation works

9 Any parameter in a MMNL can be modelled as a random variable not just the constants10Note that when it is assumed that there is no correlation over time in the error terms eg

in the absence of unobserved heterogeneity including lagged dependent variables Yt-1 does not pose a problem

16 Annual transitions between labour market states for young Australians

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 11: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

full-time employment was in the middle at around 82 for completers and 79 for non-completers Completion of an apprenticeship has the strongest positive impact on obtaining full-time work which is not surprising as apprenticeships are themselves considered a form of full-time work

However earlier research has suggested that full-time vocational study is not particularly beneficial in terms of labour market outcomes for at least some types of courses Analyses of the three older cohorts from the Youth in Transition (YIT) cohorts born in 1961 1965 and 1970 the youngest YIT cohort born in 1975 and the 1995 Longitudinal Surveys of Australian Youth Year 9 cohort have not found strong positive effects for VET on labour market outcomes (Long McKenzie amp Sturman 1996 Marks Hillman amp Beavis 2003 McMillan amp Marks 2003 but see Ryan 2002) The exception is apprenticeships which do substantially improve labour market outcomes By contrast according to Long (2004 pp20ndash1) the benefits of TAFE qualifications are at best equivocal

In the year after completing their qualification 25 of TAFE graduates were not in full-time work and were not enrolled in further study For those TAFE graduates not studying (47 of all graduates) some form of marginal attachment or no attachment to the labour force is a more likely outcome in the short term than is full employment(Long 2004 p6)

These findings for vocational educationmdashthat apprenticeships improve employment prospects but that other forms of vocational education in general do not substantially improve labour market outcomesmdashis consistent with other work in Australia conducted by economists (Dockery amp Norris 1996 Nevile amp Saunders 1998) and is similar to international research (Ryan 2001)

More recently similar conclusions were reached by Marks (2006) comparing full-time vocational study in the first year after leaving school with full-time work (full-time vocational study includes study at a TAFE or private institution but excludes apprentices whom are classified as full-time workers) He found that full-time vocational study increases the chances of being in full-time study or part-time work in the fourth year after leaving school It does not however increase the chances of full-time work Full-time vocational study in the first year increases the odds of being in full-time study rather than full-time work three years later about three times for men and twice as much for women It also increases the odds of being in part-time rather than full-time work in the fourth year Among men full-time vocational study in the first year (relative to full-time work) increases the likelihood of unemployment and lsquootherrsquo activities in the fourth year The lack of positive effects for full-time study on full-time work several years later is an important finding

It is not clear whether the differences between the conclusions of the earlier and later studies on the impact of VET reflect differences in emphasis placed on estimation results methodological differences the choice of comparison group or that VET is more useful for employment outcomes now than in the 1990s

The approach taken in this study is different from previous studies in one major respect Most of the previous research can be characterised as taking a lsquowhat did they do then and where are they now approachrsquo that is comparing outcomes for those with a VET qualification with those without a VET qualification This is eminently sensible for studying the outcome for

NCVER 11

individuals in a given year (for example what are the most likely labour market states for individuals given their post-school qualifications six years after leaving school) but it is not very well suited to study labour market dynamics Instead we seek to model year-on-year transitions between labour market states and investigate the role education and training has on the probability of making these transitions

12 Annual transitions between labour market states for young Australians

MethodologyAs individuals are interviewed annually information is available on their labour market status on a year-by-year basis In answering the first research question we therefore use an annual timeframe to evaluate the influence of labour market status in one period on labour market status in the subsequent period Note that this implies that we not only capture persistence in particular labour market states but also all transitions from one labour market state to another The different labour market states defined are full-time permanent employment part-time permanent employment casual employment5 unemployment and not being in the labour force

Many factors influence labour market outcomes and it is common not to have indicators for some of them When such unobserved variables are not controlled for in the analysis their effects may be attributed erroneously to observed variables This is what triggers the third research question on the nature of the observed persistence of labour market states

The labelling of genuine and spurious persistence6 is widely used in studies of labour market dynamics and dates back to Heckman (1978) The idea paraphrased here is that there are potentially two mechanisms at work that will lead to the same empirical observation that people unemployed in the previous period are more likely to be unemployed in the current period compared with individuals who were not7 The first mechanism genuine persistence captures that there is something intrinsic about unemployment that has a causal effect on the probability of being unemployed next period It may for instance act to diminish the job-ready skills of an individual or give an individual a stigma that makes them less likely to be hired by employers

The second mechanism spurious persistence captures the fact that there are underlying non-observed factors that drive the probability of being unemployed now and in the future Because these underlying factors affect the probability of being unemployed in every period it only appears that previous unemployment leads to future unemployment In actual effect being unemployed in both periods is driven by the same underlying (unobserved) process This underlying process is typically labelled lsquounobserved heterogeneityrsquo indicating that different people are unique in their own special way and that this uniqueness is not observable to the researcher who is trying to model the individualrsquos labour market dynamics

5 Includes both full-time and part-time casual employment6 Persistence is often referred to as lsquostate dependencersquo7 Unemployment is chosen here to illustrate the point One can readily insert any other

labour market outcome instead

NCVER 13

The methods used to control for the influence of unobserved variables are described later In controlling for these influences using a random effects specification we make two assumptions namely that the unobserved variables are constant over time and secondly that unobserved characteristics are not related to those that are observed As these assumptions are not only necessary but also contentious they will also be discussed later in the methodology section Further in light of the assumptions we make much of what would typically be regarded as unobserved heterogeneity explicit where possible (for example personality traits and ability) and we limit the analysis to a reasonably short four-year window when individuals have completed their post-school education and training making the assumption that the unobserved heterogeneity does not vary over time more palatable

The dataThe data used for this report are based on responses from a nationally representative sample of the cohort of young people who were in Year 9 in 1995 (the Y95 cohort) This cohort sample forms part of the LSAY program The young people in this sample responded to a printed questionnaire and to reading comprehension and numeracy tests conducted in their schools in 1995 to a mailed questionnaire administered in 1996 and to telephone interviews conducted annually since then

The initial sample included 13 613 respondents from approximately 300 government Catholic and independent schools from all Australian states and territories In the 2001 survey responses were received from 6876 individuals and by 2006 3914 individuals provided useable responses The modal age of respondents in 2006 was 25 years Because attrition was not uniform sample weights were used to reflect the original population characteristics

Defining mutually exclusive labour market statesIn order to study the annual transitions between different labour market states it is necessary to identify mutually exclusive labour market states We use the time-consistent version of the LSAY Y95 cohort data as provided by NCVER8 and create mutually exclusive labour market states based on this (1) full-time permanently employed (2) part-time permanently employed (3) casually employed (4) unemployed and (5) not in the labour force

Before describing the model used for the statistical analysis we discuss those factors that are included as explanatory variables for the labour market states we observe

8 This is the version of LSAY95 that is publicly available at the Australian Social Science Data Archive (ASSDA) subject to approval It contains the original data elements in the previous release of the LSAY95 data but also includes a series of derived variables in relation to labour market status education etc that have been made consistent over all 12 waves of the LSAY Y95

14 Annual transitions between labour market states for young Australians

Factors that can help explain labour market status post-qualificationPrevious period labour market stateThe research questions we seek to answer immediately lead to one set of factors that need to be included as explanatory variables lagged versions of our dependent variable Without the lags we cannot say anything about transitions between labour market states

Details of post-school qualificationsIn addition to the lagged labour market states that are included as explanatory factors we also include details on the highest level of post-school qualifications obtained distinguishing between certificates I or II certificate III certificate IV a certificate but at an unknown level an advanced diplomadiploma (including associate degree) and bachelor degree or higher

Personal characteristics of the respondentA small number of personal characteristics of the respondent are included They are indicators for being male and indicators for being married or being in a de facto relationship Those individuals not born in Australia are classified as having been born in either a mainly English-speaking country or a non-English speaking country Furthermore we use a categorised parental occupational class indicator distinguishing between upper upper-middle lower-middle and lower

Reading and maths abilityStudentsrsquo reading and maths ability was tested in wave 1 of LSAY Y95 that is at age 15 These are expressed as achievement scores on a scale from 0 to 20 Self-rated proficiencies are also available but these are not used in favour of the objectively measured achievement levels

Personality traitsStudents were asked in wave 3 (that is at approximately age 17) to rate themselves on a 1 to 4 scale according to specific personality traits with 1 corresponding to lsquoveryrsquo and 4 relating to lsquonot at allrsquo The traits distinguished were being agreeable open popular intellectual calm hardworking outgoing and confident

The general structure of the modelThe method used in the statistical analysis to model the labour market dynamics is a multinomial logit model (MNL) The inclusion of the respondentsrsquo previous labour market states makes it a dynamic multinomial logit model The role of unobserved heterogeneity is accounted for by the inclusion of random effects An alternative way to interpret random effects is to think of them as the constant terms in the multinomial logit model being random The resulting model with random alternative specific constants is known as a form of the lsquomixed multinomial logitrsquo (MMNL) or

NCVER 15

lsquorandom parameter logitrsquo (RPL) model9 The random parameters in this case are the constants in the model This model has considerable advantages when analysing state dependence in the presence of unobserved heterogeneity and will be briefly discussed here

To discuss the modelrsquos structure and to illustrate why this specification is the best choice we first introduce some notation Let the dependent variable Yit denote the labour market state at time t for individual i Factors that influence the choice for Yit are denoted by Xit and lagged values of the dependent variable Yit-1 The explanatory variables in Xit as outlined previously imply a clear direction of the association between Xit and Yit That is Xit influences Yit but the reverse cannot hold The unobserved heterogeneitymdashin this study captured by an individual specific random effectmdashis denoted by μi

Why a logit specificationWhen it comes to modelling discrete choice variables the two main types of models are the logit and the probit models Usually the inclusion of lagged dependent variables creates an endogeneity problem but not so in a mixed logit framework Train (2003 p118) explains the issue in plain language Using our notation the Yit depends on Xit and the error term εit If thatrsquos the case then Yit-1 depends on Xit-1 and the error term εit-1 In the presence of unobserved heterogeneity the error term εit includes the unobserved heterogeneity component μi This μi component in the error terms makes these error terms correlated over time (Train 2003 p115) When one includes the lagged dependent variable Yt-1 on the right-hand side as an explanatory variable it will thus violate the independence assumption between the right-hand side variables and the error term in the equation Yt = γYt-1 + β Xt + εt This is easy to see when one realises that Yt-1 depends on εt-1 and εt and εt-1 are correlated because both include μi If a probit specification were to be used the error structure used in the estimation would have to be corrected to account for this Standard statistical software packages such as Stata now have routines that make this correction for binary probits that include lagged values of the dependent variable in the presence of unobserved heterogeneity10 However these routines have not yet been extended to multinomial probits

Train (2003 p150) explains why choosing a mixed logit set-up including lagged dependent variables as explanatory variables on the right-hand side does not pose a problem in the presence of unobserved heterogeneity unlike the case of a probit specification The crux of it is that conditional on the unobserved heterogeneity μi the estimation boils down to estimating a standard multinomial logit modelmdashwith a single complication the unobserved heterogeneity μi on which it was conditioned needs to be lsquointegrated outrsquo Fortunately this can be done using standard numerical simulation making the mixed logit approach (in our case multinomial mixed logit due to a multiple of choices) an ideal candidate to model state dependence in the presence of unobserved heterogeneity In the next subsections we provide more detail on how the estimation works

9 Any parameter in a MMNL can be modelled as a random variable not just the constants10Note that when it is assumed that there is no correlation over time in the error terms eg

in the absence of unobserved heterogeneity including lagged dependent variables Yt-1 does not pose a problem

16 Annual transitions between labour market states for young Australians

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 12: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

individuals in a given year (for example what are the most likely labour market states for individuals given their post-school qualifications six years after leaving school) but it is not very well suited to study labour market dynamics Instead we seek to model year-on-year transitions between labour market states and investigate the role education and training has on the probability of making these transitions

12 Annual transitions between labour market states for young Australians

MethodologyAs individuals are interviewed annually information is available on their labour market status on a year-by-year basis In answering the first research question we therefore use an annual timeframe to evaluate the influence of labour market status in one period on labour market status in the subsequent period Note that this implies that we not only capture persistence in particular labour market states but also all transitions from one labour market state to another The different labour market states defined are full-time permanent employment part-time permanent employment casual employment5 unemployment and not being in the labour force

Many factors influence labour market outcomes and it is common not to have indicators for some of them When such unobserved variables are not controlled for in the analysis their effects may be attributed erroneously to observed variables This is what triggers the third research question on the nature of the observed persistence of labour market states

The labelling of genuine and spurious persistence6 is widely used in studies of labour market dynamics and dates back to Heckman (1978) The idea paraphrased here is that there are potentially two mechanisms at work that will lead to the same empirical observation that people unemployed in the previous period are more likely to be unemployed in the current period compared with individuals who were not7 The first mechanism genuine persistence captures that there is something intrinsic about unemployment that has a causal effect on the probability of being unemployed next period It may for instance act to diminish the job-ready skills of an individual or give an individual a stigma that makes them less likely to be hired by employers

The second mechanism spurious persistence captures the fact that there are underlying non-observed factors that drive the probability of being unemployed now and in the future Because these underlying factors affect the probability of being unemployed in every period it only appears that previous unemployment leads to future unemployment In actual effect being unemployed in both periods is driven by the same underlying (unobserved) process This underlying process is typically labelled lsquounobserved heterogeneityrsquo indicating that different people are unique in their own special way and that this uniqueness is not observable to the researcher who is trying to model the individualrsquos labour market dynamics

5 Includes both full-time and part-time casual employment6 Persistence is often referred to as lsquostate dependencersquo7 Unemployment is chosen here to illustrate the point One can readily insert any other

labour market outcome instead

NCVER 13

The methods used to control for the influence of unobserved variables are described later In controlling for these influences using a random effects specification we make two assumptions namely that the unobserved variables are constant over time and secondly that unobserved characteristics are not related to those that are observed As these assumptions are not only necessary but also contentious they will also be discussed later in the methodology section Further in light of the assumptions we make much of what would typically be regarded as unobserved heterogeneity explicit where possible (for example personality traits and ability) and we limit the analysis to a reasonably short four-year window when individuals have completed their post-school education and training making the assumption that the unobserved heterogeneity does not vary over time more palatable

The dataThe data used for this report are based on responses from a nationally representative sample of the cohort of young people who were in Year 9 in 1995 (the Y95 cohort) This cohort sample forms part of the LSAY program The young people in this sample responded to a printed questionnaire and to reading comprehension and numeracy tests conducted in their schools in 1995 to a mailed questionnaire administered in 1996 and to telephone interviews conducted annually since then

The initial sample included 13 613 respondents from approximately 300 government Catholic and independent schools from all Australian states and territories In the 2001 survey responses were received from 6876 individuals and by 2006 3914 individuals provided useable responses The modal age of respondents in 2006 was 25 years Because attrition was not uniform sample weights were used to reflect the original population characteristics

Defining mutually exclusive labour market statesIn order to study the annual transitions between different labour market states it is necessary to identify mutually exclusive labour market states We use the time-consistent version of the LSAY Y95 cohort data as provided by NCVER8 and create mutually exclusive labour market states based on this (1) full-time permanently employed (2) part-time permanently employed (3) casually employed (4) unemployed and (5) not in the labour force

Before describing the model used for the statistical analysis we discuss those factors that are included as explanatory variables for the labour market states we observe

8 This is the version of LSAY95 that is publicly available at the Australian Social Science Data Archive (ASSDA) subject to approval It contains the original data elements in the previous release of the LSAY95 data but also includes a series of derived variables in relation to labour market status education etc that have been made consistent over all 12 waves of the LSAY Y95

14 Annual transitions between labour market states for young Australians

Factors that can help explain labour market status post-qualificationPrevious period labour market stateThe research questions we seek to answer immediately lead to one set of factors that need to be included as explanatory variables lagged versions of our dependent variable Without the lags we cannot say anything about transitions between labour market states

Details of post-school qualificationsIn addition to the lagged labour market states that are included as explanatory factors we also include details on the highest level of post-school qualifications obtained distinguishing between certificates I or II certificate III certificate IV a certificate but at an unknown level an advanced diplomadiploma (including associate degree) and bachelor degree or higher

Personal characteristics of the respondentA small number of personal characteristics of the respondent are included They are indicators for being male and indicators for being married or being in a de facto relationship Those individuals not born in Australia are classified as having been born in either a mainly English-speaking country or a non-English speaking country Furthermore we use a categorised parental occupational class indicator distinguishing between upper upper-middle lower-middle and lower

Reading and maths abilityStudentsrsquo reading and maths ability was tested in wave 1 of LSAY Y95 that is at age 15 These are expressed as achievement scores on a scale from 0 to 20 Self-rated proficiencies are also available but these are not used in favour of the objectively measured achievement levels

Personality traitsStudents were asked in wave 3 (that is at approximately age 17) to rate themselves on a 1 to 4 scale according to specific personality traits with 1 corresponding to lsquoveryrsquo and 4 relating to lsquonot at allrsquo The traits distinguished were being agreeable open popular intellectual calm hardworking outgoing and confident

The general structure of the modelThe method used in the statistical analysis to model the labour market dynamics is a multinomial logit model (MNL) The inclusion of the respondentsrsquo previous labour market states makes it a dynamic multinomial logit model The role of unobserved heterogeneity is accounted for by the inclusion of random effects An alternative way to interpret random effects is to think of them as the constant terms in the multinomial logit model being random The resulting model with random alternative specific constants is known as a form of the lsquomixed multinomial logitrsquo (MMNL) or

NCVER 15

lsquorandom parameter logitrsquo (RPL) model9 The random parameters in this case are the constants in the model This model has considerable advantages when analysing state dependence in the presence of unobserved heterogeneity and will be briefly discussed here

To discuss the modelrsquos structure and to illustrate why this specification is the best choice we first introduce some notation Let the dependent variable Yit denote the labour market state at time t for individual i Factors that influence the choice for Yit are denoted by Xit and lagged values of the dependent variable Yit-1 The explanatory variables in Xit as outlined previously imply a clear direction of the association between Xit and Yit That is Xit influences Yit but the reverse cannot hold The unobserved heterogeneitymdashin this study captured by an individual specific random effectmdashis denoted by μi

Why a logit specificationWhen it comes to modelling discrete choice variables the two main types of models are the logit and the probit models Usually the inclusion of lagged dependent variables creates an endogeneity problem but not so in a mixed logit framework Train (2003 p118) explains the issue in plain language Using our notation the Yit depends on Xit and the error term εit If thatrsquos the case then Yit-1 depends on Xit-1 and the error term εit-1 In the presence of unobserved heterogeneity the error term εit includes the unobserved heterogeneity component μi This μi component in the error terms makes these error terms correlated over time (Train 2003 p115) When one includes the lagged dependent variable Yt-1 on the right-hand side as an explanatory variable it will thus violate the independence assumption between the right-hand side variables and the error term in the equation Yt = γYt-1 + β Xt + εt This is easy to see when one realises that Yt-1 depends on εt-1 and εt and εt-1 are correlated because both include μi If a probit specification were to be used the error structure used in the estimation would have to be corrected to account for this Standard statistical software packages such as Stata now have routines that make this correction for binary probits that include lagged values of the dependent variable in the presence of unobserved heterogeneity10 However these routines have not yet been extended to multinomial probits

Train (2003 p150) explains why choosing a mixed logit set-up including lagged dependent variables as explanatory variables on the right-hand side does not pose a problem in the presence of unobserved heterogeneity unlike the case of a probit specification The crux of it is that conditional on the unobserved heterogeneity μi the estimation boils down to estimating a standard multinomial logit modelmdashwith a single complication the unobserved heterogeneity μi on which it was conditioned needs to be lsquointegrated outrsquo Fortunately this can be done using standard numerical simulation making the mixed logit approach (in our case multinomial mixed logit due to a multiple of choices) an ideal candidate to model state dependence in the presence of unobserved heterogeneity In the next subsections we provide more detail on how the estimation works

9 Any parameter in a MMNL can be modelled as a random variable not just the constants10Note that when it is assumed that there is no correlation over time in the error terms eg

in the absence of unobserved heterogeneity including lagged dependent variables Yt-1 does not pose a problem

16 Annual transitions between labour market states for young Australians

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 13: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

MethodologyAs individuals are interviewed annually information is available on their labour market status on a year-by-year basis In answering the first research question we therefore use an annual timeframe to evaluate the influence of labour market status in one period on labour market status in the subsequent period Note that this implies that we not only capture persistence in particular labour market states but also all transitions from one labour market state to another The different labour market states defined are full-time permanent employment part-time permanent employment casual employment5 unemployment and not being in the labour force

Many factors influence labour market outcomes and it is common not to have indicators for some of them When such unobserved variables are not controlled for in the analysis their effects may be attributed erroneously to observed variables This is what triggers the third research question on the nature of the observed persistence of labour market states

The labelling of genuine and spurious persistence6 is widely used in studies of labour market dynamics and dates back to Heckman (1978) The idea paraphrased here is that there are potentially two mechanisms at work that will lead to the same empirical observation that people unemployed in the previous period are more likely to be unemployed in the current period compared with individuals who were not7 The first mechanism genuine persistence captures that there is something intrinsic about unemployment that has a causal effect on the probability of being unemployed next period It may for instance act to diminish the job-ready skills of an individual or give an individual a stigma that makes them less likely to be hired by employers

The second mechanism spurious persistence captures the fact that there are underlying non-observed factors that drive the probability of being unemployed now and in the future Because these underlying factors affect the probability of being unemployed in every period it only appears that previous unemployment leads to future unemployment In actual effect being unemployed in both periods is driven by the same underlying (unobserved) process This underlying process is typically labelled lsquounobserved heterogeneityrsquo indicating that different people are unique in their own special way and that this uniqueness is not observable to the researcher who is trying to model the individualrsquos labour market dynamics

5 Includes both full-time and part-time casual employment6 Persistence is often referred to as lsquostate dependencersquo7 Unemployment is chosen here to illustrate the point One can readily insert any other

labour market outcome instead

NCVER 13

The methods used to control for the influence of unobserved variables are described later In controlling for these influences using a random effects specification we make two assumptions namely that the unobserved variables are constant over time and secondly that unobserved characteristics are not related to those that are observed As these assumptions are not only necessary but also contentious they will also be discussed later in the methodology section Further in light of the assumptions we make much of what would typically be regarded as unobserved heterogeneity explicit where possible (for example personality traits and ability) and we limit the analysis to a reasonably short four-year window when individuals have completed their post-school education and training making the assumption that the unobserved heterogeneity does not vary over time more palatable

The dataThe data used for this report are based on responses from a nationally representative sample of the cohort of young people who were in Year 9 in 1995 (the Y95 cohort) This cohort sample forms part of the LSAY program The young people in this sample responded to a printed questionnaire and to reading comprehension and numeracy tests conducted in their schools in 1995 to a mailed questionnaire administered in 1996 and to telephone interviews conducted annually since then

The initial sample included 13 613 respondents from approximately 300 government Catholic and independent schools from all Australian states and territories In the 2001 survey responses were received from 6876 individuals and by 2006 3914 individuals provided useable responses The modal age of respondents in 2006 was 25 years Because attrition was not uniform sample weights were used to reflect the original population characteristics

Defining mutually exclusive labour market statesIn order to study the annual transitions between different labour market states it is necessary to identify mutually exclusive labour market states We use the time-consistent version of the LSAY Y95 cohort data as provided by NCVER8 and create mutually exclusive labour market states based on this (1) full-time permanently employed (2) part-time permanently employed (3) casually employed (4) unemployed and (5) not in the labour force

Before describing the model used for the statistical analysis we discuss those factors that are included as explanatory variables for the labour market states we observe

8 This is the version of LSAY95 that is publicly available at the Australian Social Science Data Archive (ASSDA) subject to approval It contains the original data elements in the previous release of the LSAY95 data but also includes a series of derived variables in relation to labour market status education etc that have been made consistent over all 12 waves of the LSAY Y95

14 Annual transitions between labour market states for young Australians

Factors that can help explain labour market status post-qualificationPrevious period labour market stateThe research questions we seek to answer immediately lead to one set of factors that need to be included as explanatory variables lagged versions of our dependent variable Without the lags we cannot say anything about transitions between labour market states

Details of post-school qualificationsIn addition to the lagged labour market states that are included as explanatory factors we also include details on the highest level of post-school qualifications obtained distinguishing between certificates I or II certificate III certificate IV a certificate but at an unknown level an advanced diplomadiploma (including associate degree) and bachelor degree or higher

Personal characteristics of the respondentA small number of personal characteristics of the respondent are included They are indicators for being male and indicators for being married or being in a de facto relationship Those individuals not born in Australia are classified as having been born in either a mainly English-speaking country or a non-English speaking country Furthermore we use a categorised parental occupational class indicator distinguishing between upper upper-middle lower-middle and lower

Reading and maths abilityStudentsrsquo reading and maths ability was tested in wave 1 of LSAY Y95 that is at age 15 These are expressed as achievement scores on a scale from 0 to 20 Self-rated proficiencies are also available but these are not used in favour of the objectively measured achievement levels

Personality traitsStudents were asked in wave 3 (that is at approximately age 17) to rate themselves on a 1 to 4 scale according to specific personality traits with 1 corresponding to lsquoveryrsquo and 4 relating to lsquonot at allrsquo The traits distinguished were being agreeable open popular intellectual calm hardworking outgoing and confident

The general structure of the modelThe method used in the statistical analysis to model the labour market dynamics is a multinomial logit model (MNL) The inclusion of the respondentsrsquo previous labour market states makes it a dynamic multinomial logit model The role of unobserved heterogeneity is accounted for by the inclusion of random effects An alternative way to interpret random effects is to think of them as the constant terms in the multinomial logit model being random The resulting model with random alternative specific constants is known as a form of the lsquomixed multinomial logitrsquo (MMNL) or

NCVER 15

lsquorandom parameter logitrsquo (RPL) model9 The random parameters in this case are the constants in the model This model has considerable advantages when analysing state dependence in the presence of unobserved heterogeneity and will be briefly discussed here

To discuss the modelrsquos structure and to illustrate why this specification is the best choice we first introduce some notation Let the dependent variable Yit denote the labour market state at time t for individual i Factors that influence the choice for Yit are denoted by Xit and lagged values of the dependent variable Yit-1 The explanatory variables in Xit as outlined previously imply a clear direction of the association between Xit and Yit That is Xit influences Yit but the reverse cannot hold The unobserved heterogeneitymdashin this study captured by an individual specific random effectmdashis denoted by μi

Why a logit specificationWhen it comes to modelling discrete choice variables the two main types of models are the logit and the probit models Usually the inclusion of lagged dependent variables creates an endogeneity problem but not so in a mixed logit framework Train (2003 p118) explains the issue in plain language Using our notation the Yit depends on Xit and the error term εit If thatrsquos the case then Yit-1 depends on Xit-1 and the error term εit-1 In the presence of unobserved heterogeneity the error term εit includes the unobserved heterogeneity component μi This μi component in the error terms makes these error terms correlated over time (Train 2003 p115) When one includes the lagged dependent variable Yt-1 on the right-hand side as an explanatory variable it will thus violate the independence assumption between the right-hand side variables and the error term in the equation Yt = γYt-1 + β Xt + εt This is easy to see when one realises that Yt-1 depends on εt-1 and εt and εt-1 are correlated because both include μi If a probit specification were to be used the error structure used in the estimation would have to be corrected to account for this Standard statistical software packages such as Stata now have routines that make this correction for binary probits that include lagged values of the dependent variable in the presence of unobserved heterogeneity10 However these routines have not yet been extended to multinomial probits

Train (2003 p150) explains why choosing a mixed logit set-up including lagged dependent variables as explanatory variables on the right-hand side does not pose a problem in the presence of unobserved heterogeneity unlike the case of a probit specification The crux of it is that conditional on the unobserved heterogeneity μi the estimation boils down to estimating a standard multinomial logit modelmdashwith a single complication the unobserved heterogeneity μi on which it was conditioned needs to be lsquointegrated outrsquo Fortunately this can be done using standard numerical simulation making the mixed logit approach (in our case multinomial mixed logit due to a multiple of choices) an ideal candidate to model state dependence in the presence of unobserved heterogeneity In the next subsections we provide more detail on how the estimation works

9 Any parameter in a MMNL can be modelled as a random variable not just the constants10Note that when it is assumed that there is no correlation over time in the error terms eg

in the absence of unobserved heterogeneity including lagged dependent variables Yt-1 does not pose a problem

16 Annual transitions between labour market states for young Australians

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 14: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

The methods used to control for the influence of unobserved variables are described later In controlling for these influences using a random effects specification we make two assumptions namely that the unobserved variables are constant over time and secondly that unobserved characteristics are not related to those that are observed As these assumptions are not only necessary but also contentious they will also be discussed later in the methodology section Further in light of the assumptions we make much of what would typically be regarded as unobserved heterogeneity explicit where possible (for example personality traits and ability) and we limit the analysis to a reasonably short four-year window when individuals have completed their post-school education and training making the assumption that the unobserved heterogeneity does not vary over time more palatable

The dataThe data used for this report are based on responses from a nationally representative sample of the cohort of young people who were in Year 9 in 1995 (the Y95 cohort) This cohort sample forms part of the LSAY program The young people in this sample responded to a printed questionnaire and to reading comprehension and numeracy tests conducted in their schools in 1995 to a mailed questionnaire administered in 1996 and to telephone interviews conducted annually since then

The initial sample included 13 613 respondents from approximately 300 government Catholic and independent schools from all Australian states and territories In the 2001 survey responses were received from 6876 individuals and by 2006 3914 individuals provided useable responses The modal age of respondents in 2006 was 25 years Because attrition was not uniform sample weights were used to reflect the original population characteristics

Defining mutually exclusive labour market statesIn order to study the annual transitions between different labour market states it is necessary to identify mutually exclusive labour market states We use the time-consistent version of the LSAY Y95 cohort data as provided by NCVER8 and create mutually exclusive labour market states based on this (1) full-time permanently employed (2) part-time permanently employed (3) casually employed (4) unemployed and (5) not in the labour force

Before describing the model used for the statistical analysis we discuss those factors that are included as explanatory variables for the labour market states we observe

8 This is the version of LSAY95 that is publicly available at the Australian Social Science Data Archive (ASSDA) subject to approval It contains the original data elements in the previous release of the LSAY95 data but also includes a series of derived variables in relation to labour market status education etc that have been made consistent over all 12 waves of the LSAY Y95

14 Annual transitions between labour market states for young Australians

Factors that can help explain labour market status post-qualificationPrevious period labour market stateThe research questions we seek to answer immediately lead to one set of factors that need to be included as explanatory variables lagged versions of our dependent variable Without the lags we cannot say anything about transitions between labour market states

Details of post-school qualificationsIn addition to the lagged labour market states that are included as explanatory factors we also include details on the highest level of post-school qualifications obtained distinguishing between certificates I or II certificate III certificate IV a certificate but at an unknown level an advanced diplomadiploma (including associate degree) and bachelor degree or higher

Personal characteristics of the respondentA small number of personal characteristics of the respondent are included They are indicators for being male and indicators for being married or being in a de facto relationship Those individuals not born in Australia are classified as having been born in either a mainly English-speaking country or a non-English speaking country Furthermore we use a categorised parental occupational class indicator distinguishing between upper upper-middle lower-middle and lower

Reading and maths abilityStudentsrsquo reading and maths ability was tested in wave 1 of LSAY Y95 that is at age 15 These are expressed as achievement scores on a scale from 0 to 20 Self-rated proficiencies are also available but these are not used in favour of the objectively measured achievement levels

Personality traitsStudents were asked in wave 3 (that is at approximately age 17) to rate themselves on a 1 to 4 scale according to specific personality traits with 1 corresponding to lsquoveryrsquo and 4 relating to lsquonot at allrsquo The traits distinguished were being agreeable open popular intellectual calm hardworking outgoing and confident

The general structure of the modelThe method used in the statistical analysis to model the labour market dynamics is a multinomial logit model (MNL) The inclusion of the respondentsrsquo previous labour market states makes it a dynamic multinomial logit model The role of unobserved heterogeneity is accounted for by the inclusion of random effects An alternative way to interpret random effects is to think of them as the constant terms in the multinomial logit model being random The resulting model with random alternative specific constants is known as a form of the lsquomixed multinomial logitrsquo (MMNL) or

NCVER 15

lsquorandom parameter logitrsquo (RPL) model9 The random parameters in this case are the constants in the model This model has considerable advantages when analysing state dependence in the presence of unobserved heterogeneity and will be briefly discussed here

To discuss the modelrsquos structure and to illustrate why this specification is the best choice we first introduce some notation Let the dependent variable Yit denote the labour market state at time t for individual i Factors that influence the choice for Yit are denoted by Xit and lagged values of the dependent variable Yit-1 The explanatory variables in Xit as outlined previously imply a clear direction of the association between Xit and Yit That is Xit influences Yit but the reverse cannot hold The unobserved heterogeneitymdashin this study captured by an individual specific random effectmdashis denoted by μi

Why a logit specificationWhen it comes to modelling discrete choice variables the two main types of models are the logit and the probit models Usually the inclusion of lagged dependent variables creates an endogeneity problem but not so in a mixed logit framework Train (2003 p118) explains the issue in plain language Using our notation the Yit depends on Xit and the error term εit If thatrsquos the case then Yit-1 depends on Xit-1 and the error term εit-1 In the presence of unobserved heterogeneity the error term εit includes the unobserved heterogeneity component μi This μi component in the error terms makes these error terms correlated over time (Train 2003 p115) When one includes the lagged dependent variable Yt-1 on the right-hand side as an explanatory variable it will thus violate the independence assumption between the right-hand side variables and the error term in the equation Yt = γYt-1 + β Xt + εt This is easy to see when one realises that Yt-1 depends on εt-1 and εt and εt-1 are correlated because both include μi If a probit specification were to be used the error structure used in the estimation would have to be corrected to account for this Standard statistical software packages such as Stata now have routines that make this correction for binary probits that include lagged values of the dependent variable in the presence of unobserved heterogeneity10 However these routines have not yet been extended to multinomial probits

Train (2003 p150) explains why choosing a mixed logit set-up including lagged dependent variables as explanatory variables on the right-hand side does not pose a problem in the presence of unobserved heterogeneity unlike the case of a probit specification The crux of it is that conditional on the unobserved heterogeneity μi the estimation boils down to estimating a standard multinomial logit modelmdashwith a single complication the unobserved heterogeneity μi on which it was conditioned needs to be lsquointegrated outrsquo Fortunately this can be done using standard numerical simulation making the mixed logit approach (in our case multinomial mixed logit due to a multiple of choices) an ideal candidate to model state dependence in the presence of unobserved heterogeneity In the next subsections we provide more detail on how the estimation works

9 Any parameter in a MMNL can be modelled as a random variable not just the constants10Note that when it is assumed that there is no correlation over time in the error terms eg

in the absence of unobserved heterogeneity including lagged dependent variables Yt-1 does not pose a problem

16 Annual transitions between labour market states for young Australians

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 15: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Factors that can help explain labour market status post-qualificationPrevious period labour market stateThe research questions we seek to answer immediately lead to one set of factors that need to be included as explanatory variables lagged versions of our dependent variable Without the lags we cannot say anything about transitions between labour market states

Details of post-school qualificationsIn addition to the lagged labour market states that are included as explanatory factors we also include details on the highest level of post-school qualifications obtained distinguishing between certificates I or II certificate III certificate IV a certificate but at an unknown level an advanced diplomadiploma (including associate degree) and bachelor degree or higher

Personal characteristics of the respondentA small number of personal characteristics of the respondent are included They are indicators for being male and indicators for being married or being in a de facto relationship Those individuals not born in Australia are classified as having been born in either a mainly English-speaking country or a non-English speaking country Furthermore we use a categorised parental occupational class indicator distinguishing between upper upper-middle lower-middle and lower

Reading and maths abilityStudentsrsquo reading and maths ability was tested in wave 1 of LSAY Y95 that is at age 15 These are expressed as achievement scores on a scale from 0 to 20 Self-rated proficiencies are also available but these are not used in favour of the objectively measured achievement levels

Personality traitsStudents were asked in wave 3 (that is at approximately age 17) to rate themselves on a 1 to 4 scale according to specific personality traits with 1 corresponding to lsquoveryrsquo and 4 relating to lsquonot at allrsquo The traits distinguished were being agreeable open popular intellectual calm hardworking outgoing and confident

The general structure of the modelThe method used in the statistical analysis to model the labour market dynamics is a multinomial logit model (MNL) The inclusion of the respondentsrsquo previous labour market states makes it a dynamic multinomial logit model The role of unobserved heterogeneity is accounted for by the inclusion of random effects An alternative way to interpret random effects is to think of them as the constant terms in the multinomial logit model being random The resulting model with random alternative specific constants is known as a form of the lsquomixed multinomial logitrsquo (MMNL) or

NCVER 15

lsquorandom parameter logitrsquo (RPL) model9 The random parameters in this case are the constants in the model This model has considerable advantages when analysing state dependence in the presence of unobserved heterogeneity and will be briefly discussed here

To discuss the modelrsquos structure and to illustrate why this specification is the best choice we first introduce some notation Let the dependent variable Yit denote the labour market state at time t for individual i Factors that influence the choice for Yit are denoted by Xit and lagged values of the dependent variable Yit-1 The explanatory variables in Xit as outlined previously imply a clear direction of the association between Xit and Yit That is Xit influences Yit but the reverse cannot hold The unobserved heterogeneitymdashin this study captured by an individual specific random effectmdashis denoted by μi

Why a logit specificationWhen it comes to modelling discrete choice variables the two main types of models are the logit and the probit models Usually the inclusion of lagged dependent variables creates an endogeneity problem but not so in a mixed logit framework Train (2003 p118) explains the issue in plain language Using our notation the Yit depends on Xit and the error term εit If thatrsquos the case then Yit-1 depends on Xit-1 and the error term εit-1 In the presence of unobserved heterogeneity the error term εit includes the unobserved heterogeneity component μi This μi component in the error terms makes these error terms correlated over time (Train 2003 p115) When one includes the lagged dependent variable Yt-1 on the right-hand side as an explanatory variable it will thus violate the independence assumption between the right-hand side variables and the error term in the equation Yt = γYt-1 + β Xt + εt This is easy to see when one realises that Yt-1 depends on εt-1 and εt and εt-1 are correlated because both include μi If a probit specification were to be used the error structure used in the estimation would have to be corrected to account for this Standard statistical software packages such as Stata now have routines that make this correction for binary probits that include lagged values of the dependent variable in the presence of unobserved heterogeneity10 However these routines have not yet been extended to multinomial probits

Train (2003 p150) explains why choosing a mixed logit set-up including lagged dependent variables as explanatory variables on the right-hand side does not pose a problem in the presence of unobserved heterogeneity unlike the case of a probit specification The crux of it is that conditional on the unobserved heterogeneity μi the estimation boils down to estimating a standard multinomial logit modelmdashwith a single complication the unobserved heterogeneity μi on which it was conditioned needs to be lsquointegrated outrsquo Fortunately this can be done using standard numerical simulation making the mixed logit approach (in our case multinomial mixed logit due to a multiple of choices) an ideal candidate to model state dependence in the presence of unobserved heterogeneity In the next subsections we provide more detail on how the estimation works

9 Any parameter in a MMNL can be modelled as a random variable not just the constants10Note that when it is assumed that there is no correlation over time in the error terms eg

in the absence of unobserved heterogeneity including lagged dependent variables Yt-1 does not pose a problem

16 Annual transitions between labour market states for young Australians

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 16: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

lsquorandom parameter logitrsquo (RPL) model9 The random parameters in this case are the constants in the model This model has considerable advantages when analysing state dependence in the presence of unobserved heterogeneity and will be briefly discussed here

To discuss the modelrsquos structure and to illustrate why this specification is the best choice we first introduce some notation Let the dependent variable Yit denote the labour market state at time t for individual i Factors that influence the choice for Yit are denoted by Xit and lagged values of the dependent variable Yit-1 The explanatory variables in Xit as outlined previously imply a clear direction of the association between Xit and Yit That is Xit influences Yit but the reverse cannot hold The unobserved heterogeneitymdashin this study captured by an individual specific random effectmdashis denoted by μi

Why a logit specificationWhen it comes to modelling discrete choice variables the two main types of models are the logit and the probit models Usually the inclusion of lagged dependent variables creates an endogeneity problem but not so in a mixed logit framework Train (2003 p118) explains the issue in plain language Using our notation the Yit depends on Xit and the error term εit If thatrsquos the case then Yit-1 depends on Xit-1 and the error term εit-1 In the presence of unobserved heterogeneity the error term εit includes the unobserved heterogeneity component μi This μi component in the error terms makes these error terms correlated over time (Train 2003 p115) When one includes the lagged dependent variable Yt-1 on the right-hand side as an explanatory variable it will thus violate the independence assumption between the right-hand side variables and the error term in the equation Yt = γYt-1 + β Xt + εt This is easy to see when one realises that Yt-1 depends on εt-1 and εt and εt-1 are correlated because both include μi If a probit specification were to be used the error structure used in the estimation would have to be corrected to account for this Standard statistical software packages such as Stata now have routines that make this correction for binary probits that include lagged values of the dependent variable in the presence of unobserved heterogeneity10 However these routines have not yet been extended to multinomial probits

Train (2003 p150) explains why choosing a mixed logit set-up including lagged dependent variables as explanatory variables on the right-hand side does not pose a problem in the presence of unobserved heterogeneity unlike the case of a probit specification The crux of it is that conditional on the unobserved heterogeneity μi the estimation boils down to estimating a standard multinomial logit modelmdashwith a single complication the unobserved heterogeneity μi on which it was conditioned needs to be lsquointegrated outrsquo Fortunately this can be done using standard numerical simulation making the mixed logit approach (in our case multinomial mixed logit due to a multiple of choices) an ideal candidate to model state dependence in the presence of unobserved heterogeneity In the next subsections we provide more detail on how the estimation works

9 Any parameter in a MMNL can be modelled as a random variable not just the constants10Note that when it is assumed that there is no correlation over time in the error terms eg

in the absence of unobserved heterogeneity including lagged dependent variables Yt-1 does not pose a problem

16 Annual transitions between labour market states for young Australians

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 17: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Unobserved heterogeneity identification and estimationUsing our notation we briefly outline how unobserved characteristics enter the model how the model is identified and how it is estimated Let the probability that an individual i chooses a particular labour market state j in period t conditional on the unobserved random effect μi be denoted by

This is the familiar form for the probability in a standard multinomial logit model except it now includes Yit-1 and the unobserved heterogeneity terms in addition to the standard Xit There is only one complication that we need to clarify in order to match the tables with parameter estimates to the model as outlined here The previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time employment casual employment unemployment and not in the labour force However we combine full-time and part-time permanent employment into a single lsquopermanent employmentrsquo outcome for Yit (that is Y as the dependent variable) because each extra choice outcome will greatly complicate the analysis whereas an extra explanatory variable only has a relatively modest impact by comparison11 Also and only in the case of women the previous period labour market status (that is Y as an explanatory variable) can take on the values of permanent full-time employment permanent part-time or casual employment unemployment and not in the labour force That is in the case of women we combine casual and part-time permanent employment into a single state The more detailed specification was not supported by the data

The probability that we observe an individualrsquos history to be Yi = Yi1 Yi2 Yi3hellipYiT given unobserved heterogeneity μi is

where I() denotes the indicator function and T is the end of our data window Specifically the number of choices in our model (J) is four The number of time periods (T) is five These five years are the last five years in the LSAY Y95 data12 In a final step the unobserved heterogeneity μi needs to be integrated out of the above equation to get the unconditional probability Prob(Yi) We do so numerically by taking random draws from the distribution for μi evaluate Prob(Yi | μi) for each of these draws (which with the drawn μi given is a standard MNL probability) and then average over those to get Procircb(Yi)

11For instance in a model with four choices and 25 explanatory variables one needs to estimate (4-1)25 = 75 parameters (one of the choices needs to be normalised to zero as in any multinomial logit) By introducing an extra choice the number of parameters to be estimated is increased by another 25 to 100 An extra explanatory variable increases the number of parameters to be estimated by only three to 78

12Due to the inclusion of the labour market state in the previous period as an explanatory variable we lose one year (2002) Hence the period over which we model labour market dynamics is four years corresponding to waves 9 to 12 when the respondents were approximately 22 to 25 spanning the calendar years 2003 to 2006

NCVER 17

1

11

expProb( | )

exp

j it j it iit i J

m it m it im

X YY j

X Y

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 18: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

The model is thus estimated by simulated maximum likelihood with the pseudo log-likelihood to be maximised defined as

The unobserved random effects are assumed to be multivariate normal and correlated13 Further the random effects also assume two more conditions that were highlighted in the discussion of the research objectives earlier namely (1) that the unobserved heterogeneity is constant over time and (2) that these unobserved characteristics are not related to those that are observed It is important to spell these assumptions out as the validity of the results depends on them

Unfortunately these two assumptions are inevitable when including random effects It is important to realise that they are assumptions that cannot be directly tested Indeed if the variables are not observed it cannot be asserted that there is no relationshipmdashit has to be assumed The second assumption that the unobservables are uncorrelated with the observed explanatory variables is the most contentious even in the literature It is important however to not over-dramatise the issue Even in straightforward OLS regressions the assumption that the error is uncorrelated with the X variables is assumed to hold Because of assumption (2) when possible researchers tend to prefer a fixed effects specification over a random effects specification Unfortunately available standard software only has the capacity to estimate fixed-effects panel data models if the dependent variable is continuous or dichotomous but not discrete multinomial

The first assumption that unobserved heterogeneity is stable over time is really inescapable and on a par with the assumption that economic agents behave rationally If it were not even fixed effects approaches would break down

Discussing these assumptions may paint a somewhat sombre picture but in reality we explicitly account for two factors that in most typical studies are unobserved (and hence would be part of the unobserved heterogeneity) personality and ability Furthermore we model a relatively short four-year window in the post-qualification period where it is more reasonable to expect unobserved heterogeneity to be stable (recall that the assumption is that the unobserved heterogeneity terms are time-independent)

The initial conditions problemCommon to studies of labour market dynamics is the so-called initial condition problem The initial condition problem arises when lagged dependent variables are included and the data observation window starts (much) later than the first opportunity to observe the labour market state of interest Because current choices depend on lagged choices it follows that the first choice we observe depends on lagged choices also However those lagged choices are typically not observed for the first observation in

13Other assumptions are possible but normally distributed random effects are most commonly used Letting the random effects be correlated breaks the so-called Independence of Irrelevant Alternatives (IIA) assumption a rigidity that in the past has traditionally led to multinomial probit models being preferred over multinomial logit specifications

18 Annual transitions between labour market states for young Australians

iPseudoLL Procircb(Y )i

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 19: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

(nationally representative) longitudinal data sets therefore creating a bias in the estimates One possible approach to solve the initial conditions problems is based on a suggestion by Wooldridge (2005) In the Wooldridge approach the relationship between Yit and μi is accounted for by modelling the distribution of μi given Yi1 (that is the labour market state in the initial period) The assumption in Wooldridgersquos approach is that the distribution of the individual specific effects conditional on the exogenous individual characteristics is correctly specified14 Although other approaches to deal with the problem of initial conditions are available (Heckman 1981 Orme 2001) Monte Carlo simulations reported in Arulampalam and Stewart (2009) suggest

14As outlined in the previous discussion on unobserved heterogeneity we assume these to be normally distributed

NCVER 19

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 20: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

that all three estimators display similar results and perform satisfactorily We are therefore satisfied that our chosen estimation strategy is sensible Just to clarify in our specification the initial condition is the labour market state in 2002 (wave 8) on which we condition The labour market states that are modelled are those for 2003 to 2006 (waves 9 to 12)

20 Annual transitions between labour market states for young Australians

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 21: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

ResultsIn this section we discuss the results obtained from estimating the multinomial logit model as outlined earlier We report two types of results The first set of results consists of the parameter estimates of the model These show the statistical significance of the various factors described in the methodology section such as previous labour market state that were included in the model as explanatory variables The estimates in themselves however are not very informative For starters the multinomial logit model requires a normalisation for identification that sets all parameter estimates for the normalised outcome to zero The choice of the normalised outcome is arbitrary but it does affect the appearance of the table with the parameter estimates Parameter estimates are only obtained for the three remaining outcomes (We use lsquonot in the labour forcersquo as the normalised outcome) Similarly in the set of explanatory variables we need to choose certain characteristics as reference categories in order to avoid perfect multi-collinearity Again this only affects the appearance of the table with parameter estimates and is otherwise inconsequential

To overcome the interpretation issue of raw multinomial logit parameter model estimates we instead discuss a different set of results These results consist of simulations based on the parameter estimates The simulations have a very natural interpretation and show what we predict to happen to peoplersquos labour market status if we were to give them a different (chosen) set of characteristics They also show the impact on all labour market outcomes (that is not affected by a normalisation) as well as the impacts of all factors (again no normalisation issue here) We will discuss how these simulations work in practice in more detail The raw parameter estimates are reported for the sample as a whole and for males and females separately in the appendix

Results from scenario analysesIntroductionOne way to address the economic importance of the variables is to perform scenario analyses The process is rather straightforward and will be briefly discussed using the indicators for country of birth and parentsrsquo occupation class as examples The scenarios for the other variables all follow the same process

After estimating the model the parameter estimates are preserved Using the actual observed values for the variables in the data the probability of being in each of the four labour market states is predicted for each of the

NCVER 21

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 22: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

individuals in the sample These are then averaged over all individuals and form the base predictions with which the scenarios are compared The scenarios operate as follows for each individual the indicator for being born overseas is set equal to 0 This altered data set is then used to again predict the probability of being in each of the four labour market states for each of the respondents These are averaged over all respondents and can then be readily compared with the base predictions A second scenario is repeated but this time setting the indicator for being

22 Annual transitions between labour market states for young Australians

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 23: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

born overseas equal to 1 for all respondents resulting in a second distribution to be compared with the base prediction15

For the variables that indicate the parentsrsquo occupation class it is necessary to condition on three variables at once because if the parentsrsquo occupation class is upper-middle it cannot simultaneously be upper lower-middle or lower Thus simulating all individuals having parentsrsquo occupation class as upper-middle amounts to setting both remaining indicators for lower and lower-middle to zero simulating having parentsrsquo occupation class as lower amounts to setting that variable to 1 while setting the parentsrsquo occupation class as lower-middle and upper-middle equal to 0 and so on

In tables 1 to 4 we present the results (for males and females separately) of the lsquowhat ifrsquo scenarios categorised by the effects of personal characteristics (including ability) personality post-school qualifications and previous period labour market state and finally post-school qualifications and previous labour market state combined The latter addresses the role of post-school qualifications on the persistence of labour market states In each of the tables the top line displays the base predictions Each of the scenarios is expressed as deviations from these base predictions in percentage points Because the states are mutually exclusive and each individual has to be in exactly one of them the percentage point changes in each scenario have to sum to zero So for example if being married or in a de facto relationship increases the probability of being observed in permanent employment then this needs to be offset by an equally reduced probability of being in any of the other three labour market states How much each of these labour market states is affected in turn is an empirical matter That is not all are necessarily affected in equal proportion We report results for males and females separately following the same grouping as outlined earlier in the discussion of the factors that can help explain labour market status post-qualification

The role of personal characteristicsIn table 1 the results from the scenario analyses for the personal characteristics are displayed for males and females separately They relate to gender country of birth parental occupational status partner status and achievement scores for reading and maths There are too many numbers for them to be discussed in detail so we highlight the overall impression of them One thing that stands out is that all effects are relatively small The largest percentage point change to predicted probabilities is only in the order of 1ndash3 percentage points which is achieved by the scenarios that simulate respondents being married or in a de facto relationship Being partnered it seems increases the proportion of respondents working casually or being out of the labour force at the expense of being employed permanently but only for women For men the reverse holds true that is being partnered increases the proportion of respondents working permanently (or casually) at the expense of being out of the labour force Furthermore we also note that the magnitudes of the 15This is why they are called simulations as we are effectively comparing the predicted

probabilities for the actual data with the predicted probabilities when everyone would be Australian-born and when everyone would be foreign-born However this is precisely what would be wanted if one is interested in the role of country of birth on the predicted probabilities of being in a particular labour market state

NCVER 23

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 24: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

effects do not change much between the specification with and without controls for unobserved heterogeneity

24 Annual transitions between labour market states for young Australians

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 25: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Table 1a Males Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereBorn in Australia 002 -007 006 000 005 -010 005 -001

Born overseas -040 080 -041 000 -080 116 -042 006

Parentsrsquo occupation class ndash upper -155 -125 106 174 -132 -127 114 145

Parentsrsquo occupation class ndash upper-middle -045 115 -005 -065 -096 148 005 -057

Parentsrsquo occupation class ndash lower-middle 000 067 -031 -036 -017 085 -037 -031

Parentsrsquo occupation class ndash lower 162 -189 004 023 207 -231 -003 027

Not cohabitating -044 -015 028 031 -052 -011 034 030

Married or de facto 172 036 -103 -105 184 026 -118 -092

Ability score reading 10 (on scale of 0ndash20) 007 -034 -003 029 -005 -028 001 032

Ability score reading 15 (on scale of 0ndash20) 010 -023 -005 018 026 -038 -008 020

Ability score maths 10 (on scale of 0ndash20) 041 -035 014 -020 050 -044 012 -019

Ability score maths 15 (on scale of 0ndash20) -065 038 029 -002 -076 051 030 -006

Source LSAY 1995 cohort

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 26: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Table 1b Females Results from what-if scenarios based on parameter estimatesmdashPersonal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereBorn in Australia 012 -009 -002 -001 -001 000 -003 003

Born overseas -258 203 027 029 010 -005 040 -044

Parentsrsquo occupation class ndash upper 196 -031 -116 -049 280 -071 -221 012

Parentsrsquo occupation class ndash upper-middle -024 -042 020 045 -078 -022 037 063

Parentsrsquo occupation class ndash lower-middle 045 057 -057 -045 096 048 -085 -059

Parentsrsquo occupation class ndash lower -113 -043 110 046 -191 -033 181 043

Not cohabitating 232 -082 065 -215 127 -037 111 -201

Married or de facto -247 114 -067 200 -117 048 -121 190

Ability score reading 10 (on scale of 0ndash20) -016 027 043 -054 010 002 076 -088

Ability score reading 15 (on scale of 0ndash20) -021 019 -050 052 -031 030 -077 078

Ability score maths 10 (on scale of 0ndash20) 056 -117 -039 100 046 -095 -071 120

Ability score maths 15 (on scale of 0ndash20) -096 113 086 -104 -070 090 109 -128

Source LSAY 1995 cohort

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 27: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Personality traitsTable 2 displays the results for the scenario analyses related to personality traits Before discussing a number of them we note that in general the magnitudes of the effects for personality are larger than those for the personal characteristics with some of them in the order of 5ndash10 percentage points

For each of the personality traits we simulate rating possession of these traits from the highest level to the lowest level The one personality trait that appears to have a relatively strong effect for both males and females is being lsquoagreeablersquo Males who are less agreeable are more likely to be employed as a casual at the expense of working permanently The scenario in which all males would report the lowest level of being agreeable would reduce the proportion of men employed permanently by about five to six percentage points but increase the proportion employed casually by about seven percentage points16 Women who are less agreeable are more likely to be employed casually or to leave the labour market at the expense of working permanently Unlike the case for men the overall employment rate for women would be reduced in this case

Confidence seems to play a much bigger role for females than it does for males for whom it has little impact The scenario in which all women would report the lowest level of confidence would compared with the status quo reduce the proportion of women employed (either casually or permanently) by about four or five percentage points and lift the proportion of women not in the labour force by a similar margin17 Where men are more sensitive than women is popularity Being less popular means males are much less likely to be employed permanently and more likely to be employed in the three remaining states of casual employment unemployment or out of the labour force

16In table 2 the numbers for lsquoagreeable (least)rsquo point to a reduction in the proportion employed permanently of 490 percentage points in the specification without random effects and 574 percentage points in the specification with random effects respectively

17Using the numbers for lsquoconfidentrsquo in table 2 the reduction in the proportion employed is 584 percentage points (384 + 201) in the specification without random effects and 686 percentage points (545 + 141) in the specification with random effects

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 28: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Table 2a Males Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereAgreeable (most) 075 -182 036 071 090 -197 028 079

Agreeable (least) -490 686 -074 -121 -574 763 -063 -127

Open (most) -106 020 -022 108 -105 032 -026 099

Open (least) 202 -078 062 -186 206 -117 080 -169

Popular (most) 319 -163 -076 -080 387 -212 -081 -093

Popular (least) -849 413 196 240 -1116 596 195 325

Intellectual (most) -131 -026 044 113 -173 003 047 124

Intellectual (least) 173 046 -080 -140 242 -015 -086 -141

Calm (most) -127 128 -019 018 -147 158 -020 009

Calm (least) 277 -283 054 -048 293 -322 058 -028

Hard-working (most) -090 117 019 -046 -125 141 030 -047

Hard-working (least) 184 -335 -050 202 234 -365 -078 209

Outgoing (most) -031 064 -014 -019 -069 099 -015 -016

Outgoing (least) 082 -173 033 058 162 -243 035 047

Confident (most) -005 015 -046 036 014 -010 -049 045

Confident (least) -026 -046 157 -086 -096 029 165 -097

Source LSAY 1995 cohort

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 29: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Table 2b Females Results from what-if scenarios based on parameter estimatesmdashPersonality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereAgreeable (most) 181 -058 -010 -113 203 -060 -009 -135

Agreeable (least) -611 216 025 370 -729 243 -001 487

Open (most) -025 014 003 008 -029 021 -002 010

Open (least) 102 -063 -010 -030 119 -089 006 -037

Popular (most) -255 238 083 -066 -214 193 102 -081

Popular (least) 268 -254 -117 104 240 -211 -177 149

Intellectual (most) 024 -096 -051 124 -007 -075 -071 153

Intellectual (least) -210 304 119 -214 -131 232 139 -240

Calm (most) -056 -007 024 039 -101 014 046 041

Calm (least) 118 017 -048 -087 220 -034 -095 -091

Hard-working (most) -120 118 -020 022 -117 136 -054 036

Hard-working (least) 267 -257 070 -081 202 -249 172 -125

Outgoing (most) -004 013 -035 026 -021 035 -049 035

Outgoing (least) 005 -052 125 -078 056 -119 163 -101

Confident (most) 102 107 -029 -180 174 066 -067 -172

Confident (least) -383 -201 059 525 -545 -141 137 549

Source LSAY 1995 cohort

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 30: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Post-school qualifications and previous labour market stateTable 3 shows the results for the scenario analyses for post-school qualifications and previous period labour market states separately for males and females The magnitudes of the effects are much larger than those in the scenarios discussed up to now In particular previous period labour market state has a large impact as would be expected from the literature on labour market dynamics Furthermore the inclusion of random effects has a much larger effect here dampening the strong persistence that is found when not controlling for unobserved heterogeneity The latter finding on dampening the persistence is also a common result in the literature on labour market dynamics

In relation to the qualifications when it comes to the probability of being in work (that is permanent and casual combined) any qualification is better than none but the strongest boost for women is provided by a certificate IV closely followed by bachelor degree or better For males the employment effects are smaller but the biggest boost to overall employment is provided by a bachelor degree or better A scenario in which all men and women would have a certificate IV increases the employment probability for both relative to the status quo but it affects men and women differently For men it increases the proportion employed permanently but reduces the proportion employed as a casual For women the reverse holds

When interpreting the effects of the scenarios it is important to remember that they are expressed as percentage point changes from the base projections A fair comparison of the role of post-school qualifications would be to compare it with having no post-school qualifications For instance in the model without random effects not having any post-school qualifications reduces the probability of being in permanent employment by 58 percentage points for women and 18 percentage points for men all else being equal Having a bachelor degree or better increases it by 27 percentage points for men and 55 percentage points for women Therefore the net effect of having a bachelor degree or better vis-agrave-vis no qualification on the probability of being employed permanently is an increase of 45 percentage points for men (on a base of about 86) and 113 percentage points for women (on a base of about 85)

When it comes to the previous period labour market state it is shown that the most persistent states are casual employment for men and not being in the labour force for women These scenarios show the largest percentage point changes with respect to the base predictions Although the magnitudes of the persistence differ when unobserved heterogeneity is controlled for or not (with persistence deemed much larger in the case of the latter) the trade-offs operate the same way By that we mean that simulating a scenario whereby women are not in the labour force increases the predicted proportion of women not in the labour force next period almost completely at the expense of a reduced proportion of respondents employed in a permanent job This actually holds true for men as well Similarly simulating men in casual employment this period will lead to an increased predicted proportion of men employed casually next period but this comes exclusively at the expense of a reduced proportion of respondents working in permanent jobs

30 Annual transitions between labour market states for young Australians

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 31: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

As an example to bring the magnitudes of the simulated scenarios to life consider the following compared with not being in the labour force being full-time permanently employed in the previous period would increase the proportion of women permanently employed this period by a very large 386 percentage points in the specification without random effects and a more modest but still large 194 percentage points when unobserved heterogeneity is controlled for18 These numbers are large but this is a direct result of using a rather radical scenario comparison full-time permanent employment compared with not being in the labour force (in the previous period) For men the corresponding effects are smaller at 174 and 100 percentage points respectively

Comparing the scenario results for lagged labour market states as a group it is clear that not controlling for unobserved heterogeneity will severely exaggerate the influence (including persistence) of being out of the labour force for women and casual employment for men The effects of the other labour market states are much less sensitive to the inclusion of the random effects Furthermore the fact that lagged labour market states still have an impact after controlling for unobserved heterogeneity by the inclusion of random effects shows that part of the observed persistence should be considered genuine

18The number 3856 is obtained by comparing the difference between the numbers -3004 and +852 which are the effects in table 3b on the predicted proportion in permanent employment for the not in labour force and full-time permanent employment scenarios respectively Similarly the number 1938 is the difference between -1465 and +473

NCVER 31

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 32: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Table 3a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone wereNo post-school qualifications -182 -055 116 121 -168 -062 128 103

Certificate I or II 021 -101 -103 183 044 -102 -111 170

Certificate III -074 093 -021 002 -034 060 -015 -011

Certificate IV 309 -220 -142 053 311 -210 -173 072

Certificate ndash level unknown -299 412 -117 004 -454 587 -112 -021

Advanced diplomadip (includes assoc dip) -068 066 118 -115 -072 039 137 -104

Bachelor degree or higher 268 -101 -055 -111 295 -138 -062 -095

1 period lagged status ndash Not in labour force -988 -342 211 1119 -554 -154 248 460

1 period lagged status ndash FT permanent 749 -553 -087 -109 447 -288 -087 -072

1 period lagged status ndash PT permanent -137 -216 145 209 -441 049 175 217

1 period lagged status ndash Casual -4494 4452 138 -097 -1570 1513 144 -086

1 period lagged status ndash Unemployed -554 -103 284 373 -337 -174 198 313

Initial condition (2002) ndash Not in labour force -440 -084 312 212 -591 -160 371 381

Initial condition (2002) ndash FT permanent 161 -097 -072 008 352 -268 -087 002

Initial condition (2002) ndash PT permanent 408 -191 -103 -114 592 -362 -124 -106

Initial condition (2002) ndash Casual -846 784 -138 200 -2841 2721 -053 173

Initial condition (2002) ndash Unemployed -227 -238 466 -001 -370 -187 591 -033

Source LSAY 1995 cohort

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 33: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Table 3b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone wereNo post-school qualifications -577 136 127 314 -654 109 195 350

Certificate I or II -384 041 164 179 -470 034 252 184

Certificate III -209 304 -112 017 -040 204 -197 033

Certificate IV -220 905 -197 -488 031 756 -380 -407

Certificate ndash level unknown -379 000 150 229 -574 084 256 234

Advanced diplomadip (includes assoc dip) -083 -066 -023 172 -255 118 -056 193

Bachelor degree or higher 551 -135 -061 -355 560 -130 -077 -354

1 period lagged status ndash Not in labour force -3004 -144 425 2723 -1465 -138 492 1112

1 period lagged status ndash FT permanent 852 -247 -126 -479 473 -063 -130 -279

1 period lagged status ndash PT permanent or casual -371 625 -023 -231 -147 140 067 -060

1 period lagged status ndashUnemployed -659 -097 517 239 -598 166 493 -061

Initial condition (2002) ndash Not in labour force -494 010 183 300 -1250 022 450 778

Initial condition (2002) ndash FT permanent 401 -105 -123 -173 567 -139 -173 -255

Initial condition (2002) ndash PT permanent or casual -102 122 -039 019 -184 264 -065 -015

Initial condition (2002) ndash Unemployed -396 -340 436 300 -927 -257 874 310

Source LSAY 1995 cohort

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 34: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Post-school qualifications and previous labour market state combinedTable 4 presents the role of post-school qualifications and previous labour market state independently That is scenarios changed only one of these while keeping the other at observed values Table 4 reports results for scenarios that include both qualifications and lagged outcomes simultaneously This will provide an insight into the labour market dynamics for different levels of post-school qualifications We limit our discussion to the results based on the specification with controls for unobserved heterogeneity as omitting the random effects leads to exaggerated rates of persistence That is we focus on the genuine effect of past labour market states

The scenarios in table 4 compare findings for two undesirable labour market states that is not being in the labour force and unemployment and one less desirable labour market outcome casual employment In the specification for women casual employment was combined with part-time permanent employment because the data did not support the more detailed model for women19 By conducting a scenario analysis of being in each of these labour market states in the previous period while simultaneously setting a chosen level of post-school qualification we can study the effect of qualifications on the persistence in labour market outcomes

When simulating being out of the labour force in the previous period the effect on the probability of being permanently employed in the current period was found to be -55 percentage points for men (table 3a with random effects) and -146 percentage points for women (table 3b with random effects) When related to the post-school qualification level we see that this negative effect of the permanent employment probability ranges from almost no effect when combined with having a bachelor degree or higher (-02 percentage points for men -48 percentage points for women) to a maximum of -96 percentage points for men (-273 percentage points for women) if being out of the labour force in the previous period is compounded by having no post-school qualifications (table 4) In line with the independent effect of qualifications in table 4 having a bachelor degree for men and women or a certificate IV for women is shown to provide the best buffer against being out of the labour force becoming a persistent state

The story for the unemployment scenario is very much the same as that for being out of the labour force when it comes to the probability of being permanently employed The only difference is that the impacts are much smaller ranging from negligible when unemployment is combined with

19Strictly speaking each of these states could be considered the right choice under certain circumstances Because we restrict the sample to those who have completed their post-school qualifications the persons not in the labour force are not enrolled in some form of study leading to a qualification However they may have made a personal choice to become a homemaker are taking a gap year or have caring responsibilities Furthermore casual employment is to a large extent voluntary and does not by definition imply that it is a second-choice outcome Casual jobs are jobs with fewer benefits such as paid leave and sick leave but they are a flexible form of employment which may be attractive to young people and typically attract a wage premium to compensate for the absence of job security and lack of benefits Even unemployment if frictional can be beneficial in providing a means to search for a better job match

34 Annual transitions between labour market states for young Australians

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 35: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

having a bachelor degree or better to -69 percentage points for men when unemployment is combined with having no post-school qualifications and -146 percentage points for women (table 4)

It would be tempting to conclude that being unemployed is better than being out of the labour force because the effects of the former on the probability of being permanently employed are smaller There is an important gender effect here since for men the difference is not particularly large For men the effects on permanent employment of a bachelor degree are 14 and -02 percentage points and the effects of no qualifications are -69 and -96 percentage points depending on being related to unemployment or being out of the labour force For women the corresponding figures are -48 and 09 percentage points and -273 and -146 percentage points respectively Thus it is indeed the case that being unemployed is better than being out of the labour force when only looking at the probability of being permanently employed but what these results are really indicating is that not being in the labour market is a much more persistent state in particular for women That is why simulating being out of the labour force in the previous period is so damaging to the current permanent employment prospects of women the respondents are likely to still be out of the labour force in the current period Unemployment is also lsquostickyrsquo in a sense but much less so20

Finally on the results for lagged casual employment related to post-school qualifications for males table 4 shows that the effects on the two (current) employment states are large This is to be expected because we know from table 3 that casual employment is a highly persistent state for men and that in scenarios where lagged labour market status was set to be casual the increased probability to be employed casual this period came at the expense of the probability to be employed permanently But apart from the larger effects in absolute terms of lagged casual employment interacted with qualification levels on the two employment outcomes the difference between the qualification levels themselves is very similar to the case where qualification levels were related to lagged unemployment or lagged not in the labour force Using the results from table 4 for males the difference between bachelor degree or higher and no qualifications on the probability to be permanently employed is 94 percentage points when related to lagged not in the labour force (difference between -96 and -02) 83 percentage points when related to lagged unemployment (difference between -69 and 14) and 66 percentage points when related to lagged casual employment (difference between -171 and -105)

20When analysing the effects of being out of the labour force or unemployed in the previous period on the probability of being unemployed today it shows that being unemployed in the previous period is worse than being out of the labour force as one would expect This is true for both males and females

NCVER 35

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 36: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Table 4a Males Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8667 858 236 240 8726 789 265 220

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -1631 -429 394 1666 -957 -237 468 726

Certificate I or II -1452 -481 -005 1937 -649 -284 024 909

Certificate III -1023 -232 171 1084 -547 -085 219 413

Certificate IV -687 -569 -064 1320 -180 -382 -088 650

Certificate ndash level unknown -1258 183 -009 1085 -930 516 035 379

Advanced diplomadip (includes assoc dip) -700 -236 464 471 -562 -094 515 141

Bachelor degree or higher -175 -428 119 483 -017 -286 134 169

1 period lagged status ndash Unemployed AND

No post-school qualifications -979 -202 528 653 -687 -250 406 532

Certificate I or II -602 -267 055 814 -383 -295 -002 680

Certificate III -641 055 238 347 -338 -108 171 275

Certificate IV -033 -419 -028 480 036 -394 -106 464

Certificate ndash level unknown -994 628 026 340 -727 475 005 247

Advanced diplomadip (includes assoc dip) -612 015 540 057 -374 -121 437 058

Bachelor degree or higher 015 -245 164 066 138 -307 091 079

1 period lagged status ndash Casual AND

No post-school qualifications -4565 4253 335 -023 -1708 1387 343 -022

Certificate I or II -4095 4067 -006 035 -1289 1280 -020 030

Certificate III -5023 5077 065 -120 -1752 1742 112 -102

Certificate IV -3208 3299 -055 -035 -787 935 -117 -031

Certificate ndash level unknown -6042 6302 -110 -150 -2895 3073 -053 -125

Advanced diplomadip (includes assoc dip) -4968 4886 262 -179 -1837 1664 330 -158

Bachelor degree or higher -3931 4034 063 -166 -1045 1146 047 -148

Source LSAY 1995 cohort

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 37: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Table 4b Females Results from what-if scenarios based on parameter estimatesmdashQualifications and (select) past labour market status combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8542 386 293 778 8448 305 598 650

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -4709 -139 607 4242 -2730 -096 693 2133

Certificate I or II -4322 -175 738 3760 -2411 -135 832 1715

Certificate III -3408 041 155 3211 -1515 -010 141 1384

Certificate IV -1538 812 -009 735 -448 452 -115 111

Certificate ndashlevel unknown -4423 -202 688 3937 -2558 -109 824 1842

Advanced diplomadip (includes assoc dip) -3894 -224 329 3789 -2045 -078 341 1783

Bachelor degree or higher -1570 -214 363 1421 -482 -211 446 247

1 period lagged status ndash Unemployed AND

No post-school qualifications -1740 -025 895 870 -1464 303 825 336

Certificate I or II -1502 -096 987 611 -1260 197 916 147

Certificate III -705 149 223 333 -616 463 151 002

Certificate IV -262 779 -043 -473 -572 1249 -215 -463

Certificate ndash level unknown -1524 -126 950 700 -1388 266 921 201

Advanced diplomadip (includes assoc dip) -911 -166 474 603 -912 330 405 177

Bachelor degree or higher 187 -209 321 -299 089 -029 346 -405

1 period lagged status ndash PT permanent or casual AND

No post-school qualifications -1180 933 118 130 -923 282 288 354

Certificate I or II -843 697 154 -008 -700 180 353 166

Certificate III -959 1334 -137 -237 -236 415 -161 -019

Certificate IV -1758 2642 -229 -655 -287 1143 -383 -474

Certificate ndash level unknown -792 596 145 050 -826 247 356 223

Advanced diplomadip (includes assoc dip) -364 430 -036 -030 -466 297 003 167

Bachelor degree or higher 387 248 -099 -536 488 -047 -033 -408

Source LSAY 1995 cohort

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 38: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

ConclusionThis study examined year-on-year transitions of labour market states for young Australians who completed their post-school education and training The analysis models year-on-year transitions and does not follow the more commonly used approach of taking a (sub) sample of individuals at one point in time (for example first year after leaving school) and then modelling the labour market state say five years later Of key interest are the role that post-school qualifications play in transitions between labour market states and how much of the persistence can be assigned to the previous period labour market state per se and how much is attributable to unobserved heterogeneity In studies of labour market transitions it is often found that not controlling for such unobserved heterogeneity tends to overstate the role of past labour market states on current labour market states We find this to indeed be the case but mainly for two labour market states casual employment for men and being out of the labour force for women

Most studies on labour market dynamics for the adult labour market show that observed state dependence (occupying the same labour market state in consecutive periods) is much reduced when controlling for unobservable heterogeneity So while at first glance it appears that getting people into work and out of unemployment will lead to a large proportion of these people being employed in the future (lsquohigh state dependencersquo lsquowork leads to workrsquo etc) controlling for unobservables now changes the perspective and indicates that this lsquowork leads to workrsquo mechanism is only temporary and people will revert to their previous state Some will stay in employment of course but fewer than would appear at first glance The greater the roleimpact of unobservables (as captured here by random effects) the less permanent the effect of public policy will be To use a vintage toy analogy the greater the roleimpact of unobservables in driving transitions between labour market states the more the distribution of labour market states will resemble a roly-poly doll21 Policy will only be effective in temporarily altering the distribution of labour market states but preferences will return the distribution back towards the previous state unless the policy intervention is sustained

What we find in our study is that the observed state dependence is indeed reduced when controlling for unobservables In the context of the labour market states other than casual employment in the case of men and not in the labour force in the case of women the reduction in persistence is modest when compared with these latter two cases The direct implication is that public policy would still be effective since we do find there is genuine state dependence in labour market states but that the effect will be less than could be expected based on observed persistence in the (raw) data

21A roly-poly doll is defined as a tumbler toy When such a toy is pushed over it wobbles for a few moments while it seeks to return to an upright position

38 Annual transitions between labour market states for young Australians

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 39: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

That is a sizable chunk of the persistence is due to a common underlying process (unobserved heterogeneity) that will claw back much of the initial policy response over time In particular any policy that would momentarily increase the proportion of men working casually or would lead women to leave the labour market will have a lasting positive effect on the proportion of men working casually or women being out of the labour force in the subsequent periodmdashas we do find there is such a genuine impactmdashbut these will be much smaller than what might be expected if that certain je ne sais quoi captured by the controls for unobserved heterogeneity is ignored

Although previous period labour market states trump all other factors that are important in predicting current labour market states this does not mean the other factors should be dismissedmdashthese still matter A policy leading to all young Australian females collectively taking a gap year this period (that is place them in the lsquonot in labour forcersquo state or lsquounemploymentrsquo) would be completely offset in terms of impact on next periodrsquos labour market outcome if this policy resulted in these womenrsquos post-school qualification levels being lifted to at least a certificate IV And to give an example of a soft-skills policy lifting young Australian womenrsquos confidence to a point where they all respond as being very confident would increase their proportion in permanent employment by close to 1ndash2 percentage points (on top of a base prediction of about 85) and about one percentage point extra in casual employment while reducing unemployment and exits from the labour force

NCVER 39

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 40: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

ReferencesAinley J amp Corrigan M 2005 Participation in and progress through New

Apprenticeships LSAY research report no44 ACER Camberwell VicArulampalam W amp Stewart M 2009 lsquoSimplified implementation of the Heckman

estimator of the dynamic probit model and a comparison with alternative estimatorsrsquo Oxford Bulletin of Economics and Statistics vol71 no5 pp659ndash81

Curtis DD 2008 VET pathways taken by school leavers LSAY research report no52 ACER Camberwell Vic

Curtis DD amp McMillan J 2008 School non-completers Profiles and initial destinations LSAY research report no54 ACER Camberwell Vic

Dockery AM amp Norris K 1996 lsquoThe lsquorewardsrsquo for apprenticeship training in Australiarsquo Australian Bulletin of Labour vol22 no2 pp109ndash25

Fullarton S 2001 VET in schools Participation and pathways LSAY research report no21 ACER Camberwell Vic

Heckman JJ 1978 lsquoSimple statistical models for discrete panel data developed and applied to tests of the hypothesis of true state dependence against the hypothesis of spurious state dependencersquo Annales de lrsquoINSEE vol3031 pp227ndash69

mdashmdash1981 lsquoThe incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic processrsquo in Structural analysis of discrete data with econometric applications eds CF Manski and D McFadden MIT Press Cambridge

Long M 2004 How young people are faring 2004 Dusseldorp Skills Forum Sydney

Long M McKenzie P amp Sturman A 1996 Labour market participation and income consequences in TAFE ACER Camberwell Vic

Marks GN 2006 The transition to full-time work of young people who do not go to university LSAY research report no49 ACER Camberwell Vic

Marks GN Hillman KJ amp Beavis A 2003 Dynamics of the Australian youth labour market The 1975 cohort 1996ndash2000 LSAY research report no34 ACER Camberwell Vic

McMillan J amp Marks GN 2003 School leavers in Australia Profiles and pathways LSAY research report no31 ACER Camberwell Vic

McMillan J Rothman S amp Wernert N 2005 Non-apprenticeship VET courses Participation persistence and subsequent pathways LSAY research report no47 ACER Camberwell Vic

Nevile JW amp Saunders P 1998 lsquoGlobalisation and the return to education in Australiarsquo The Economic Record vol74 no226 pp279ndash85

Orme CD 2001 lsquoTwo-step inference in dynamic non-linear panel data modelsrsquo unpublished mimeo University of Manchester

Ryan C 2002 Longer-term outcomes for individuals completing vocational education and training qualification NCVER Adelaide

Ryan P 2001 lsquoFrom school-to-work A cross-national perspectiversquo Journal of Economic Literature vol39 pp34ndash92

Train KE 2003 Discrete choice methods with simulation Cambridge University Press Cambridge UK

40 Annual transitions between labour market states for young Australians

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 41: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Wooldridge JM 2005 lsquoSimple solutions to the initial conditions problem in dynamic nonlinear panel data models with unobserved heterogeneityrsquo Journal of Applied Econometrics vol20 pp39ndash54

NCVER 41

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 42: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

AppendixTable A1 Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 0716 -2272 -0071 1247 -2562 0239

[0524] [0165] [0963] [0515] [0351] [0915]

Male 0708 1108 0456 0865 1308 0546

[0000] [0000] [0068] [0003] [0001] [0131]

Born overseas ndash Mainly English speaking -0414 -0192 -0362 -0313 -0152 -0227

[0282] [0738] [0568] [0584] [0884] [0785]

Born overseas ndash Non-English speaking 0386 0242 0236 0481 0289 0271

[0397] [0680] [0676] [0490] [0782] [0713]

Parentsrsquo occupation class ndash Upper-middle

0322 0507 0402 0335 0624 0437

[0246] [0194] [0319] [0401] [0293] [0390]

Parentsrsquo occupation class ndash Lower-middle

0346 0630 0210 0414 0763 0307

[0179] [0084] [0580] [0285] [0175] [0528]

Parentsrsquo occupation class ndash Lower 0158 0168 0428 0182 0142 0488

[0551] [0666] [0272] [0646] [0808] [0354]

Married -0867 -0483 -1246 -1000 -0435 -1343[0000] [0067] [0000] [0000] [0259] [0001]

De facto -0249 -0351 -0710 -0128 -0257 -0659[0170] [0178] [0014] [0639] [0502] [0098]

Ability score reading (on scale of 0ndash20) -0256 -0326 -0505 -0357 -0467 -0584[0079] [0109] [0009] [0145] [0172] [0041]

Ability score reading ndash Squared 0009 0011 0017 0012 0016 0020[0099] [0137] [0018] [0168] [0202] [0058]

Ability score maths (on scale of 0ndash20) 0020 0139 0217 0100 0243 0286

[0881] [0480] [0257] [0608] [0490] [0280]

Ability score maths ndash Squared 0000 -0002 -0006 -0002 -0005 -0008

[0930] [0764] [0453] [0766] [0691] [0434]

Agreeable (scale 1ndash4) -0123 0250 -0154 -0160 0308 -0158

[0490] [0316] [0553] [0543] [0411] [0611]

Open (scale 1ndash4) 0148 0024 0174 0180 0005 0213

[0275] [0901] [0385] [0383] [0988] [0433]

Popular (scale 1ndash4) -0208 -0162 -0208 -0307 -0274 -0282

[0195] [0500] [0382] [0185] [0486] [0405]

Intellectual (scale 1ndash4) 0211 0309 0219 0285 0379 0265

[0178] [0171] [0347] [0287] [0317] [0432]

Calm (scale 1ndash4) 0085 -0096 0081 0105 -0167 0087

[0452] [0559] [0625] [0544] [0521] [0686]

Hardworking (scale 1ndash4) -0030 -0374 -0065 -0016 -0488 -0055

42 Annual transitions between labour market states for young Australians

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 43: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

[0813] [0038] [0723] [0933] [0099] [0823]

Outgoing (scale 1ndash4) 0002 -0101 0104 0014 -0209 0097

[0985] [0573] [0586] [0943] [0445] [0726]

Confident (scale 1ndash4) -0244 -0378 -0018 -0264 -0318 -0007

[0062] [0046] [0926] [0191] [0295] [0978]

Certificate I or II 0130 -0276 -0061 0135 -0331 -0106

[0590] [0472] [0872] [0713] [0606] [0828]

Certificate III 0500 0466 -0407 0664 0494 -0299

[0058] [0237] [0390] [0106] [0441] [0640]

Certificate IV 1135 1305 -0549 1259 1500 -0554

[0009] [0015] [0510] [0045] [0061] [0604]

Certificate ndash Level unknown 0242 0764 -0156 0270 1052 -0147

[0361] [0030] [0720] [0511] [0047] [0791]

Advanced dipdip (incl assoc dip) 0397 0137 0100 0457 0219 0133

[0120] [0737] [0798] [0275] [0722] [0814]

Bachelor degree or higher 1482 0936 0578 1792 1004 0765[0000] [0002] [0054] [0000] [0027] [0052]

initial condition (2002) ndash FT permanent 0919 0329 -0458 2065 0865 0206

[0000] [0433] [0208] [0000] [0279] [0701]

initial condition (2002) ndash PT permanent 0680 0413 -0408 1728 1024 0210

[0007] [0334] [0278] [0000] [0197] [0687]

initial condition (2002 ndash Casual 0091 0870 -1676 0218 2957 -1583

[0858] [0156] [0151] [0829] [0040] [0409]

initial condition (2002) ndash Unemployed 0036 -0705 0252 0615 -0462 0688

[0899] [0196] [0505] [0179] [0615] [0185]

1 period lagged status ndash FT permanent 3330 2619 1400 2383 2246 0854[0000] [0000] [0000] [0000] [0014] [0054]

1 period lagged status ndash PT permanent 2483 2389 1325 1520 2000 0777

[0000] [0000] [0000] [0000] [0033] [0112]

1 period lagged status ndash Casual 1998 5566 1303 1699 4472 1081

[0000] [0000] [0102] [0014] [0001] [0382]

1 period lagged status ndash Unemployed 1741 1913 1567 1385 1832 1283[0000] [0002] [0000] [0001] [0111] [0014]

Standard deviation of random effect (permanent) 1434[0000]

Standard deviation of random effect (casual) 1424[0097]

Standard deviation of random effect (unemployed) 0747[0076]

Correlation between random effects (casual permanent) -0208

Correlation between random effects (unemployed permanent) -0927

Correlation between random effects (casual unemployed) 0264

N (1482 individuals 4 waves) 5928 5928Log likelihood -2146255 -2126594Log likelihood (constants only) -3276128 -3276128R-squared 0345 0351R-squared (adjusted) 0341 0347Degrees of freedom 105 111

NCVER 43

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 44: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Note The reference (omitted) categories are female Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

44 Annual transitions between labour market states for young Australians

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 45: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Table A2 Males Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant -0340 -3600 -3865 0615 -3340 -2656

[0892] [0221] [0277] [0909] [0618] [0714]

Born overseas -0001 0198 -0222 -0047 0269 -0253

[0999] [0765] [0763] [0968] [0839] [0853]

Parentsrsquo occupation class ndash Upper-middle

0981 1501 0508 0935 1610 0504

[0037] [0010] [0413] [0274] [0161] [0647]

Parentsrsquo occupation class ndash Lower-middle

0829 1244 0226 0790 1317 0165

[0038] [0017] [0686] [0378] [0244] [0886]

Parentsrsquo occupation class ndash Lower 0577 0341 0120 0536 0136 0052

[0183] [0554] [0843] [0565] [0914] [0968]

Married or de facto 0809 0884 0030 0813 0868 -0024

[0058] [0061] [0960] [0343] [0331] [0984]

Ability score reading (on scale of 0ndash20) -0349 -0556 -0436 -0419 -0737 -0537

[0264] [0120] [0305] [0593] [0402] [0535]

Ability score reading ndash Squared 0014 0023 0018 0017 0030 0022

[0225] [0092] [0266] [0564] [0375] [0514]

Ability score maths (on scale of 0ndash20) 0087 0254 0511 0130 0347 0531

[0739] [0429] [0197] [0829] [0655] [0499]

Ability score maths ndash Squared -0004 -0010 -0021 -0006 -0013 -0021

[0655] [0425] [0159] [0795] [0667] [0468]

Agreeable (scale 1ndash4) 0329 0875 0165 0395 1069 0265

[0308] [0029] [0707] [0590] [0240] [0796]

Open (scale 1ndash4) 0680 0584 0763 0695 0537 0802

[0020] [0085] [0052] [0287] [0481] [0338]

Popular (scale 1ndash4) -0487 -0038 -0051 -0676 -0012 -0247

[0142] [0923] [0909] [0246] [0987] [0783]

Intellectual (scale 1ndash4) 0483 0519 0246 0585 0544 0342

[0156] [0200] [0596] [0490] [0601] [0769]

Calm (scale 1ndash4) 0130 -0241 0205 0098 -0400 0183

[0572] [0398] [0502] [0818] [0446] [0748]

Hardworking (scale 1ndash4) -0286 -0702 -0405 -0318 -0857 -0488

[0228] [0015] [0221] [0582] [0232] [0504]

Outgoing (scale 1ndash4) -0106 -0307 -0042 -0086 -0423 -0023

[0668] [0302] [0903] [0896] [0575] [0979]

Confident (scale 1ndash4) 0202 0157 0460 0273 0306 0530

[0447] [0628] [0202] [0616] [0640] [0465]

Certificate I or II -0113 -0265 -1173 -0151 -0284 -1208

[0807] [0651] [0179] [0876] [0808] [0497]

Certificate III 0494 0795 -0069 0549 0829 -0002

[0467] [0301] [0945] [0800] [0717] [1000]

Certificate IV 0368 -0162 -1119 0258 -0258 -1396

[0549] [0851] [0354] [0837] [0863] [0449]

Certificate ndash Level unknown 0465 1330 -0676 0528 1815 -0520

[0412] [0034] [0467] [0677] [0201] [0742]

Advanced dipdip (incl assoc dip) 1193 1438 1141 1182 1407 1155

[0122] [0097] [0204] [0519] [0486] [0513]

NCVER 45

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 46: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1255 1059 0424 1213 0915 0372

[0004] [0041] [0448] [0147] [0400] [0736]

Initial condition (2002) ndash FT permanent 0777 0640 -0566 1341 0952 -0168

[0126] [0366] [0415] [0283] [0682] [0916]

Initial condition (2002) ndash PT permanent 1534 1157 -0072 2119 1422 0326

[0014] [0152] [0931] [0113] [0511] [0844]

Initial condition (2002) ndash Casual -0003 1206 -1676 0054 3265 -0903

[0997] [0197] [0232] [0976] [0249] [0780]

Initial condition (2002) ndash Unemployed 0720 0361 0926 1379 1257 1651

[0227] [0671] [0212] [0358] [0619] [0397]

1 period lagged status ndash FT permanent 2672 1917 1294 1893 1379 0587

[0000] [0012] [0052] [0111] [0617] [0667]

1 period lagged status ndash PT permanent 1296 1412 0994 0526 0915 0317

[0011] [0094] [0189] [0681] [0737] [0836]

1 period lagged status ndash Casual 1736 4953 2093 1582 3801 1362

[0052] [0000] [0081] [0598] [0382] [0681]

1 period lagged status ndash Unemployed 0913 1251 0997 0326 0238 0175

[0111] [0193] [0196] [0792] [0935] [0918]

Standard deviation of random effect (permanent) 1240

[0150]

Standard deviation of random effect (casual) 1640[0002]

Standard deviation of random effect (unemployed) 1195

[0246]

Correlation between random effects (casual permanent) 0331

Correlation between random effects (unemployed permanent) 0779

Correlation between random effects (casual unemployed) -0333

N (647 individuals 4 waves) 2588 2588

Log likelihood -87908 -87173

Log likelihood (constants only) -132610 -132610

R-squared 034 034

R-squared (adjusted) 033 033

Degrees of freedom 96 102

Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour force

Source LSAY 1995 cohort

46 Annual transitions between labour market states for young Australians

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 47: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Table A3 Females Parameter estimates of (mixed) multinomial logits with and without (correlated) random effects

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Constant 2176 -2140 1492 2947 -7573 1899

[0104] [0338] [0405] [0202] [0268] [0484]

Born overseas -0113 0442 0038 0099 0066 0176

[0758] [0400] [0944] [0863] [0966] [0813]

Parentsrsquo occupation class ndash Upper-middle

-0266 -0267 0418 -0274 0181 0521

[0502] [0626] [0500] [0640] [0896] [0530]

Parentsrsquo occupation class ndash Lower-middle

-0050 0220 0286 0080 0897 0515

[0894] [0667] [0630] [0885] [0525] [0539]

Parentsrsquo occupation class ndash Lower -0303 -0296 0676 -0299 0128 0800

[0427] [0582] [0259] [0596] [0932] [0335]

Married or de facto -0862 -0226 -1163 -0857 -0312 -1192[0000] [0382] [0000] [0001] [0528] [0002]

Ability score reading (on scale of 0ndash20) -0199 0048 -0538 -0300 0390 -0610

[0235] [0867] [0015] [0314] [0696] [0112]

Ability score reading ndash Squared 0006 -0004 0017 0009 -0017 0019

[0308] [0733] [0039] [0389] [0624] [0168]

Ability score maths (on scale of 0ndash20) -0164 -0080 -0004 -0144 0024 0013

[0348] [0770] [0986] [0624] [0981] [0974]

Ability score maths ndash Squared 0009 0012 0006 0009 0012 0006

[0200] [0282] [0535] [0439] [0740] [0701]

Agreeable (scale 1ndash4) -0337 -0062 -0212 -0469 0119 -0321

[0124] [0842] [0532] [0183] [0887] [0439]

Open (scale 1ndash4) 0034 -0052 0009 0047 -0215 0033

[0833] [0830] [0973] [0854] [0772] [0928]

Popular (scale 1ndash4) -0065 -0664 -0352 -0103 -1145 -0356

[0733] [0027] [0232] [0748] [0247] [0472]

Intellectual (scale 1ndash4) 0214 0557 0389 0294 0852 0424

[0243] [0055] [0160] [0358] [0352] [0324]

Calm (scale 1ndash4) 0105 0119 -0012 0144 0013 -0010

[0436] [0544] [0956] [0528] [0983] [0974]

Hardworking (scale 1ndash4) 0085 -0429 0164 0129 -1054 0235

[0581] [0072] [0479] [0581] [0191] [0534]

Outgoing (scale 1ndash4) 0058 -0010 0227 0091 -0269 0216

[0708] [0968] [0354] [0701] [0715] [0601]

Confident (scale 1ndash4) -0442 -0772 -0253 -0534 -0959 -0269

[0005] [0001] [0308] [0029] [0199] [0423]

Certificate I or II 0217 -0044 0259 0280 -0137 0290

[0450] [0931] [0557] [0517] [0934] [0635]

Certificate III 0573 0841 -0461 0774 1005 -0346

[0056] [0069] [0414] [0126] [0507] [0671]

Certificate IV 1969 3011 0112 2260 4057 0205

[0003] [0000] [0926] [0033] [0017] [0917]

Certificate ndash Level unknown 0148 -0225 0163 0175 0040 0232

[0641] [0668] [0749] [0739] [0978] [0762]

Advanced dipdip (incl assoc dip) 0311 -0312 -0274 0391 0316 -0250

[0272] [0547] [0589] [0384] [0834] [0746]

NCVER 47

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 48: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Without random effects With (correlated) random effects

Permanent Casual Unemployed Permanent Casual Unemployed

Bachelor degree or higher 1554 0576 0586 1960 0091 0885

[0000] [0106] [0122] [0000] [0927] [0109]

Initial condition (2002) ndash FT permanent 1019 0507 -0257 2223 0562 0408

[0000] [0347] [0568] [0000] [0715] [0580]

Initial condition (2002) ndash PT permanent or casual

0531 0747 -0227 1429 2177 0173

[0060] [0143] [0599] [0006] [0145] [0793]

Initial condition (2002) ndash Unemployed -0008 -2302 0436 0553 -2586 0860

[0982] [0047] [0349] [0320] [0310] [0215]

1 period lagged status ndash FT permanent 3348 2215 1174 2486 2625 0686

[0000] [0001] [0005] [0000] [0118] [0213]

1 period lagged status ndash PT permanent or casual

2559 3654 1021 1758 3171 0622

[0000] [0000] [0018] [0000] [0056] [0288]

1 period lagged status ndash Unemployed 1836 1636 1514 1578 3234 1228[0000] [0085] [0001] [0002] [0175] [0045]

Standard deviation of random effect (permanent) 1356[0000]

Standard deviation of random effect (casual) 3237[0001]

Standard deviation of random effect (unemployed) 0759

[0162]

Correlation between random effects (casual permanent) 0077

Correlation between random effects (unemployed permanent) -0837

Correlation between random effects (casual unemployed) 0480

N (835 individuals 4 waves) 3340 3340

Log likelihood -132773 -124118

Log likelihood (constants only) -187900 -187900

R-squared 029 034

R-squared (adjusted) 029 033

Degrees of freedom 90 96Note The reference (omitted) categories are Australian born parentsrsquo occupation class ndash upper no post-school

qualifications initial condition (2002) ndash not in labour force and 1 period lagged status ndash not in labour forceSource LSAY 1995 cohort

48 Annual transitions between labour market states for young Australians

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 49: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Table A4 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personal characteristics ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereMale 107 072 -020 -159 110 091 -052 -149

Female -041 -069 021 089 -051 -082 050 083

Born in Australia -001 000 002 -001 -004 001 003 001

Born overseas ndash Mainly English speaking -218 061 -004 161 -144 041 011 093

Born overseas ndash Non-English speaking 173 -034 -020 -119 196 -039 -048 -109

Parentsrsquo occupation class ndash Upper -031 -055 -018 104 001 -067 -040 106

Parentsrsquo occupation class ndash Upper-middle 011 005 011 -026 -021 023 014 -016

Parentsrsquo occupation class ndash Lower-middle 029 039 -039 -029 043 054 -070 -027

Parentsrsquo occupation class ndash Lower -035 -052 057 030 -049 -086 112 023

Not cohabitating 062 -009 046 -098 024 -023 091 -092

Married -295 100 -078 273 -283 161 -155 277

De facto 100 -039 -068 007 195 -042 -132 -021

Ability score reading 10 (on scale of 0ndash20) -010 009 023 -022 -022 015 042 -035

Ability score reading 15 (on scale of 0ndash20) -001 -009 -031 041 010 -013 -049 053

Ability score maths 10 (on scale of 0ndash20) 029 -043 -018 031 058 -058 -034 033

Ability score maths 15 (on scale of 0ndash20) -037 035 041 -039 -055 047 059 -051

Source LSAY 1995 cohort

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 50: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Table A5 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Personality ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereAgreeable (most) 104 -085 013 -032 114 -106 024 -031

Agreeable (least) -358 307 -033 084 -402 396 -074 080

Open (most) -046 021 -009 034 -043 031 -022 034

Open (least) 161 -079 030 -112 146 -114 077 -109

Popular (most) 076 -010 009 -074 078 -003 010 -085

Popular (least) -166 019 -016 163 -185 003 -026 208

Intellectual (most) -042 -033 -009 084 -050 -035 -008 093

Intellectual (least) 052 071 016 -139 063 074 007 -143

Calm (most) -065 045 -004 024 -082 071 -010 021

Calm (least) 153 -105 008 -056 184 -154 020 -050

Hardworking (most) -062 070 005 -013 -085 099 -002 -012

Hardworking (least) 179 -206 -015 042 230 -262 -005 037

Outgoing (most) -004 022 -021 002 -019 050 -036 004

Outgoing (least) 016 -070 061 -006 053 -147 106 -012

Confident (most) 075 038 -042 -072 110 027 -083 -054

Confident (least) -213 -100 114 199 -300 -074 224 150

Source LSAY 1995 cohort

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 51: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Table A6 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and past labour market status ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone wereNo post-school qualifications -383 033 120 230 -446 026 216 204

Certificate I or II -173 -081 070 184 -211 -097 124 183

Certificate III 005 044 -085 036 087 034 -152 031

Certificate IV 207 134 -174 -167 243 234 -369 -108

Certificate ndash Level unknown -370 249 007 114 -472 387 -016 101

Advanced dip dip (includes assoc dip) -080 -033 050 063 -140 -020 101 058

Bachelor degree or higher 406 -091 -060 -255 444 -123 -101 -220

1 period lagged status ndash Not in labour force -2540 -315 343 2511 -1032 -272 432 871

1 period lagged status ndash FT permanent 803 -373 -110 -320 452 -165 -122 -165

1 period lagged status ndash PT permanent 242 -215 046 -072 -154 -022 150 026

1 period lagged status ndash Casual -4545 4781 -032 -203 -1334 1488 -012 -142

1 period lagged status ndash Unemployed -605 -151 453 303 -481 -082 541 023

Initial condition (2002) ndash Not in labour force -565 067 268 230 -1194 030 615 549

Initial condition (2002) ndash FT permanent 301 -098 -103 -100 485 -200 -154 -131

Initial condition (2002) ndash PT permanent 083 003 -058 -028 195 -066 -060 -069

Initial condition (2002) ndash Casual -543 513 -173 202 -2342 2518 -441 265

Initial condition (2002) ndash Unemployed -442 -182 406 217 -788 -293 868 213

Source LSAY 1995 cohort

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 52: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Table A7 Results from lsquowhat-ifrsquo scenarios based on parameter estimates Qualifications and (select) past labour market status ndash combined ()

Specification without random effects Specification with (correlated) random effectsLabour market state (dependent variable) Permanent Casual Unemployed Not in

labour forcePermanent Casual Unemployed Not in

labour force

Predicted labour market states at observed data values 8597 592 268 543 8376 594 620 410

Changes to these base predictions if everyone were1 period lagged status ndash Not in labour force AND

No post-school qualifications -3916 -334 513 3737 -1901 -289 675 1515

Certificate I or II -3564 -407 431 3540 -1627 -365 540 1453

Certificate III -2729 -281 143 2867 -951 -247 171 1027

Certificate IV -1573 -139 -034 1746 -397 -048 -157 601

Certificate ndash Level unknown -3449 -119 321 3247 -1632 000 383 1250

Advanced dip dip (includes assoc dip) -3060 -357 432 2985 -1355 -300 559 1097

Bachelor degree or higher -1130 -354 269 1214 -218 -334 332 219

1 period lagged status ndash Unemployed AND

No post-school qualifications -1428 -126 778 775 -1099 -077 907 269

Certificate I or II -1067 -264 648 683 -807 -198 756 250

Certificate III -530 -087 229 387 -318 -035 280 072

Certificate IV -037 060 -019 -005 -007 222 -121 -094

Certificate ndash Level unknown -1219 204 474 541 -1004 340 517 148

Advanced dipdip (includes assoc dip) -829 -201 590 439 -697 -113 714 096

Bachelor degree or higher 138 -261 276 -153 076 -203 352 -225

1 period lagged status ndash Casual AND

No post-school qualifications -5123 5078 063 -018 -1842 1613 204 025

Certificate I or II -4295 4201 069 025 -1389 1204 157 029

Certificate III -4896 5217 -122 -198 -1325 1631 -182 -125

Certificate IV -5219 5799 -206 -374 -1523 2193 -421 -250

Certificate ndash Level unknown -5995 6321 -097 -229 -2396 2633 -126 -111

Advanced dipdip (includes assoc dip) -4513 4616 023 -125 -1464 1460 094 -090

Bachelor degree or higher -3704 4151 -076 -371 -695 1097 -107 -295

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix
Page 53: National Centre for Vocational Education Research - Annual …€¦  · Web view2016-03-29 · In other words, an event that would lead to all of Australia’s youth collectively

Source LSAY 1995 cohort

  • Annual transitions between labour market states for young Australians
    • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research
    • Gary Marks Australian Council for Educational Research
      • Publisherrsquos note
          • About the research
            • Annual transitions between labour market states for young Australians
            • Hielke Buddelmeyer Melbourne Institute of Applied Economic and Social Research and Gary Marks Australian Council for Educational Research
              • Contents
              • Tables
              • Executive summary
              • Introduction
                • Objective of the research
                • Previous studies
                  • Methodology
                    • The data
                    • Defining mutually exclusive labour market states
                    • Factors that can help explain labour market status post‑qualification
                      • Previous period labour market state
                      • Details of post-school qualifications
                      • Personal characteristics of the respondent
                      • Reading and maths ability
                      • Personality traits
                        • The general structure of the model
                          • Why a logit specification
                          • Unobserved heterogeneity identification and estimation
                          • The initial conditions problem
                              • Results
                                • Results from scenario analyses
                                  • Introduction
                                  • The role of personal characteristics
                                  • Personality traits
                                  • Post-school qualifications and previous labour market state
                                  • Post-school qualifications and previous labour market state combined
                                      • Conclusion
                                      • References
                                      • Appendix