2

1 Introduction1

Reservation wages are the minimum wage level that individuals are willing to accept

to work. As such, they can be interpreted as a measure of individuals’ willingness to work

and as a summary indicator of individuals’ labour supply behaviour. Reservation wages can

the be a useful tool for the analysis of labour supply trends, for comparing the labour supply

behaviour of different groups in the population or to evaluate relevant policy issues, like the

effect of unemployment benefits on individuals’ job search behaviour.

Empirically, two main approaches to analysing reservation wages can be identified.

The first approach attempts to infer the main characteristics of reservation wages from the

observed wage offer distribution, or from the duration of the unemployment spells, or by

introducing hypotheses about the matching process between firms and workers (reviews of

these studies are Wolpin, 1995, and Mortensen and Pissarides, 1999). The second focuses on

self-reported reservation wages, collected by many surveys both in US and in Europe. More

precisely, these surveys ask to job seekers to report the minimum net salary they would

accept to give up search and start working. This information has been interpreted as the job

seekers’ reservation wage and has been used, for instance, by Holzer (1985) to compare the

reservation wages of white and black unemployed youth. Similarly, Jones (1989) focuses on

the relationship between reservation wages, market wages and measures related to the cost

of unemployment and benefits in Great Britain. Self-reported reservation wages have also

been used to test the main findings of the job search theory (Lancaster and Chesher, 1983)

and the relationship between unemployment duration and reservation wages (Jones, 1988).

More recently, Prasad (2003) studies the relationship between reservation wages and

macroeconomic variables in Germany, including the generosity of the unemployment

compensation system.

Both these approaches can be criticised on various grounds. On the one hand, the

indirect identification of reservation wages is based on strong assumptions on the shape of

1 We thank Gianna Barbieri, Piero Cipollone and Roberto Torrini and the participants at the EALE conference- Lisbon 2004 for helpful comments. The views expressed in this paper are those of the authors and to not

3

the wage offer distribution (for a critical discussion, see Flinn and Heckman, 1982), the job

offer arrival rate (see, for example, Kiefer and Neumann, 1979) or the way in which firms

set their wage level (see, for example, van der Berg and Ridder, 1998). On the other hand,

when analysing self-reported reservation wages one must be willing to accept the reliability

of these self-reported data.

In this paper we follow this last approach and we look at self-reported reservation

wages collected by the Italian LFS. The Italian data are characterised by a counterintuitive

and persistent relationship between reservation wages and the level of unemployment.

Reservation wages of people living in the Southern part of Italy (or “Mezzogiorno”) are

higher than those of people residing in the Northern and Central regions, in spite of the fact

in the Southern regions the unemployment rate is 4 time higher than in the rest of the

country. One might argue that the lower willingness of people living in the South might

itself be among the causes of the high unemployment prevailing in the area (see for example

Boeri and Garibaldi, 2000). A similar conclusion was drawn for instance by Holzer (1986)

for black youth reservation wages. He finds that, when controlling for the unemployment

rate and personal characteristics, the reservation wages of black youth are relatively higher

than those of white youth. However, in Italy the reservation wages of people living in the

South are higher not only in relative, but also in absolute terms. This evidence has been

usually considered as rather paradoxical and invalidating the reliability of the variable as

collected by the LFS. In what follows we will refer to this phenomenon as the “reservation

wage paradox”.

In spite of this paradoxical pattern, since 1993 the Italian National Statistical Office

(Istat) collects data on self-reported reservation wages for around 10,000 unemployed job

seekers each years. Using this large sample and this relatively long time series, we propose a

simple empirical model for interpreting self-reported reservation wages. Similarly to a large

part of the empirical literature we refer to the job search theory as our theoretical framework.

First, we show that the Italian “reservation wage paradox” is mainly determined by a

selection process due to the way in which self-reported reservation wages are usually

involve the responsibility of the Institutions of affiliation.

4

collected. This selection bias, far from being specific to the Italian data, may potentially

affect all the surveys collecting reservation wages, since they follow a similar scheme.

Second, we show that, when selection is taken into account, individuals determine

their reservation wages consistently with the market factors that the theory predicts as

influencing reservation wages. They are also consistent with future movements in the labour

market, being reservation wages a useful tool to predict transition probabilities from the non-

working conditions into employment.

The paper is organised as follows. Section 2 describes the characteristics of the data

and presents some statistics illustrating the “reservation wage paradox”. In Section 3 we

present our empirical model and we identify the determinants of the “reservation wage

paradox”. In Section 4 we describe the relationship between self-reported reservation wages

and the factors affecting the expected utility from employment and the relationship between

reservation wages and the transition probabilities into employment. Finally, Section 5

concludes.

2 Data and evidence on the Italian labour market

2.1 The data

Data on self-reported reservation wages are collected with different regularity by

various surveys. In Europe, for example, they are collected by the European Community

Household Panel (ECHP), the German Socio-Economic Household Panel, The British

Household Panel Survey and the Italian Labour Force Survey (LFS). In US they are

sporadically collected by the Current Population Survey and the National Longitudinal

Youth Survey.

In collecting reservation wages all these surveys follow a similar scheme. First, all

working-age individuals –both employed and not employed- are asked whether they are

looking for a job. If they answer affirmatively, they are further required to indicate the

characteristics of the desired job. Finally they have to report the minimum net monthly

5

salary they would accept for a job with the desired characteristics.2

In this paper we focus on the reservation wages collected by the Italian LFS. The

Italian data have very suitable characteristics. First, a continuos series of data on reservation

wages is available since 1993. This allows us to study trends in reported reservation wages

and to verify over the time the robustness of our findings. Second, differently from the other

European longitudinal surveys, the LFS sample size is very high and it amounts each year to

around 10,000 job seekers. Finally, thank to the sample design, the Italian LFS has a

longitudinal dimension and it allows us to study the relationship between reservation wages

and future labour market movements.

More in detail, the Italian LFS is conducted quarterly (in January, April, July and

October) by the Italian national statistical office (Istat) and it involves almost 80,000

households and 200,000 individuals. Households participate to the survey according to a

rotating scheme of the type 2-2-2. Because of this scheme half of the original sample is

interviewed for two subsequent waves, left out of the sample for two other waves and

interviewed again for two other surveys. People can then be followed for a period up to 6

quarters (for a critical description of the LFS see Cannari and Sestito, 1995, and Paggiaro

and Torelli, 1999). Thanks to the survey design, Istat produces two type of public user LFS

files: a cross section and a yearly matched panel. The first type is available for the period

October 1992-October 2003 and it reports information on individual labour market

behaviour at the time of the interview. The second type of files refers to the period April

1993 - April 2002 and report the labour market conditions of individuals in April of each

year and 12 months after. Both files contain detailed data on job search behaviour and

reservation wages. In this paper we use the data collected in April from 1993 to 2002. Data

are used in both the cross-sectional and longitudinal version.3

Before reporting their reservation wages, job seekers are required to specify a wide set

2 Reservation wages are observed for both the ILO unemployed and people who look for employment but they

do not sought employment during the four week preceding the interview.3 From January 2004 Istat has launched a new version of the Italian LFS. Data on reservation wages are not

comparable with those collected before 2003, because of changes in the survey questionnaire. The quality ofthe data collected in 2003 however is lower than the quality of data of the preceding surveys and they are notused in this paper.

6

of job desired characteristics. First, job seekers have to state whether they are looking for a

dependent or a self employment. Possible answers are: (1) permanent dependent

employment, (2) fixed term job - training program, (3) fixed term job - other, (4) any

dependent employment; (5) self-employment. Second they have to indicate whether they are

looking (1) exclusively for a full time employment, (2) exclusively for a part-time

employment, (3) preferably for a full time employment, (4) preferably for a part-time

employment (5) for any working time. Finally, job seekers are required to indicate the

desired location of the job ((1) the same municipality of living; (2) within a daily commuting

distance; (3) everywhere in Italy; (4) everywhere in Italy and abroad). Individuals are then

required to report their reservation wages conditionally on the declared preferences, which

obviously affect the observed distribution. As a consequence, modelling the reservation

wages of individuals looking for a wide set of different jobs can be particularly difficult.

In this paper we look at a quite homogeneous subset of job seekers, selected not only

by looking at the features of the data but also at the main findings of job search theory.

First, similarly to other papers on reservation wages (see; for example; Holzer, 1986),

we exclude employed job seekers, since their reservation wage depends also on the current

wage level, but this information is not collected by the LFS.4 Second, consider Table 1 that

reports the average reservation wage of non-employed job seekers by type of desired job and

working time (standard deviation are reported within brackets). To take into account

differences in the desired working hours, reservation wages have been divided by the

average number of hours worked by full time and part-time employees in 2000, i.e. by 40

weekly hours for full time workers, 24 for part time workers, 39 hours per week for people

with no preferences about the working time, equal to the average hours worked over the

sample of whole employees.

Confirming the hypothesis that the preferences on job characteristics affect the

reported reservation wages, the average reservation wage of people trying to set up a self-

employment is around 20 percentage points higher than the average reservation wage of

people looking for a dependent position. People looking for a part time job are characterised

4 From 2004 wages will instead be regularly collected.

7

by higher reservation wages, while people with no preferences about the type of job are

available to accept lower wages.

As a first approximation, the data suggest that the effects of heterogeneity of the

preferences can be partially offset by selecting people looking for a dependent employment,

for a full time position or available to accept any working time. These job seekers amount on

average to around 80 per cent of total Italian job seekers. This sample selection can be

justified not only by looking at mean reservation wages, but also from a theoretical

viewpoint. In standard job search models, both the job offer arrival rate and the wage offer

distribution are exogenous to the individual. This type of model cannot be adapted to self-

employment, where both these variables cannot be considered as truly exogenous.

Analogously, part-time job seekers can be excluded because, even if in principle it is

possible to estimate their “full-time equivalent” reservation wage (or to model their hourly

minimum net monthly salary). Nevertheless, their reservation wages would be affected by a

different set of parameters determining the expected value of employment. In particular, in

Italy, where part-time workers amount to 10 per cent of the employees, the probability to

receive a part-time job offer is lower than the probability to receive a full-time offer.

Instead, we do not exclude individuals on the basis of the desired job location, but in

the empirical analysis presented in the next sections we will explicitly model the effects on

reservation wages of the availability (and costs) of moving.

Summing, up, in what follows we analyse the reservation wages of non-employed job

seekers who are neither looking for a self employment nor looking for a part time job.

Table 2 reports the average reservation wages of our sample, by area of residence

(North-Centre and South5) sex and age group (15-24, 25 and over). The Table also reports

the unemployment rate, the employment rate and the share of job seekers in total non-

working population. Reservation wages are the mean of the net monthly salary, expressed in

euro. Confirming the existence of the “reservation wage paradox”, reservation wages are

5 The Southern part of Italy includes the following regions: Campania, Molise, Puglia, Basilicata, Calabria,

Sicily and Sardinia. The North-Centre includes: Piedmont, Val d’Aosta, Lombardy, Trentino Alto Adige,Veneto, Friuli Venezia Giulia, Liguria, Emilia Romagna, Tuscany, Lazio, Marche, Umbria.

8

significantly higher in the Southern part of Italy than in the North-Centre, except for men

aged at least 25. For this subgroup we cannot reject the hypothesis that average reservation

wage are statistically similar. However, since people living in the Southern part of the

country are characterised by a lower employment probability and a remarkably higher

unemployment rate, one might expect a positive (not a negative, neither a nil) North-South

differential.

2.2 Decomposing observed reservation wages

The analysis of sample averages cannot help to determine whether the differences in

the average reservation wages are due to differences in population composition and/or to

differences in unobservable characteristics. For instance, the negative reservation wage

differential could be due to a large portion of men among the job seekers in the

Mezzogiorno, or a large share of people with a higher education. Alternatively, the

difference could be caused by unobservable characteristics, related to a lower willingness to

work of people residing in the Southern regions.

To evaluate the relative importance of the two factors, in this section we have carried

out a simple Oaxaca’s decomposition (Oaxaca, 1973). First, for each area A (A=N for North-

Centre and A=S for South), we have estimated the equation AAAA Xw εβ += '* where *Aw is the

log of reservation wages and AX is a matrix of independent variables. Then we have

calculated the following statistics:

(1) SSNNSNSN XXXww )'ˆˆ(ˆ)'(** βββ −−−=−

where *Aw is the average (log) reservation wage in area A=N,S, AX is the vector of the

average independent variables in area A and Aβ̂ is the vector of estimated OLS coefficients.

The first term of the right hand side is the contribution of the observed characteristics to the

reservation wage differential. The second term measures the effect of differences in labour

supply behaviour of the two groups. This is calculated using the coefficients of job seekers

living in the North-Centre along with the mean characteristics of job seekers living in the

South and by subtracting to this value the average value observed in the South. This

difference can be interpreted as a measure of the differential in job seekers’ “choosiness”

9

Table 3 reports the estimated coefficients of the reservation wage equation. Data refer

to April 2000. The last row of Table 3 reports the two components of the Oaxaca’s

decomposition. The last two columns report the contribution to the differential of each

independent variable.

Independent variables are standard socio-demographic variables like sex, potential

experience (age – years of schooling) and the square of the potential experience, educational

attainment (primary, secondary, high school, university), the presence of past work

experiences, the number of household members, the role of individuals within the household

(single living alone, household head, spouse of the head, son/daughter, other), the self-

defined economic status (job seeker, student, housekeeper, retired, other), the ILO status

(being a job seeker in/out of the labour force) and search duration. The estimated equation

includes the log of the unemployment rate, separately calculated for men and women, age

group (15-24, 25+) and region of living. To account for heterogeneity in preferences, the

model includes a dummy equal to 1 if the person is available to any type of dependent

employment and any type of working time, and a set of dummies representing the implicit

cost of moving (in the same place of living, in the neighbourhood, everywhere in Italy,

abroad).

Since the estimated equation is a simple reduced form model, the corresponding

coefficients are not easily interpretable. However, the last two columns of Table 3 suggest

that among the observables, all the differences are almost compensated and that the

population composition does not contribute to explain the differential. Among the

unobservable factors, the most part of the differential is due to the constant term, while the

socio-demographic, household background and preference-related variables have a

remarkably lower impact.6

Table 4 reports the Oaxaca’s decomposition, derived by applying the same model

reported in Table 3 to all surveys conducted in April from 1993 to 2002. First, Table 4

6 A relatively sizeable contribution to the differential is due to the higher “choosiness” of men living in theSouth and to persons with a secondary attainment. The effect of the linear term of potential experience isfully compensated by the effect of the squared term. This evidence is clearly in contrast with the finding ofGiraldo and Trivellato (2003), who find the reservation wage differential is caused by differences in

10

confirms the existence of a negative reservation wage differential for the entire period 1993-

2002. Second, this difference on average is due to both the observable and the unobservable

components. Third, from 1997 onwards, the weight of the unobservable is higher than the

weight of observables, apparently confirming the hypothesis that differences in reservation

wages may be due to a higher, non-measurable, “choosiness” of job seekers residing in the

Southern part of Italy.

3 Modelling reservation wages

In the last section we showed that the pattern of observed reservation wage, as

summarised by standard empirical analysis, may support the hypothesis that the high

unemployment rate plaguing the Southern part of Italy may be due to a higher job seekers’

“choosiness”. Standard empirical analysis however does not take into account some relevant

characteristics of the observed reservation wage distribution.

As already stressed, reservation wages are generally collected only for people who

state to seek for work. However job search theory demonstrates that the choice of searching

depends on the utility from non-participating to the labour market and ultimately on the

same set of variables influencing reservation wages (see, for example, Pissarides, 2000,

ch. 7). This implies a standard problem of selection in the observed reservation wage

distribution.

Job search theory allows us also to specify the type of truncation affecting the

reservation wage distribution. People search for job only if the expected value of searching

is higher than the value of staying out of the labour force. Since reservation wages are an

increasing function of the value of non-market activities, on average more wealthy people

are expected to have higher reservation wages. However, the higher is the reservation wages,

the lower is the probability to get a suitable job. Thus, when job seekers have to pay a cost

for searching, more wealthy people are less probable to search for work. As a consequence,

the observed reservation wage distribution is truncated from above and both the mean and

the variance of the truncated (observed) variable are expected to be smaller than the mean of preferences about the desired job.

11

the original variable.

Truncation from above, however, is not the only source of bias in our data. Since we

analyse the reservation wages of those who are not employed, we have to consider that the

probability of being employed at a given time depends upon the same set of variables

influencing reservation wages. In particular, one might expect that on average reservation

wages of the employed population are lower than the reservation wages of the non-working

population. The observed reservation wage distribution is then truncated both from above

and from below.

The relevance of selection can be addressed also by the evidence presented in Table 2.

For instance, the last rows of the Table report the share of job seekers in total non-working

individuals. As already mentioned, the job search theory suggests that less wealthy

individuals have a higher probability to look for employment. Coherently with the findings

of the theory, the share of job seekers in non-working population, this share is remarkably

higher in the Southern part of the country than in the North-Centre for all the groups

considered, suggesting that truncation from above can be a relevant issue for the distribution

of reservation wages in the North-Centre.7 Similarly, the higher employment rate in this area

signals differences in the second truncation process. In turns, differences between the North-

Centre and the South may then be caused by differences in the selection process, i.e. by both

differences in labour demand and the value of non–market activities in the two areas.

Empirically, to take into account the existence of differences in the selection process,

we define the variable iI1 which is equal to 1 if the i-th individual is not employed and it is

zero otherwise. Similarly, we define iI 2 = 1 if the i-th individual looks for a dependent and

full time job and 2I = 0 otherwise.8 In this way we also model the sample selection we

7 The higher reservation wages observed in the Southern part of the country can be caused not only bydifferences in the mean value of the determinants of reservation wages (e.g. the value of non-market activities),but also on their variability.

8 Note that alternatively we could define the following variables: iE1 =1 if the i-th person participates to thelabour market (as employed or non-employed job seeker), iE2 =1 if the i-th person is a job seeker, and the

expected reservation wage as )1,1,|( 21* == EEXwE . This model specification gives similar results, that

can be obtained upon request.

12

introduced in our data by looking only at reservation wages of people looking for a full-time

dependent employment.

The expected value of the truncated reservation wage is equal to:

(2) )1,1Pr(')1,1,|( 212* ===== IIXIIXwE β

Under the assumption that ),,()1,1Pr( 221121 ργγ TT ZZFII === is a standard

cumulative bivariate normal distribution with correlation coefficient ρ , Maddala (1983)

shows that:

(3) 221121* )1,1,|( MMXIIXwE T ϑϑβ ++===

where 1M and 2M are two Inverse Mill’s ratios “adjusted” to take into account that the

underlying distribution is a bivariate normal (see Maddala, 1983 pp. 282-284 and Duca and

Rosenthal, 1993, for the formulae). The vector of parameters β can then be estimated by a

standard two-stage procedure: at the first stage we estimate the probability to observe a job

seeker in the set of non-working individuals, and at the second stage we estimate the (3).9

It is straightforward to show that, similarly to the standard Oaxaca’s formula, Equation

(3), separately estimated for the two areas, can be used to decompose observed reservation

wages in three terms, one related to observed characteristics, one to differences in job search

behaviour and one imputed to selection (see also Yun, 2000, for a discussion) . More

formally,

(4) )()'ˆˆ(ˆ)'(**SNSSNNSNSN XXXww Λ−Λ+−+−=− βββ

where ∑+

=ΛA

AjAjiAjAjA n

MM 221ˆˆ ϑϑ

and An equal to the number of observation in area A,

A=N,S ( AΛ is simply the sample mean of the estimated selection terms).

The decomposition allows us to estimate the relative importance of the three factors in

determining the reservation wage differential.

9 Since we observe 11 =I only if 12 =I , this is a model with partial observability (see Maddala, 1983, p.

13

Empirically, we have first estimated ),,()1,1Pr( 221121 ργγ TT ZZFII === , where 1Z

and 2Z are vectors of variables used to model the probability to be not employed and the

probability to search for work. As before, the independent variables are standard socio-

demographic variables, potential experience, educational attainment, the number of

household members and the role of individuals household relationships (single living alone,

household head, spouse of the head, son/daughter, other), the log of the unemployment rate.

It is well known that all models with selection are subject to strong assumptions to ensure

the identifiability of parameters. In our case the estimated coefficients 1γ , 2γ and β are

identified only if XZZ ≠≠ 21 . In practise however, it is difficult to find variables which

affect the probability of being not employed, with no effect on the probability to search for

work and reservation wages. We assume that the probability of being not employed depends

on a set of regional dummies reflecting local market characteristics, like the sectoral

composition of employment. Instead, we assume that the probability of being a job seeker

depends of individuals’ past working history and in particular on the causes of job separation

(never employed, fired, separated for personal reasons, retired, etc.). Neither of these two

sets of variables enter the reservation wage equation that includes dummies for preferences

about the desired job.

Estimated coefficients refer to a reduced form and they are not reported in this paper

(results are available upon request).10 The reservation wage equation is reported in Table 5.

The model is analogous to the OLS estimates presented in Table 3, except for the inclusion

of the two “adjusted” Mill’s ratios 1M and 2M .

The two Oaxaca’s components are now both positive. Coherently with the prediction

of the theory a central role in determining the term NSN XX β̂)'( − is played by the local

unemployment rate. Among the unobservables, the differences in the constant term are now

fully compensated by the expected returns to potential experience, higher in the North-

Centre than in the South and by a higher sensitivity to the level of the unemployment rate in

282). We have estimated this probability by applying the Stata command: heckprob.

10 It is worth mentioning that, as suggested by the theory the estimated correlation parameter ρ is alwaysnegative and comprised between -.25 and -.35 in the North-Centre and between –35 and –50 in the South.

14

the North-Centre. Thus, when considering selection, the expected reservation wages have no

paradoxical pattern.

Table 6 reports the Oaxaca’s decomposition in the presence of selectivity for the years

1993-2002. For all the years considered the negative reservation wage differential is

explained by selection. Coherently with overall labour market conditions, people living in

the North-Centre is more “choosy” than people living in the South and the term SSN X)'ˆˆ( ββ −

is always positive. Finally, the differences in average characteristics had a negative impact in

years 1993-94, their effect was nil between 1995 and 1997 and then positive, as a

consequence of the sharp reduction in the level of the unemployment rate registered in the

North-Centre after 1998. Thus, differently from Giraldo and Trivellato (2003), variables

related to tastes, like the availability to accept any type of job, have a consistent but lower

impact of the reservation wage differential. Overall, the South-North differential in the

reservation wage is mostly explained by the differences in the sample selection leading only

some groups in the population to be a non-employed job-seeker.

Figure 1 reports the average predicted reservation wages from 1993 to 2002, estimated

for the whole working age population by the use of the model presented in Section 4.

For each area predicted values are calculated as AAA Xw β̂'ˆ * = , i.e. by letting the two

“adjusted” inverse Mill’s ratios equal to zero and the vector of independent variables at the

mean value. Averages are evaluated on the sample of the entire population, i.e. working and

non-working individuals. Predicted values can then be considered as an overall labour

supply index. They are the log of reservation wages, expressed in euro, at constant prices

(the base year is 2000).

If we take into account that on average the standard deviation of the prediction is

around .08 in the North-Centre and to .04 in the South, the figure suggests that in the

Northern and Central regions reservation wages remained roughly constant from 1993 to

2002, while they increased in the Southern part of the country, contemporaneously to the

growth of employment recorded in the South after 1999. Consequently, also the North-South

reservation wage differential, higher than 15 per cent at the beginning of the period,

decreased to around 4 per cent in 2001-2002. Additionally, this evidence leads to exclude

15

that the employment growth registered in Italy after the mid of the Nineties was due to

reduction of job seekers’ “choosiness”, i.e. as measured by self-reported reservation wages.

4 Reservation wages and individual behaviour.

As already stressed, the estimated equations presented in Tables 3 and 5 are simple

reduced form which do not allow to determine how individuals set their reservation wages.

In the last section we have showed that taking into account selection, the pattern of

reservation wages in Italy is consistent with the prediction of the theory and it is lower in

those regions characterised by a higher unemployment rate. In this section we would like to

verify whether (and how) reservation wages respond to the complex set of variables

determining the expected value of employment and selection itself. Further we would like to

verify whether, given the results presented in Section 3, reservation wages can be used to

predict future labour market movements.

4.1 The determinants of reservation wages

For job search theory, reservation wages are a complex function of the present utility

of searching for job and the expected value of employment. The first element is related to the

individual condition during search and non-labour income. The value of employment is

affected by the current wage offer distribution and the matching technology between firms

and workers. In particular, any increase in the expected value of employment typically

induces workers to increase their reservation wage. A rise in the expected value from

employment may be due to 3 factors: (1) a shift to the right of the wage offer distribution,

i.e. an increase in the average of the potential wages distribution;11 (2) a mean preserving

increasing spread of the wage distribution12 (i.e. an increase in the average of that portion of

the wage distribution lying to the right of the given reservation wage with no effect on the

overall wage offer mean); (3) a rise in the probability of receiving a job offer for a given

11 In this paper we define as potential wage the wage level an individual can aspire, given her socio-demographic characteristics (educational attainment, gender, age, etc.) and the features of the labour market.

12 For a formal definition of a mean preserving increase in the spread of the wage distribution see Rothschild-

16

shape of the wage offer distribution.

These predictions hold in the simplest setting in which jobs last forever and there is no

search on the job (i.e. after a given offer has been accepted).13 Instead, in models where

people may be separated from their job, a rise in the risk of involuntary separation and a

shortening in the expected duration of the corresponding job relationship reduce the present

discounted value of employment, with a negative effect on reservation wages (see Wolpin,

1995, and Mortensen, 1986, for a review). The risk of involuntary separation may arise also

when people can be offered temporary job contracts, such a feature being a relevant aspect

of the jobs currently available to job-seekers. In this case no clear prediction emerges.

Assume that employed workers may search for a better job during employment, but assume

that on-the-job-search is more expensive. Assume also that fixed-term workers search for a

permanent position. Thus, it cannot be excluded that the acceptance of a fixed term contract

may require a higher wage.14 On the contrary, reservation wages tend to decrease when

employed workers have higher chances than the unemployed of finding new and better jobs

in the future, getting access to a more dynamic career.15

Among all possible factors, together with the traditional measures of market tightness

(typically, the unemployment rate), in this paper we want to analyse the effects of the shape

of the wage offer distribution and of some job characteristics, like the temporary or

permanent nature of employment and the possibility to access to dynamic careers. All these

market factors, which in the next will be referred as the “market determinants” of the

Stiglitz (1970 and 1971).

13 Another assumption is that workers may not “recall” an offer previously refuted. The possibility of recall ofthe offers previously discarded induces the worker to be choosier, as he or she has a chance of accepting anoffer previously rejected and the refusal itself being less costly. However, the no-recall assumption does notdrastically change the results of the simplest setting so far described.

14 In the simple setting in which there is on-the-job search is equally costly and productive (in terms of theprobability of locating job offers) the reservation wage is driven down to the minimum acceptable wagewhich would result in a standard static labour supply model (see Mortensen, 1986). The possibility thataccepting the offer currently under scrutiny may increase the chances of better jobs in the future may furtherlower the reservation wage below that minimum, the acceptance wage becoming a sort of entry wage.

15 In the simple setting in which the on-the-job search is equally costly and productive (in terms of theprobability of locating job offers) the reservation wage is driven down to the minimum acceptable wagewhich would result in a standard static labour supply model (see Mortensen, 1986). The possibility thataccepting the offer currently under scrutiny may increase the chances of better jobs in the future may furtherlower the reservation wage below that minimum, the acceptance wage becoming a sort of entry wage.

17

reservation wage, are very heterogeneous across Italian regions and may potentially explain

at least part of the reservation wage paradox.16

These variables however are not directly observed. In the previous sections we have

modelled reservation wages just as a function of those individual variables like gender, age,

educational attainment, etc. and local labour market characteristics, that are correlated to the

relevant market determinants. Here we adopt a different strategy.

We estimate the expected wages, the corresponding variance, the probability to get a

temporary job and the probability of a job-to-job move by the use of external datasets. These

variables are modelled as functions of variables collected by the LFS, which are then used to

impute the estimated market determinants to our sample. In this way it is possible to verify

whether reservation wages are set coherently with measures of the expected value of the

market variables that the theory indicates as influencing reservation wages. We estimate the

shape of the distribution of the potential wages a given individual may aspire, as summarised

by the mean and the variance. Another assumption is that workers may not “recall” an offer

previously refuted. The possibility of recall of the offers previously discarded induces the

worker to be choosier, as he or she has a chance of accepting an offer previously rejected

and the refusal itself being less costly. However, the no-recall assumption does not

drastically change the results of the simplest setting so far described. The potential wage is

estimated on the basis of a simple net monthly salary equation of full time dependent

employees. Data are drawn from the Survey on Household Income and Wealth conducted by

the Bank of Italy (see Brandolini and Cannari, 1994 and Cannari and Sestito, 1995 for

details on the survey). Estimates are corrected to take into account selection and the effects

of job tenure on observed wage paid. The variability of the wage offer distribution is

estimated on the square of the estimated residuals of the wage equation.17

We capture job characteristics by trying to measure the probability that the available

job offers may lead to further job-to-job transitions, and by measuring the probability that

the received job offer concerns a temporary position. The first variable has been estimated

16 For example in 2000 the share of temporary workers on total employment was equal to 8.6 in the Northcentre and 10.5 in the South.

18

on a LFS sample of newly employed persons, identified as those among employed with job

tenure lower than two years in April 2000. Thus, we estimate the probability that a person is

a fixed term worker by a simple probit model. The probability to change a job when already

employed is estimated on observed quarterly job-to-job-flows collected by the LFS between

July 1999 and January 2000. Also this probability is estimated by a probit model.18 Apart

from the practical difficulties we met in these measurement exercises, it has to be stressed

that neither of the two measures may be considered fully satisfactory. As for the job-to-job

transitions, we are considering a subsequent event which is in itself endogenous and might

be due to the fact that the first job was unsatisfactory, not only to the job-search advantages

(and employability enhancing effects) of the first job. Moreover, focusing upon job-to-job

transitions says nothing about the earnings career prospects, these possibly being present

also on-the-job (through a tenure effect). As for the temporary nature of the jobs available, it

has to be recognised that such a factor might also capture other unsatisfactory features of the

temporary contracts per se. The average value of imputed variables is reported in Table 7, by

sex, gender and geographical area.

The estimated model is totally similar to the model presented in Section 3, except for

having estimated a single equation for people living in the North-Centre and in the South

and having instead included a dummy for the area of living.

Table 8 reports some of the estimated coefficients of the standard reduced form and of

the model with market determinants (since they are analogous to those presented in Table 5,

the other coefficients are not reported). As noted above, the estimated coefficients of the

reduced form model are not easily interpretable. The estimated coefficients of the market

determinants have instead a plausible sign. As suggested by the theory, individuals’ potential

wages, the variance and the probability to get a fixed-term job are highly significant and

with a positive and sizeable sign. As expected, the probability to find a new job (when

employed) negatively influences reservation wages. Finally the dummy relative to the area

of living, North-Centre, negative in the reduced form model, is now positive, suggesting

that, when controlling for some –even very imprecise- measure related to the market

17 Estimated results and further details are available on request.

19

determinants, the estimated coefficients of the reservation wages appear to be interpretable.

The model presented in Table 8 allows us also to estimate the ratio between

reservation wages and potential wages. This index is traditionally interpreted as a synthetic

measure of job seekers’ willingness to work (see, for instance, Jones, 1988). Figure 2 reports

the ratio between observed reservation wages and potential wages, averaged by sex and age.

The figure also plots the ratio between predicted reservation wages and potential wages.

Predicted values, derived by the model presented in Table 8, are calculated by keeping all

the independent variables, expect sex and potential wages, at their average value. Averages

are calculated on whole the sample of both working and non-working individuals. The two

indices confirm that Italian job seekers are characterised by a ration lower than 1. The

estimated ratio is however lower than the observed one, because of the effects of selection.

The age profile of the two indices is U-shaped. In general, women are more “choosy” than

men and their reservation/potential wage ratio is higher than the value estimated for men.

4.2 Transition probabilities

Job search theory predict a clear and negative relationship between reservation wages

and the probability to get a job. This simple prediction has been tested for other countries

(see for example Jones, 1988), but in Italy it is often found that reservation wages are not

correlated to future movements into employment. Also for this reason the usefulness of

reservation wages in explaining job search behaviour has been widely criticised. Giraldo and

Trivellato (2003), for example, estimate the relationship between transition probabilities and

their measure of job seekers’ “choosiness”. Because they do not find a strong relationship

between their index and individual labour market movements, they argue that individual

transitions are affected by unobserved heterogeneity.

In this section we show that also when selection in observed reservation wages is

explicitly taken into account, the relationship between expected reservation wages and

transitions into employment is restored. We have estimated a simple probit model for one

year-apart transitions from non-employment into employment, conditional on a set of

18 Estimated results and further details are available on request.

20

characteristics affecting the probability to receive a job offer (sex, potential experience,

educational attainment, local non-employment rate, etc.) and four alternative measures of the

(log) reservation wage: (1) the observed reservation wage; (2) a standard OLS predicted

reservation wage, estimated by the model analogous to that presented in Section 2 (i.e. not

controlling for selection); (3) the predicted reservation wage estimated by the model

presented in Section 3 and referred to the subset of job seekers; (4) the predicted reservation

wage estimated by the model presented in Section 3 and referred to all the non-working

population. This last exercise has been carried out since in Italy, as in many other countries,

the most part of the annual transitions into employment are due to movements from out of

the labour force. Thus, we would like to verify whether our measure of reservation wages,

that can be calculated for the entire population, can be useful to predict also these transitions.

The estimated marginal effects, based on standard public longitudinal to year 2000

LFS file, are reported in Table 9. The elasticities to the observed reservation wages and the

OLS estimate of the reservation wages are positive and not significant. When controlling for

selection, the elasticity to the (estimated) reservation wage is negative and around –0.35 for

the set of job seekers and to – .12 for the sample non-working population.

5 Conclusions

In this paper we have shown that data on self-reported reservation wages are affected

by a multiple selection process, which limits the interpretability of the observed reservation

wage distribution. Of course, our results may be considered specific to the Italian labour

market. However, consider Table 9, which reports the average reservation wages collected

by the ECHP in 2000 for the following countries: France, Germany, Italy and Spain. These

countries have been selected on the basis of the sample size and of heterogeneity in labour

market conditions across different areas.

Similarly to the Italian LFS, in the ECHP working age individuals are asked whether

they search for work. If they answer affirmatively, they must indicate the desired number of

working hours and the corresponding minimum net monthly salary they would accept for

that job. We have selected all non-working people who would like to work at least 30 and no

more than 46 hours per week (that is, people looking for a full time employment). Data are

21

then comparable to those presented in Table 2.

For each country two main geographical areas are identified: “High unemployment

areas” and “Low unemployment areas”. Low unemployment areas have been identified as

those areas where the unemployment rate was higher than the national average in 2000 (see

Eurostat, 2001). They are defined in accordance with NUTS aggregates and corresponds to:

(1) Bassin Parisien, Nord–Pas–de–Calais and Mediterranée for France; (2) Berlin,

Brandeburg, Bremen, Mecklenburg-Vorpommern, Sachsen, Sachsen-Anhalt, Thurigen in

Germany; (3) Abruzzi-Molise Campania, Sud, Sicilia, Sardegna, that fully corresponds to

the Southern part of Italy, as defined in the previous sections; (4) Noroeste, Centro, Sur in

Spain. For these countries we have estimated a simple model equivalent to the model we

have used for the Italian LFS.19

For each country the sample size amounts of average to 280 job seekers. Similarly to

Table 2, Table 10 reports also the employment rate and the share of job seekers in total non-

working population. Reservation wages expressed in euros.

In all the countries, the share of job seekers in total non-working population is higher

in those areas where the employment rate is lower, suggesting that the selection problem

affecting Italian data could be more general. Consider for example Germany. In 2000 the

share of job seekers was around 16 per cent in the high unemployment areas and to 5 per

cent in the rest of the country. In spite of this striking difference, the average observed

reservation wages were quite similar in the two areas. Similar, even if less evident,

conclusions hold also for Spain and France. Data and estimates for Italy20 confirm the

existence of a reservation wage paradox.

In general, the results of our empirical exercise confirm that that individual labour

supply behaviour cannot be fully understood without analysing the determinants of the non-

working condition and the related decision to look for employment. For this reason, looking

19 Because the ECHP contains information on household wealth and individual non-employment income, alsothese data have been added to the model the model presented in Section 4. Estimates are available uponrequest.

20 Differences from the LFS estimates may be due to various factors, such as the lower ECHP sample size,implying less accurate estimates and the reference population, equal to people aged at least 15 in the LFS and

22

just at job seekers’ reservation wages can distort our understanding of the labour market.

This bias can be even more relevant when comparing the willingness to work of different

groups in the population and when determining whether unemployment is “voluntary” or

not.

at least 16 in the ECHP.

23

Tables and figuresTab. 1

Reservation wages by type of desired job characteristics(Year 2000, euro averages and standard deviations within brackets)

Full time Part time Any workingtime

Total

Dependent permanent 6 7 6 6(1) (2) (1) (2)

Dependent fixed term 5 6 4 5(1) (2) (1) (2)

Any type of dependent employment 5 7 5 5(1) (3) (2) (2)

Self-employment 7 9 7 7(3) (4) (3) (3)

Total 6 7 5 6(1) (3) (2) (2)

Source: authors’ elaboration on LFS data

24

Tab. 2

Reservation wages by job seekers’ characteristics(Year 2000, euro averages and standard deviations within brackets)

Men Women

Age 15—24 Age 25+ Age 15--24 Age 25+

Average reservation wages

North-Centre 821.2 918.5 794.2 837.9

South 866.4 918.0 831.6 846.2

Difference: North-Centre-South -45.2 0.5 -37.4 -8.3

Standard error 1.3 1.9 1.5 2.0Employment rate

North-Centre 0.365 0.762 0.302 0.498

South 0.188 0.713 0.100 0.300Unemployment rate

North-Centre 0.060 0.025 0.081 0.038

South 0.172 0.091 0.161 0.094Share of job seekers

North-Centre 0.127 0.139 0.160 0.119

South 0.305 0.455 0.283 0.231Source: authors’ elaboration on LFS data.

25

Tab. 3

Reservation wages: OLS estimated coefficients(logs)

North-Centre South

Coefficient StandardError

Coefficient StandardError NSN XX β̂)'( − SSN X)'ˆˆ( ββ −

Constant 6.739 0.032 6.841 0.024 -0.102

Man 0.007 0.015 0.025 0.011 0.000 -0.009

Primary educational attainment -0.220 0.021 -0.214 0.016 0.008 -0.001

Secondary educational attainment -0.207 0.016 -0.178 0.013 0.007 -0.012

High school attainment -0.129 0.015 -0.127 0.012 -0.006 -0.001

Potential experience*Man 0.013 0.002 0.006 0.001 -0.015 0.061

Potential experience ^2*Man 0.000 0.000 0.000 0.000 0.006 -0.027

Potential experience*Woman 0.006 0.002 0.002 0.001 0.005 0.026

Potential experience ^2*Woman 0.000 0.000 0.000 0.000 -0.002 -0.006

Log(Unemployment rate) -0.005 0.006 0.005 0.007 0.005 0.010

No. Household members 0.005 0.004 -0.001 0.003 -0.002 0.023

Housewife -0.003 0.016 0.006 0.010 0.000 -0.001

Student 0.097 0.019 0.046 0.013 0.000 0.003

Retired -0.022 0.052 0.071 0.069 0.000 0.000

Other condition 0.044 0.039 0.024 0.032 0.000 0.000

Single living alone 0.013 0.028 -0.010 0.023 0.000 0.001

Household head 0.042 0.023 0.066 0.017 -0.002 -0.004

Spouse of the head -0.012 0.022 -0.012 0.018 0.000 0.000

Son/Daughter 0.002 0.020 -0.013 0.016 0.000 0.009

Desired job location: commutingdistance

0.029 0.008 0.015 0.006 0.003 0.005

Desired job location: everywhere in Italy 0.076 0.014 0.091 0.007 -0.010 -0.003

Desired job location: everywhere 0.129 0.016 0.105 0.012 0.002 0.001

Past work experiences -0.005 0.009 0.002 0.007 -0.001 -0.003

Less selective job seeker -0.123 0.013 -0.120 0.007 0.008 -0.001

Search duration: 1-3 months 0.017 0.013 -0.024 0.011 0.001 0.003

Search duration: 4-12 months 0.013 0.010 -0.002 0.008 0.001 0.002

Search duration: 12 months+ -0.004 0.010 -0.016 0.007 0.000 0.002

Job seeker out of the labour force 0.028 0.008 0.026 0.005 -0.002 0.001

No. observations 2625 6730

R2 0.206 0.155

Oaxaca’s components 0.006 -0.026Source: authors’ elaboration on LFS data.

26

Tab. 4

Reservation wages: Oaxaca’s decomposition(logs)

North-Centre South DifferenceNSN XX β̂)'( − SSN X)'ˆˆ( ββ −

1993 6.524 6.571 -0.047 -0.026 -0.021

1994 6.566 6.620 -0.053 -0.026 -0.027

1995 6.582 6.652 -0.070 -0.024 -0.046

1996 6.635 6.657 -0.021 -0.020 -0.001

1997 6.654 6.682 -0.028 -0.010 -0.018

1998 6.662 6.690 -0.027 -0.012 -0.015

1999 6.685 6.693 -0.008 0.006 -0.014

2000 6.740 6.762 -0.020 0.006 -0.026

2001 6.772 6.806 -0.034 -0.015 -0.019

2002 6.809 6.830 -0.021 0.020 -0.041Source: authors’ elaboration on LFS data.

27

Tab. 5Reservation wages: estimated coefficients with multiple selection criteria

(logs)

North-Centre South

Coefficient Stand. Err. Coefficient Stand. Err.NSN XX β̂)'( − SSN X)'ˆˆ( ββ −

Constant 6.819 0.036 6.927 0.029 0.000 -0.108

Man -0.023 0.016 0.032 0.011 0.001 -0.028

Primary educational attainment -0.256 0.022 -0.291 0.028 0.010 0.005

Secondary educational attainment -0.284 0.022 -0.243 0.024 0.010 -0.017

High school attainment -0.152 0.015 -0.170 0.016 -0.007 0.006

Potential experience*Man 0.040 0.006 0.014 0.003 -0.045 0.205

Potential experience ^2*Man -0.001 0.000 0.000 0.000 0.023 -0.119

Potential experience*Woman 0.021 0.003 0.007 0.002 0.015 0.089

Potential experience ^2*Woman 0.000 0.000 0.000 0.000 -0.010 -0.048

Log(Unemployment rate) -0.041 0.010 -0.006 0.007 0.042 0.038

No. Household members -0.001 0.004 -0.001 0.003 0.000 0.001

Housewife -0.064 0.042 0.100 0.037 0.001 -0.017

Student 0.028 0.062 0.188 0.055 0.000 -0.008

Retired -0.131 0.085 0.271 0.097 -0.001 -0.001

Other condition -0.024 0.061 0.155 0.057 0.000 -0.001

Single living alone 0.022 0.028 0.010 0.024 0.000 0.000

Household head 0.075 0.024 0.102 0.021 -0.004 -0.005

Spouse of the head -0.060 0.025 -0.007 0.019 0.001 -0.009

Son/Daughter -0.003 0.020 -0.024 0.016 0.000 0.013

Desired job location: commutingdistance

0.029 0.008 0.015 0.006 0.003 0.005

Desired job location: everywhere in Italy 0.077 0.014 0.091 0.007 -0.010 -0.003

Desired job location: everywhere 0.125 0.016 0.105 0.012 0.002 0.001

Past work experiences 0.003 0.009 -0.002 0.007 0.001 0.002

Less selective job seeker -0.123 0.013 -0.121 0.007 0.008 0.000

Search duration: 1-3 months 0.020 0.013 -0.024 0.011 0.001 0.003

Search duration: 4-12 months 0.015 0.010 -0.002 0.008 0.001 0.002

Search duration: 12 months+ -0.002 0.010 -0.015 0.007 0.000 0.002

Job seeker out of the labour force 0.027 0.008 0.027 0.005 -0.002 0.000

M1 -0.209 0.042 -0.123 0.032

M2 0.023 0.032 -0.075 0.032

No. observations 2625 6730

R2 0.214 0.158

Oaxaca’s components 0.040 0.009Source: authors’ elaboration on LFS data.

28

Tab. 6

Reservation wages: decomposition with selectivity terms(logs)

North-Centre South DifferenceNSN XX β̂)'( − SSN X)'ˆˆ( ββ − Selection

1993 6.524 6.571 -0.047 -0.026 0.048 -0.069

1994 6.566 6.620 -0.053 -0.012 0.048 -0.089

1995 6.582 6.652 -0.070 -0.002 0.041 -0.109

1996 6.635 6.657 -0.021 -0.002 0.126 -0.145

1997 6.654 6.682 -0.028 0.005 0.002 -0.035

1998 6.662 6.690 -0.027 0.010 0.071 -0.108

1999 6.685 6.693 -0.008 0.033 0.030 -0.071

2000 6.740 6.762 -0.020 0.040 0.009 -0.069

2001 6.772 6.806 -0.034 0.028 0.023 -0.085

2002 6.809 6.830 -0.021 0.060 0.018 -0.099Source: authors’ elaboration on LFS data.

Fig. 1

Predicted Reservation wages: 1993-2002(logs)

6.70

6.75

6.80

6.85

6.90

6.95

7.00

7.05

1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

South

North-Centre

Source: authors’ elaboration on LFS data.

29

Tab. 7

The determinants of reservation wages(averages)

Men Women

Age 15—24 Age 25+ Age 15--24 Age 25+

Average potential wages

North-Centre 7.355 7.834 7.297 7.557

South 7.173 7.666 7.104 7.375Potential wage variance

North-Centre 0.187 0.187 0.167 0.166

South 0.261 0.251 0.244 0.233Probability to find a temporary employment

North-Centre 0.379 0.209 0.414 0.345

South 0.534 0.331 0.571 0.490Probability of a job to job move

North-Centre 0.126 0.048 0.121 0.045

South 0.131 0.052 0.127 0.050Source: authors’ elaboration on LFS data.

30

Tab. 8

Reservation wages: the determinants(logs)

Reduced form Model with marketdeterminants

Coefficients Standarderrors

Coefficients Standarderrors

Constant 6.818 0.023 3.083 1.651

Potential wage 0.392 0.189

Variance of the potential wage 2.114 0.532

Probability to find a temporary employment 0.669 0.338

Probability of a job to job movement -2.049 0.417

Man 0.009 0.009 -0.022 0.013

Primary educational attainment -0.274 0.017 0.122 0.112

Secondary educational attainment -0.251 0.017 0.102 0.086

High school attainment -0.153 0.011 0.084 0.054

Potential experience*Man 0.022 0.003 0.014 0.008

Potential exp ^2*Man 0.000 0.000 0.000 0.000

Potential experience*Woman 0.011 0.002 0.006 0.005

Potential exp ^2*Woman 0.000 0.000 0.000 0.000

Log(Unemployment rate) -0.023 0.006 -0.025 0.009

Living in the Norh-Centre -0.011 0.007 0.160 0.049

M1 -0.113 0.024 -0.099 0.026

M2 0.085 0.021 0.081 0.0219,355 9,3550.168 0.170

Source: authors’ elaboration on LFS data.

31

Fig. 2

Reservation wage/Potential wage ratio(percentages)

.85

.9.9

5R

atio

15 25 35 45 55 65Age

Estimated Observed

Men

.85

.9.9

5R

atio

15 25 35 45 55 65Age

Estimated Observed

Women

Reserv ation w./Potential w.

Source: authors’ elaboration on LFS data.

32

Tab. 9

Reservation wages and transition probabilities(marginal effects)

Log observed rw OLS prediction Prediction with selection(unemployed)

Prediction with selection(all)

Coefficients StandardError

Coefficients StandardError

Coefficients StandardError

Coefficients StandardError

Man -0.131 0.027 -0.104 0.022 -0.173 0.026 0.005 0.005

Primary educational att. -0.065 0.028 -0.072 0.027 -0.273 0.024 -0.050 0.007Secondary educational att. -0.081 0.026 -0.097 0.027 -0.317 0.027 -0.066 0.008High school attainment -0.066 0.024 -0.080 0.023 -0.212 0.025 -0.041 0.006Potential exp.*Man 0.013 0.002 0.013 0.002 0.038 0.002 0.008 0.001Potential exp ^2*Man 0.000 0.000 0.000 0.000 -0.001 0.000 0.000 0.000Potential exp.*Woman -0.002 0.003 -0.001 0.002 0.011 0.002 0.002 0.000Potential exp ^2*Woman 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000Log(Unemployment rate) -0.123 0.010 -0.110 0.008 -0.207 0.009 -0.021 0.002

Reservation wage w* 0.069 0.030 0.045 0.090 -0.355 0.106 -0.124 0.034

No. observations 3,911 3,911 3,911 26,430

Pseudo R2 0.066 0.066 0.126 0.100Source: authors’ elaboration on LFS data.

33

Tab. 10Reservation wages in European countries

(Euro, percentage values)Low unemployment areas High unemployment areas (1)

FranceAverage observed reservation wage 1,077 972Employment rate 63.9 59.4Share of job seekers 10.4 11.6

GermanyAverage observed reservation wage 1,081 1,055Employment rate 63.9 62.1Share of job seekers 4.9 16.1

ItalyAverage observed reservation wage 867 891Employment rate 57.8 42.4Share of job seekers 13.0 21.7

SpainAverage observed reservation wage 761 683Employment rate 56.8 49.5Share of job seekers 11.9 18.4Source: authors’ elaboration on ECHP data. Year 2000. (1) Bassin Parisien, Nord – Pas – de – Calais and Mediterranée for France; (2)Berlin, Brandeburg, Bremen, Mecklenburg-Vorpommern, Sachsen, Sachsen-Anhalt, Thurigen in Germany; (3) Abruzzi-MoliseCampania, Sud, Sicilia, Sardegna (the Southern part of the country) in Italy; (4) Noroeste, Centro, Sur in Spain.

34

References

Berg van der G. J. and Ridder G., (1998), “An Empirical Equilibrium Search Model of theLabor Market”, Econometrica, Vol. 66, No. 5, pp. 1183-1221.

Boeri, T. and P. Garibaldi, (2000), “Shadow activity and unemployment in a depressed labormarket” in IGIER and Fondazione Rodolfo De Benedetti, Le nuove frontiere dellapolitica economica 2000. Dimensioni del lavoro e del non lavoro nel Mezzogiornod’Italia, Proceedings of the Seminar held at Ministero del Tesoro, del Bilancio e dellaProgrammazione Economica, Rome, 19 September 2000.

Cannari, L. and P. Sestito, (1995), “Income, Consumption and Labour Force Surveys: AnOutlook on Italy”, Annali di statistica, Anno 124, Serie X, Vol. 5, pp. 25-45.

Casavola, P. and P. Sestito, (1995), Come si cerca e come si ottiene un lavoro? Un quadrosintetico sull’Italia e alcune implicazioni macroenomiche, in “Disoccupazione: analisimacroeconomica e mercato del lavoro”, a cura di A. Amendola, Ed. Scientificheitaliane.

Duca J. V. and S. S. Rosenthal, (1993), “The Effects of Borrowing Constraints on ConsumerLiabilities”, Journal of Money, Credit and Banking, Vol. 25, pp. 197-213.

Eurostat, (2001), Unemployment in the regions of the European Union, 2000.

Flinn, C. and J. J. Heckman, (1982), “New Methods for Analyzing Structural Model ofLabor Force Dynamics ”, Advances in Econmetrics, No. 1, pp. 35-95.

Giraldo A. and U. Trivellato, (2003), “Assessing the “choosiness” of job seerkers. Anexplanatory approach and evidence for Italy”, CESifo Working Paper n. 925, Munich,Germany, April 2003.

Holzer, H. J., (1986), “Reservation Wages and Their Labor Market Effects for Black andWhite Male Youth”, The Journal of Human Resources, Vol. 21, No. 2, pp. 157-177.

Jones, S. R. G., (1988), “The Relationship Between Unemployment Spells and ReservationWages as a Test for Search Theory”, The Quarterly Journal of Economics, Vol. 103,No. 4, pp. 741-765.

Jones, S. R. G., (1989), “Reservation Wages and the Cost of Unemployment ”, Economica,New Series, Vol. 56, No. 222, pp. 225-246.

Kiefer, N. M. and G. R. Neumann, (1979), “An Empirical Job Search Model with a test ofthe Constant Reservation Wage Hypothesis”, Journal of Political Economy, No. 87,pp. 89-107.

Istat, (2001), Documentazione tecnica del file standard della Rilevazione trimestrale delleforze di lavoro.

Lancaster, T. and A. Chesher T., (1983), “An Econometric Analysis of Reservation Wages”,Econometrica, No. 51, Vol. 6, pp. 1661-1776.

35

Maddala, G. S., (1983), Limited-Dependent and Qualitative Variables in Econometrics,Cambridge University Press, New york.

Mortensen, D. T., (1986), “Job Search and Labour Market Analysis”, Handbook of LabourEconomics, Vol. II, edited by O. Ashenfelter and R. Layard, North Holland,Amsterdam, pp 849-919.

Mortensen, D.T. and C. Pissarides, (1999), “New Developments in Models of Search in theLabour Market ”, Handbook of Labour Economics, Vol. III, edited by O. Ashenfelterand D. Card, Elsevier Science B.V ., pp. 2568-2627.

Oaxaca R., (1973) , “Male-Female Wage Differentials in Urban Labor Markets ”,International Economic Review, Vol. 14, pp. 693-706.

Paggiaro, A. and N. Torelli, (1999), “Una procedura per l’abbinamento di record nellaRilevazione trimestrale delle forze di lavoro” in Documentazione tecnica del filestandard panel della Rilevazione trimestrale delle forze di lavoro, Istat, Rome.

Pissarides, C. A., (2000), Equilibrium Unemployment Theory, 2nd ed. MIT Press,Cambridge, Mass..

Prasad E. E. (2003), “What determines the reservation wages of unemployed workers? Newevidence from German micro data”, IZA Discussion paper, No. 694.

Rothschild M. and J. Stiglitz, (1970), “Increasing Risk I: A definition”, Journal of EconomicTheory, Vol. 2, pp. 225-43.

Rothschild M., J. Stiglitz (1971), “Increasing Risk II: Its economic consequences”, Journalof Economic Theory, Vol. 3, pp. 66-84.

Sestito, P. (1988), “Flussi e mobilità nel mercato del lavoro italiano (1979-1985)”, Ricercheeconomiche, Vol. 42, n. 1.

Yun, M. S., (2000), “SElection Bias and the Decomposition of Wage Differentials”, w.p.University of Western Ontario.

Wolpin K. A. (1995), Empirical Methods for the Study of Labour Force Dynamics, HarwoodAcademic Publisher GmbH, Luxembourg.