the labour supply behaviour of sole mothers and married mothers in australia: evidence from the 1986...

THE LABOUR SUPPLY BEHAVIOUR OF SOLE MOTHERS AND MARRIED MOTHERS IN AUSTRALIA:

EVIDENCE FROM THE 1986 INCOME DISTRIBUTION SURVEY*

RUSSELL ROSS and PETER SAUNDERS

University of Sydney University of New South Wales

I. INTRODUCTION

The increased prevalence of sole parenthood has created a number of dilemmas for social policy. There is now a body of empirical research showing for a number of countries that sole parent families have high poverty rates - well above those for two parent families with children or for the population as a whole (Smeeding and Torrey, 1988). Since many sole parent families are reliant on income support provided through government social security schemes, this has raised questions about the adequacy of these payments. At the same time, others have argued that the level and availability of income support for sole parents has contributed to the growth in the number of sole parent families (Murray, 1984; McDonald and Spindler, 1988). In the Australian context, it has also been argued that the income-tested nature of social security payments for sole parents has created a poverty trap that reinforces income support dependency once that support is received. The design of income support policy for sole parents thus confronts head-on the conflict between issues of adequacy and incentives.

It is, however, also recognised that these issues are particularly complex in the context of income support provisions for sole parents. Work decisions revolve around a far greater range of considerations than just the level, availability and conditions attached to income support payments (Cass, 1986: Brown, 1989). Factors such as the presence of children, particularly younger children, access to jobs, market wage rates and the availability and affordability of child care are also significant in the overall calculus that ultimately influences the decision to work (Social Security Review (SSR), 1986; Raymond, 1987). Income support clearly plays a key role in influencing the relative financial rewards from work and thus the incentive to undertake paid work, but decisions are also framed and action pursued within an environment shaped by these other considerations.

* Previous versions of this paper were presented at seminars at La Trobe University and the Research School of Social Sciences at the Australian National University. We are grateful to participants at those seminars for their comments. We wish to acknowledge the assistance of Toni Payne, George Matheson and Marina Paxman in the preparation of the tables and statistical analysis contained in the paper. We are especially indebted to Bruce Chapman for his comments and suggestions on Sections 111 and IV.

116

1993 THE LABOUR SUPPLY BEHAVIOUR OF MOTHERS IN AUSTRALIA 117

However, despite the merits of the broader strategy for tackling what is a very difficult area of social policy, the extent to which the structure of income support provisions for sole parents is a factor - for some, the factor - causing high income support dependency remains unresolved. To argue that other factors are also important in the work decisions of sole parents does not, of itself, imply that the strength of any benefit-induced work disincentive effects are insignificant. Rather, it suggests that investigation of the presence and size of disincentive effects needs to be undertaken within a framework that also encompasses these other factors. Yet there has to date in Australia been no serious attempt to investigate the determinants of the work decisions df sole parents within such a framework. Numerous studies have highlighted statistical trends between aspects of sole parent labour force behaviour and other variables (parent's occupation or education status; number and ages of dependent children, and so on) but these comparisons can be no more than suggestive unless analysed within a multidimensional decision-making framework.

This paper represents a first step in this direction by analysing and comparing the labour supply behaviour of sole mothers and married mothers. Mothers are defined as women living with and caring for dependent children, and married mothers are those actually living with a spouse. The framework of analysis adopted is the conventional microeconomic theory of female labour supply. It is used to evaluate how well' the model explains the behaviour of married mothers as compared with sole mothers and to highlight behavioural differences between the two groups. The paper thus attempts to gain an insight into the question of whether or not there are additional factors relevant to the work behaviour of sole mothers that are not also relevant to the work decisions of married mothers. Or to put this differently, does being a sole mother make any substantial difference to work decisions over and above those factors associated with being a mother, per se. Since one of the differences between sole mothers and married mothers is that most of the former are eligible for specific income support provisions while most of the latter are not, posing the research question in this way will hopefully shed some light on to the impact of income support arrangements on work decisions.

The paper is organised as follows: Section I1 presents some comparative data on the labour force status of sole mothers and married mothers; Section I11 describes the labour supply model and the data and variables used to estimate the parameters of the model; Section IV presents and discusses the results; while Section V summarises the main conclusions of the paper.

11. PARENTS AND THE LABOUR MARKET

Two features have characterised changes in employment patterns in Australia since 1973. The first is the rapid growth in part-time employment, both in absolute terms and relative to full-time employment growth. The second is the growth in full-time female employment relative to full- time male employment, particularly for unmarried females. Between 1973 and 1989, total male employment grew at an annual average rate of 1.10 per cent, while total female employment grew almost three times faster, at 3.08 per cent a year. The annual growth rate of part-time employment (5.47 per cent) was almost five times the growth of full-time employment (1.14 per cent).' While the trend towards part-time jobs generally and towards female jobs relative to male jobs have both

' Saunders and Matheson (1990). Table 20.

118 AUSTRALIAN ECONOMIC PAPERS JUNE

been in train since 1973, the period since 1983 has seen a much stronger growth in full-time employment relative to earlier years, associated with much better employment performance generally. Thus as labour market performance has improved since 1983, the movement towards part-time jobs and female employment has continued, but at a somewhat slower pace than in earlier periods. These developments would nonetheless appear initially to be particularly beneficial to sole parents, most of whom are women, particularly those sole parents who prefer part-time employment.

In recent years, the proportion of employed sole mothers in full-time and part-time employment has been approximately in the ratio 60:40, while for employed married mothers the ratio has been the reverse, i.e. 40:60.2 There is a common view in Australia that the relatively low part-time employment status of sole parents generally (and sole mothers in particular) results from the financial disincentives associated with the poverty trap. Certainly, as Whiteford, Bradbury and Saunders (1989) have shown, the potential for the poverty trap facing supporting parent beneficiaries to adversely affect work incentives is considerable. Cass (1 986) however, argues that a range of other factors are also relevant to the work decisions of sole parents:

In their daily attempt to combine child care, household duties and part-time employment, single mothers may find that the increased expenditure of time, effort and income involved in labour force participation does not yield a commensurately increased level of disposable income. As a result, the two most economically feasible options of income support are clarified: full-time employment or full benefit. (Cass, 1986, p.8)

Table I presents evidence that, while not confirming this view, is broadly consistent with it. The table compares the labour force states of sole mothers and mamed mothers by the age of youngest child. For both groups, labour force attachment increases sharply when the youngest child reaches the age of five and enters compulsory schooling. For female sole parents, there is a further increase when the youngest child reaches fifteen, although the reverse happens at this stage for married women. For both groups, increased labour force participation for those with a youngest child over five is concentrated in full-time employment, although part-time employment also rises compared to those with younger children. For sole mothers with a youngest child over fifteen, there is a marked switch from part-time to full-time employment, with a much weaker switch apparent for married mothers.

However, perhaps the most revealing aspect of Table I is the fact that (in both 1984 and 19893) the overall percentages in full-time employment are very similar for married mothers and sole mothers. In contrast, the rate of part-time employment for sole mothers is between 60 per cent and 65 per cent of that for married mothers. Although these data suggest that there are life cycle factors that play an important role in the labour supply decisions of both married mothers and sole mothers, the difference in part-time employment rates remain important even when life cycle factors are standardised by comparing those with a youngest child in the same age range. While the evidence is not definitive as to the impact of the poverty trap on part-time employment among sole mothers, there is a presumption that this is a factor underlying the observed trends.

Ross and Saunders (1990). Table 7.

' Data are shown for these two years as they are equi-distant before and after the September 1986 date for the data upon which the analysis reported in Section V is based.

1993 THE LABOUR SUPPLY BEHAVIOUR OF MOTHERS IN AUSTRALIA

IV. LABOUR SUPPLY ANALYSIS

119

Theoretical framework The framework used here to examine the labour force differences just described is the standard

‘second generation’ static labour supply model for estimating disaggregate labour supply functions developed by Heckman, Killingsworth and MaCurdy (1981) and described in Killingsworth (1983). This model is used because of the absence of longitudinal panel data (which would facilitate fitting a dynamic life cycle model, e.g. Heckman and MaCurdy (1980)) and of data on consumption patterns (which would facilitate fitting a life cycle consistent model, e.g. Blundell and Walker (1983)).

The theoretical model is described in full in Ross and Saunders (1990). Briefly, it is a three equation system in which individual mothers are assumed to work only if the wages she is offered by employers exceeds her reservation wage. The hours she works are assumed to be related to the gap between the offered wage and the reservation wage.

The model is:

where W,, W , and H are the offered wage, the reservation wage, and (depth) of participation (typically measured by hours worked per week or per year), respectively, and where subscript i refers to the ith individual. X , and X,. are vectors of characteristics. Only those women who are offered wages in excess of their reservation wage will choose employment, and the hours they work are assumed to be related to the difference between W, and W,.

Attention is focused in the analysis presented here only on the decision to seek employment. We estimate the likelihood function that individual mothers will seek paid employment, i.e. that their offered wages exceed their (individual) reservation wages. This likelihood function is described fully in Heckman (1980) and Heckman and MaCurdy (1980). It is estimated here using probit analysis.

Two issues are worth highlighting before turning to the estimates themselves. First, the data set used, while generally containing very good information on all the independent variables, has rather poor quality wage and hours information. The only earnings data relates to total annual earnings and the only hours information concerns a full-time/part-time dichotomy on hours worked in the survey week. This precludes direct estimation of the parameters of (1) and (3). Hence, in this paper we concentrate only on the parameters of the decision to seek work, and not on the number of hours actually worked. The second issue relates to the treatment of the unemployed. They are seeking employment and therefore are labour force participants, but their hours of work are constrained to zero. This problem may lead to biased estimates if the incidence of unemployment varies systematically across the population of mothers. As this is essentially an empirical matter, estimates of two sets of probits are presented. The first set of estimates are of labour force participation, and the second set are of employment only. A comparison of the estimates obtained will indicate if unemployment is experienced randomly among mothers or whether there is in fact a systematic pattern to their incidence of unemployment. (These results are reported in Columns 1 and 2 of Table V, respectively.)


TABLE I

Labour Force Participation of Mothers by Age of Youngest Child, 1984 and 1989 (per cent)

Age of Employed Employed Unemployed In the Youngest child Full-time Part-time labour

(years) force

Married Women 0-4 5-9 10-14 15-20 Total

Female Sole Parents 0-4 5-9 10-14 15-20 Total

Married Women 0-4 5-9 10-14 15-24 Total

Female Sole Parents 0-4 5-9 10-14 15-24 Total

11.2 20.9 27.0 30.1 19.1

8.8 17.9 25.8 34.4 19.3

13.8 26.9 32.8 35.3 23.7

11.0 23.6 32.4 48.0 25.4

1984

19.1 31.3 28.7 24.0 24.8

10.4 16.6 16.9 17.6 14.8

1989

28.6 37.8 34.8 30.6 32.3

16.5 24.8 24.5 18.5 21.0

4.0 3.7 2.1

3.3 *

5.1 8.3 6.3

6.4 *

2.8 4.8 2.2 1.4 3.0

6.4 6.1 4.2 4.7 5.5

34.3 55.9 57.8 56.0 47.2

24.2 42.8 49.0 57.6 40.5

45.2 69.5 69.8 67.3 58.9

34.0 54.6 61.1 71.1 52.0

Note: An asterisk (*) indicates that the sample is too small to produce reliable estimates. Source: ABS, Labour Force Status and Other Characteristics of Families, Catalogue No. 6224.0, various issues.

The data base The data are taken from the unit record file from the 1986 Income Distribution Survey

undertaken by the Australian Bureau of Statistics (ABS). There are a total of 3202 mothers with dependent children on the file, 2841 of whom were married at the time of the survey and 361 of whom were sole mothers. Of the 361 sole mothers, 285 - almost 79 per cent - had previously been married and the remaining 76 (21 per cent) had never been married. In this analysis, all sole mcthers are grouped together because of the relatively small number who had never been married.


However, it should be noted that sole mothers are a very heterogeneous group. Compared to the previously married sole mothers, the never married mothers are much younger (their average age is 27 compared to 37 for previously married mothers, with 44 per cent of never married mothers aged less than 25 compared to only 6 per cent of previously married mothers under the same age). Never married mothers have twice as many young children but only a quarter the number of older children, reflecting their own relatively young age. The employment situation of never married mothers is also quite different to that of previously married mothers. For never married mothers, the incidence of full-time employment is 14 per cent, the incidence of part-time employment is also 14 per cent and the incidence of unemployment is 8 per cent; the corresponding figures for previously married sole mothers are 27 per cent, 15 per cent and 7 per cent, respectively.

Table I1 summarises the labour market position of married and sole mothers in the sample. Although sole mothers have a lower labour force participation rate and a higher unemployment rate than is the case for married mothers, those sole mothers who are employed are concentrated more in full-time employment. The key differences are that compared to married mothers, sole mothers have lower labour force participation but higher unemployment, and among those who are employed there is lower .part-time employment and higher full-time employment. The research question raised is whether these reflect differences in the way sole mothers respond to labour market signals compared to married mothers, or whether sole mothers respond in the same manner as do married mothers, in which case the differences reported in Table I1 simply reflect differential endowments of characteristics.

TABLE I1

The Labour Force Position of Married and Sole Mothers, 1986

Married Mothers

Sole Mothers

Labour Force Status

In Labour Force Employed Full-time Employed Part-time Unemployed

Not in Labour Force

Total

Labour Force Statistics

Labour Force Participation Rate Incidence of Full-time Employment Incidence of Part-time Employment Incidence of Unemployment Unemployment Rate

Number

1659 583 933 143

1182

162 85 50 27

199

2841 36 1

Percentage

58.1 20.5 32.8 5.0 8.6

44.9 23.6 13.9 7.5

16.7

Source: ABS, 1986 Income Distribution Survey: unit record f i l e .


TABLE I11

Mnemonics and Definitions of Variables Used in the Probit Analysis

Mnemonic Definition

AGE 15 (if aged 15), 17 (aged 16-19), 22 (20-24), 27 (25-29), 32 (30-34), 37 (35-39), 42 (40-44), 47 (45-49), 52 (50-54), 57 (55 -59), 62 (60-64)

AGE1524 dummy variable, = 1 if aged 15-24, = 0 elsewhere AGE2534 dummy variable, = 1 if aged 25-34, = 0 elsewhere AGE3544 dummy variable, = 1 if aged 35-44, = 0 elsewhere AGE4554 dummy variable, = 1 if aged 45-54, = 0 elsewhere AGE5564 dummy variable, = 1 if aged 55-64, = 0 elsewhere

EDUCN minimum number of years of formal education required to obtain highest educational qualification held (full-time equivalent years)

EXPER full-time equivalent years of employment experience; instrumental variable defined as EXPER= - 15.128 + 1.222 AGE - 0.011 (AGE)* - 0.938 KIDS04 - 1.398 KIDS59 - 1.625 KIDS10 NB: this equation was derived from a regression analysis of the experience patterns of women in the 1980 Sydney Survey of Work Patterns of Married Women; see Ross (1986) for a description of that data set.

KIDS04 KIDS59 KIDS10

number of dependent children aged 0-4 number of dependent children aged 5-9 number of dependent children aged 10 and over

EARN Woman’s weekly earnings from employment

TRANS gross annual family income from government cash transfer payments

FAMINC gross annual family income from all sources other than EARN and TRANS

SPWORK = 1 if woman’s spouse is employed full-time = 0.5 if woman’s spouse is employed part-time = 0 otherwise (i.e. if no spouse or if spouse not employed)

NM = 1 if woman is not currently married and living with her spouse (i.e. if woman is separated, divorced, widowed, or has never been married)

= 0 if woman is currently married and living with her spouse

LFPR = 1 if in the labour force (i.e. if in paid employment or unemployed) = 0 otherwise

EMP = 1 if in paid employment = 0 otherwise

LBRFT = 1 if in full-time (i.e. at least 35 hours per week) paid employment = 0 otherwise


The index, @, being estimated is:

@ = (bol - J r i ) + Wc,2 -Jr2) E D u c N +Jo3 EXPER - Jr3 KIDS04 -Jr4 (KIDSO4)' -Jrs KIDS59 -Jh KIDS10

Jr7 FAMINC -JrS SPWORK -Jd TRANS - J r l o AGE1524

J,I I AGE2534 -Jr12 AGE3544 - J r 1 3 AGE4554 -Jr14 NM

Jr15 N M . EDUCN -Jr16 N M . KIDS04 -Jr17 N M . (KIDS04)2

-

-

-

- J r l g N M . KIDS59 N M . KIDS10 -Jrzo N M . FAMINC

- J,2l N M . SPWORK -Jr22 N M . TRANS -Jr23 N M . AGE1524

- Jr24 N M . AGE2534 -Jr2s N M . AGE3544 -Jr2b N M . AGE4554 (4)

The independent variables are defined in Table 111. J,12,J,,3 are both expected to be positive, whiIe,Jr2,Jr3,Jr5, andJ, toJr13 are all expected to be negative. The expected signs onJr4 andJ, are ambiguous, and depend on the existence of (dis)economies of scale in child rearing. The signs onJ,,, t0JrZ6 are discussed later.

The age variables, AGE1524, AGE2534 etc., are included to isolate cohort effects. As the reference group chosen is those women aged 55-64 , a trend to increasing participation rates implies that each ofJrlo,Jrl ,,JrI2, andJ,.,, should be negative.4 The expectations on the relative magnitudes of the age variables are thus:

with the relative sizes ofJrl0 andJ,,, being an empirical matter.

The co-efficients on the set of cross-product terms in (4) (ix. J r I 4 to &6) will indicate the impact on labour force status of sole motherhood per se. IfJr14 is significantly different from zero, this will indicate that sole motherhood results in a 'parallel' shift of the decision function. If any ofJrls are significant, it will indicate that sole mothers respond differently to changes in that specific variable than do married mothers; a negative co-efficient will indicate that sole mothers are less likely to be in the labour force as the value of the variable increases, while a positive co-efficient indicates sole mothers are more likely to be in the labour force with increases in the variable. If all of these co-efficients are insignificant, then sole mothers don't react differently to married mothers. If this is so, then the variations in the labour force status of sole mothers compared to married mothers can be explained entirely by variations in their 'endowments' of characteristics.

Table IV compares sample descriptive statistics for married and sole mothers in the sample, classified by employment status. It is clear that there are significant differences between the two groups of mothers. Sole mothers have fewer children in all age groups than do married mothers. Employed sole mothers are marginally older, marginally more educated, and have marginally more labour force experience than do employed married mothers, but the reverse is true of

Previous research has found an inverted-U shape age effect; see Miller and Volker (1983) and Ross (1986).


mothers not in employment. Overall, the age distribution of sole mothers is more even than that of married mothers. Eighty per cent of married mothers are aged between twenty-five and forty-four, whereas only sixty-six per cent of sole mothers are in this age range. However, the biggest differences are in the income variables. Compared with married mothers, sole mothers have three times as much income from government cash transfer payments, but only around one tenth of income from all other sources (i .e. FAMINC). In addition to not having access to the earnings of an employed spouse, sole mothers have less than half as much income from all other sources than is the case for married mothers. For example, married mothers had spouses who, on average, earned around $19,142 a year and had other family income of $5,968 a year on average. By contrast, sole mothers had (by definition) no spouse income and the other sources of income averaged only $2,109 a year.

v. EMPIRICAL RESULTS

The results of the probit analysis are presented in Table V, which contains estimates for three likelihood functions. Column 1 presents the estimates of the parameters of the index of the likelihood of a mother being in the labour force ( i . e . being either employed or officially unemployed, indicated by LFPR = 1; in Table 111). Column 2 presents the estimates of the parameters of the index of the likelihood of a mother being successful in the labour market, i.e. actually being in employment (i.e. EMP = 1 in Table 111). Column 3 presents the estimates of the parameters of the index of the likelihood of an employed mother being in full-time employment rather than part-time employment (i.e. LBRFT = 1 in Table 111).

The results in each column of Table V will be discussed in turn. Within these discussions, the focus will initially be on the impact of each of the key variables and their associated co-efficients, i . e . J,, to Jo3. and J,, to Jr13. The central research question of the differential labour market behaviour of sole mothers will then be addressed by focussing on the set of cross-product terms and their co-efficientsJ,,, toJrZ6.

Labour force participation Column 1 of Table V indicates that labour force participation is determined by both economic

and demographic variables. Previous research on female participation has indicated that the presence of children, especially very young children, has a major, negative influence on participation; see, e g . Eyland et al. (1982), Miller and Volker (1983), and Ross (1986). The present results reinforce this conclusion. Pre-school aged children have a very large negative impact on participation, although there are strong economies of scale evident; it is the first such child which has the biggest impact on participation, with the marginal impact of additional pre- schoolers in the family declining as the number of children increases. Evaluated at sample means, the presence of one pre-school aged child lowers the probability of participation by 22 percentage points, while the marginal impact of a second pre-school aged child is a further reduction of 11 percentage points. The presence of children of primary school age also reduces the participation probability, but by a much smaller degree than the presence of pre-school age children; e.g. the impact of one primary school aged child reduces the participation probability by 3.7 percentage points. The estimates indicate that the presence of older children (aged ten and over) has no impact on the labour force participation decisions of their mothers.


TABLE IV

Variables Used in the Analysis, Mean Values by Employment Status for Married and Sole Mothers, 1985-86

Married Mothers

Sole Mothers

Mnemonic Employed Not Total Employed Not Total Employed(a) Employed( a)

AGE 36.3 1 35.00 35.70 37.28 33.55 34.95

AGE1524 AGE2534 AGE3544 AGE4554 AGE5564

0.0382 0.3620 0.4752 0.1189 0.0057

0.0785 0.4468 0.3464 0.1026 0.0157

0.062 1 0.0532 0.3973 0.3259 0.4100 0.4011 0.1 150 0.1980 0.0156 0.0218

0.1991 0.3673 0.3142 0.1018 0.0176

0.1462 0.3369 0.3558 0.1402 0.0209

KIDS04 KIDS59 KIDS 10 (total number of children)

0.4104 0.5210 1.0240 1.9554

0.7268 0.5925 0.79 17 2.1110

0.5614 0.2777 0.5498 0.3767 0.9144 0.9493 2.0256 1.6036

0.5885 0.5089 0.7345 1.8319

0.4792 0.466 1 0.8026 1 .I479

EDUCN (yrs) 1 1.74

11.21

0.946

11.19 11.48 12.08

10.77 11.94

0.891 0

11.02

9.85

0

11.42

10.67

0

EXPER (yrs) 10.34

SPWORK 0.828

0 198.89 107.18 276.66 0 114.31

26,573 23.937 25.1 10 2,378 1,94 1 2.109

TRANS ($ pa) 850

0

1,881 1,336 1,966 5,260 3,989

NM 0 0 1 .om

2,841 135

1 .ow

226

1 .000

36 1 Sample size Population estimate

1,516

970,622

1,325

8 17,6 17 1,788,239 85,832 114.699 200.53 1

Nore: (a) i.e. the Unemployed and Not in the Labour Force categories. Source: ABS, 1986 Income Distribution Survey: unit record file.


The co-efficients on the four age dummy variables suggest that the age profile of participation does have the familiar inverted-U shape, peaking in the 25 to 34 age range, although the profile is fairly flat around its peak. However, while all four age groups have higher participation probabilities than the reference group, the co-efficients on the first three age variables are not statistically significantly different from each other, although each is statistically significantly higher than that for AGE4554, which in turn is greater than that for the reference group. After adjusting for all other variables, the results indicate that the age profile of participation is very flat up to age forty-five, after which participation declines with age.

TABLE V

Probit Ana1.W Results, All Mothers

Co-efficient Estimate

Dependent Variable LFPR EMP LBRFT (No. of observations) 3202 3202 1651

Regressor

Constant -1.5388"

KIDS04

KIDS042

KIDS59

KIDS10

AGE1524

AGE2534

AGE3544

AGE4554

EDUCN

EXPER

(0.47 1 I )

-0.6535" (0.1 170)

0.0813** (0.047 1 )

-0.0928** (0.0480)

0.0095 (0.0485)

1.0058" (0.3435)

1.1396* (0.2632)

1.0252* (0.2285)

0.6603" (0.2288)

0.085 1 * (0.0123)

-0.0003 (0.0233)

-2.5146" (0.51 13)

-0.4643" (0.1210)

0.035 1 (0.0502)

-0.0663 (0.0494)

0.0733 (0.0503)

1.1430" (0.3738)

1.2714" (0.2867)

1.1134" (0.2470)

0.7193* (0.2441)

0.0843* (0.0122)

0.0419** (0.0247)

-1.2582"" (0.7845)

-0.2842 (0.1838)

0.007 1 (0.0856)

-0.1970" (0.0725)

-0.0086 (0.07 12)

0.9059 (0.6103)

0.548 1 (0.4930)

0.3346 (0.4494)

0.0796 (0.4463)

0.0538* (0.0152)

0.0585** (0.0358)

1993 THE LABOUR SUPPLY BEHAVIOUR OF MOTHERS IN AUSTRALIA

FAMINCI1,OOO

TRANSI1,OOO

SPWORK

NM

NM.KIDSO4

NM.KIDS042

NM.KIDS59

NM.KIDS 10

NM.AGEl524

NM.AGE2534

NMAGE3544

NM.AGE4554

NM.EDUCN

NM.EXPER

NM.FAMINCII,OOO

NM.TRANSl1 ,000

-0.0024* (0.0012)

-0.0853* (0.01 45)

0.3 159* (0.1003)

-0.1391 (1.5 160)

-0.0385 (0.3758)

0.0869 (0.1571)

-0.0276 (0.1780)

0.3220* (0.163 1)

-0.2543 (1.0670)

-0.271 1 (0.7657)

-0.6371 (0.6109)

-0.5681 (0.5970)

0.01 34 (0.0390)

0.0656 (0.0776)

-0.0710* (0.0191)

-0.0918* (0.0283)

-0.0022** 0.0012)

-0.1072* (0.0163)

0.4899* (0.1032)

0.1406 ( 1 S980)

-0.2680 (0.3999)

0.1949 (0.1637)

-0.0 100 (0.1914)

0.1353 (0.1736)

-0.2789 (1.1250)

-0.4295 (0.7984)

-0.9894 (0.6324)

-0.6902 (0.6138)

0.0188 (0.0397)

0.0835 (0.0826)

-0.0572* (0.0187)

-0.1004* (0.03 12)

127

-0.0008 0.00 16)

-0.041 3 (0.0335)

-0.4284* (0.1613)

1.1122 (2.6210)

0.0352 (0.7044)

0.1054 (0.296 1 )

0.25 18 (0.3090)

0.1169 (0.2820)

-0.6074 (1 3570)

-0.5930 ( 1.3 100)

-0.1 125 (1.05 10)

0.0801 (1.01 10)

-0.0257 (0.0594)

-0.0140 (0.1336)

-0.0473** (0.0259)

-0.196 1 * (0.0637)

Nores: Reference group are sole mothers aged 55-64 years. Standard errors in parentheses. * indicates co-efficient significant at the 95% confidence level. ** indicates co-efficient significant at the 90% confidence level.


The co-efficient on EDUCN indicates that more highly educated mothers are more likely to be in the labour force. This co-efficient is the net effect of two opposing effects, i.e. in (4) the co- efficient is -Jr2) . The fact that the net effect is positive (the co-efficient value is + 0.0851) indicates that an extra year of formal education is more highly valued by employers than by the women themselves. Curiously, the impact of previous labour force experience (EXPER) appears to have no impact on current participation, although this may simply be due to the way in which the experience variable has had to be represented (see Table 111).

The impact of the income variables is in accord with economic theory. Mothers with greater access to other sources of income are less likely to be in the labour force. The impact on participation of changes in the level of transfer payments (TRANS) is much greater than that of other sources of family income (FAMINC). The co-efficient on TRANS ( -0 .853 x 10-4) is considerably larger in absolute magnitude than that on FAMINC (-0.024 x 10-4). Evaluated at sample means of all other variables, these co-efficients suggest that a $5,000 increase in TRANS will decrease the likelihood of a mother being in the labour force by some sixteen percentage points, whereas the same increase in FAMZNC will reduce the likelihood of labour force participation by only one-half of one percentage point. Mothers with working spouses are more likely to be in the labour force themselves than are mothers without working spouses. The co- efficient on SPWORK indicates that, ceterus paribus, the likelihood that a mother with a working spouse will be in the labour force is some 18.7 percentage points higher than that for a similar woman who either has no spouse or whose spouse is not working.

The differential impact of marital status on participation can be seen in the set of co-efficients on the cross-product terms involving the NM variable, i.e. the co-efficientsJrI4 toJr2@ Any differential impact of sole parenthood will be captured in these cross-product terms. The total responsiveness of sole parents to changes in a specific variable is reflected in the sum of the co-efficients on the cross- product term and the non cross-product term for the same variable. For example, the differential impact of one primary school aged child on the labour market position of a sole mother is reflected inJr18 (i.e. the co-efficient on NM.KIDS59) where as the total impact of that child on a sole mother is reflected inJr18 +Jrs (i.e. the sum of the co-efficients on NM.KIDS59 and KIDS59).

These co-efficients indicate that sole mothers respond to labour market signals in the same manner as do married mothers, with three important exceptions. All the co-efficients on the cross- product terms, including that on NM alone, are insignificant except for those on the two income variables (FAMINC and TRANS) and that on NM.KIDS10. The insignificant co-efficient on NM indicates that there is no ‘parallel’ shift in the behaviour of sole mothers compared to their married counterparts, while the other insignificant co-efficients indicate a similar conclusion is true with respect to each of those variables individually.

However, sole mothers do react to changes in the two income variables considerably more strongly than do married mothers. The co-efficient on NM.TRANS (i.e. -0.918 x 10-4) indicates that the total effect of this variable is double that for married mothers. For a married mother, the effect is reflected completely by the co-efficient on TRANS (i.e. it is equal to -0.853 x 10-4), while for sole mothers the effect is reflected in the sum of the co-efficients on TRANS (-0.853 x 10-4) and NM.TRANS (-0.918 x lO-4), i.e. the net co-efficient is -1.771 x 10-4. Evaluated at sample means, the implied impact of a $5,000 increase in TRANS is to reduce the likelihood of participation for sole mothers by some thirty two percentage points; i.e. by double the estimated reduction for a married mother.


The hypothesis that the co-efficients on NM.FAMINC and NM.TRANS are equal cannot be rejected at usual levels of confidence. Thus, the differential impact of changes in income sources for sole mothers is independent of the source of that income. However, the overall impact of changes in FAMINC is not the same as that for TRANS. This is because the underlying impacts (as measured by the co-efficients on TRANS and FAMINC) are different. Nevertheless, after adjusting these co-efficients by the differential impact of NM.TRANS and NM.FAMINC, much of the gap is eliminated. The adjusted co-efficient on FAMINC is -0.734 x 10-4, while that for TRANS is -1.771 x 10-4. To illustrate the importance of this difference, recall the above comparison of a $5,000 increase in TRANS and the same increase in FAMINC. For a married mother, the relative impacts were 15.5 percentage points (TRANS) and 0.5 percentage points (FAMINC). For a sole mother, these impacts are 32.5 percentage points (TRANS) and 14.8 percentage points (FAMINC).

Although the co-efficient on KIDS10 is insignificant, that on N M . K I D S l 0 is significant. Thus sole mothers with children aged ten or over are more likely to be in the labour force than are married mothers, other things being equal. Whereas the impact of older children on participation is insignificant for married mothers, there is a positive, and fairly strong, impact for sole mothers. Evaluated at sample means, the impact of the presence of one older child is to raise the likelihood of a sole mother’s labour force participation by around eleven percentage points.

Employment

There are four major differences in the results presented in Column 2 of Table V compared to those in Column 1 . First, there are no economies of scale evident in the presence of pre-school aged children; i .e . the co-efficient on (KIDSO4)z is no longer significant. Second, primary school aged children appear not to have any impact on employment, i.e. the co-efficient on KIDS59 is also no longer significant. Third, previous employment experience (EXPER) is now significant, with mothers with more previous experience being more likely to be currently employed. Fourth, the co-efficient on the cross-product term between NM and KIDS10 is not statistically significant. These results indicate that unemployment, i .e. the difference between participation and employment, might not be randomly spread among mothers.

In all other respects, the results in Column 2 reflect those in Column 1 . With the exception of SPWORK, for every variable which is statistically significant in Column 2, the estimated co- efficient is within one standard error of the estimate presented in Column 1 . The co-efficient on SPWORK in Column 2 is more than one standard error greater than that in Column 1, although the difference is not quite as much as twice the standard error on the Column 1 estimate. For all the variables with significant co-efficients, the story is the same as in Column 1 .

The differential impact of sole parenthood on employment is reflected completely in the responsiveness of sole mothers to changes in the levels of other sources of income, since the cross product with KIDS10 is not significant. FAMINC and TRANS are the only two variables for which the response of sole mothers is different to that for married mothers. For both of these variables, sole mothers are more responsive than married mothers ( i . e . the co-efficients on NMFAMINC and NM.TRANS are both negative). Interestingly, unlike the estimates in Column 1 , these two co-efficients are not statistically identical, with the marginal impact on employment of transfer payments being twice that of other sources of income.

I30 AUSTRALIAN ECONOMIC PAPERS JUNE

Full-time eniployment versus part-time employment The results in Column 3 of Table V, which relate only to employed mothers, complement the

picture presented in the first two columns although there are some important differences to note. In this column, a negative co-efficient indicates that an increase in the characteristic leads to a lower likelihood of full-time employment, and therefore a correspondingly higher likelihood of part-time employment. Working mothers with primary school aged children are more likely to be in part-time employment than in full-time employment. Curiously, the presence of one or more pre-school aged children has no impact on the type of employment gained. That is, the insignificant co-efficients on the variables KIDS04 and (KIDS04)z indicate that mothers with pre- school aged children who are employed are just as likely to be in full-time employment as in part-time employment. Given that these data are from a time when the availability of long day child care was fairly scarce, it might be expected that employed mothers with pre-school aged children would be more likely to be in part-time employment than in full-time employment as these children were more likely to be in kindergartens than in long day care.5

The age profile is now completely flat. All age groups, including the reference group, are equally likely to be in full-time employment. More highly educated mothers are more likely to be in full-time employment, as are mothers with more previous employment experience. Mothers with working spouses are more likely to be in part-time employment. Interestingly, neither FAMINC nor TRANS has any impact on the type of employment. It might have been expected that these income variables would have negative co-efficients as mothers with higher levels of transfer payments or other non-wage income would, via the income effect, be more likely to be in part-time employment. That was not, however, the case.

Turning to the cross-product terms, again only the two income variables are significant. For all other characteristics, the likelihood of an employed sole mother being in full-time employment (rather than part-time employment) is exactly the same as it is for an employed married mother. Sole mothers who are employed are more likely to be in part-time employment if they have some other income, while working sole mothers in receipt of transfer payments are also less likely to be in full-time employment.

Sitnzmary of results The results presented in Table V can be summarised as supporting the following propositions:

Mothers with pre-school aged children are the most unlikely to be in the labour force and in employment, but if they are employed they are just as likely to be in full-time employment as they are to be in part-time employment;

Mothers with primary school aged children are more likely to be in the labour force and to be employed than are mothers with pre-school aged children, but less likely than mothers with older children;

More highly educated mothers are more likely to be employed than both mothers with non- working spouses and sole mothers, and are more likely to be in full-time employment rather than part-time employment;

' Long day care refers to child care which is available for at least eight hours per day five days per week, and is to be contrasted with kindergarten or pre-school care which is typically available for no more than six hours per day and for less than five days per week.


Mothers with working spouses are more likely to be in the labour force and to be employed, but are more likely to be in part-time employment than in full-time employment;

Mothers with more employment experience are no more likely to be in the labour force, but if they are in the labour force they are more likely to be employed and if employed in full-time jobs;

The age profile of employment is very flat. After allowing for all other factors, all mothers aged up to forty-five are equally likely to be in the labour force and to be employed; and

Although sole mothers respond to most labour market signals in exactly the same ways as do married mothers, there is one important exception. Sole mothers are far more responsive to income changes than are married women. This is especially true in relation to transfer payments, but is also significant in relation to other sources of income.

VI. CONCLUSIONS

There is now a body of evidence pointing to the poor financial and social circumstances of many sole parent families in Australia. Rates of dependency on income support among sole parents are high, although the duration of income support receipt is in many instances not unduly long. But such dependency, combined with the fact that this is the only form of income for many sole parent families, serves to place them towards the lower end of the income distribution and thus with a high incidence of poverty.

One way of addressing this situation is to provide sole parents with greater access to employment opportunities. A number of policy initiatives have been introduced in recent years with this aim in mind, including the provision of priority access to child care and measures to reduce the severity of the poverty trap facing sole parents in receipt of income support. It is interesting, however, to observe that the pattern of participation rates of sole mothers in full-time employment are very similar to those of married mothers with a youngest child of the same age. In contrast part-time employment among sole mothers is well below that of married mothers with a youngest child of the same age. This latter evidence is suggestive of the fact that high effective marginal tax rates arising from interactions between the tax system and income-tested social security arrangements have resulted in disincentive effects for sole mothers, at least in relation to part-time work.

This paper has investigated this issue by estimating labour supply functions for sole mothers and comparing these with the labour supply functions of married mothers. The focus of the analysis has been on establishing whether the labour supply behaviour of the two groups is different, and if so why. Specifically, the issues addressed relate to the question of whether or not being a sole mother makes any substantial difference to work decisions over and above those factors contributing to the work decisions of mothers. per se.

The evidence presented in Table V and discussed in Section V suggests that sole mothers respond to labour market signals in much the same way as do married mothers, with the important exception that they (sole mothers) are much more responsive to changes in access to other sources of income - particularly government cash transfer payments but more generally to all non-wage income sources. Overall, the findings reported in the paper are consistent with the view that the labour supply behaviour of mothers in Australia has more to do with the fact that they are mothers (and, more significantly, with the age of their youngest child), than whether or not there is an adult partner present. This suggests that the emphasis in understanding the labour supply decisions of sole mothers should be placed more on the fact that they are mothers, rather than on the fact that they are sole mothers.


REFERENCES

Australian Bureau of Statistics (1985), 1981-82 Income and Housing Survey: Unit Record File.

Australian Bureau of Statistics (1987), 1986 Income Distribution Survey: Unit Record File.

Australian Bureau of Statistics, Labour Force Status and Other Characteristics of Families, Catalogue No. 6224.0; various issues.

Australian Bureau of Statistics, The Labour Force. Catalogue No. 6203.0; various issues.

Blundell, R. and Walker, I. (1983), ‘Estimating a Life Cycle Consistent Model of Family Labour Supply with Cross Section Data’, mimeo, Departments of Econometrics and Economics, University of Manchester.

Brown, J.C. (1989), ‘Why Don’t They Go To Work? Mothers on Benefit’, Research Paper 2, Social Security Advisory Committee, London.

Cass, B. (1986), ‘The Economic Circumstances of Sole Parent Families in Australia 1974-1985: Some Implications for Child Maintenance Policies and the Social Security System’, presented to a Conference on Child Support Issues, Australian National University, Canberra, March.

Eyland, E.A., Lapsley, H.M. and Mason, C.A. (1982), ‘Determinants of Female Labour Supply’, Economic Record, vol. 58.

Heckman, J.J. (1980), ‘Sample Selection Bias as a Specification Error with an Application to the Estimation of Labour Supply Functions’ in J.P. Smith (ed.), Female Labor Supply: Theory and Estimation (Princeton, New Jersey: Princeton University Press).

Heckman, J.J. and MaCurdy, T. (1980), ‘A Life Cycle Model of Female Labor Supply’, Review of Economic Studies, vol. 47.

Heckman, J.J., Killingsworth, M.R. and MaCurdy, T. (1981), ‘Empirical Evidence on Static Labour Supply Models: A Survey of Recent Developments’, in Z. Hornstein, J. Grice and A. Webb (eds), The Economics of the Labour Market (London: HMSO).

Killingsworth, M.R. (1983), Labor Supply (New York: Cambridge University Press).

McDonald, J. and Spindler, Z.A. (1988), ‘Benefit-Induced Female Sole Parenthood in Australia, 1973-85’, Australian Economic Papers, June.

Miller, P. and Volker, P. (1983), ‘A Cross-Section Analysis of the Labour Force Participation of Married Women in Australia’, Economic Record, vol. 59.

Murray, C. (1984), Losing Ground: American Social Policy, 1950-1980 (New York: Basic Books).

Raymond, J. (1987), Bringing Up Children Alone: Policies for Sole Parents, Issues Paper No. 3, Social Security Review, Department Social Security, Canberra.

Ross, R. (1986). ‘Analysis of the 1980 Sydney Survey of Work Patterns of Married Women: Further Results’, Economic Record, vol. 62.

Ross, R. and Saunders, P. (1990), ‘The Labour Supply Behaviour of Single Mothers and Married Mothers in Australia’, Discussion Paper No. 19, Social Policy Research Centre, University of New South Wales, Kensington.


Saunders, P. and Matheson, G. (1990), ‘Sole Parent Families in Australia’, Discussion Paper No. 23, Social Policy Research Centre, University of New South Wales, Kensington.

Social Security Review (1986), Labour Force Status and Other Characteristics of Sole Parents; 1974-1985, BackgroundlDiscussion Paper No. 8, Department of Social Security. Canberra.

Smeeding, T. M. and Torrey, B. B. (1988), ‘Poor Children in Rich Countries’, Science, vol. 242, November.

Whiteford, P., Bradbury, B. and Saunders, P. (1989), ‘Poverty Traps in the Australian Social Security System’, Economic Analysis and Policy, vol. 19, no. 1.

the labour supply behaviour of sole mothers and married mothers in australia: evidence from the 1986...

Documents