The Impact of Paid Parental Leave on Labor Supply and Employment Outcomes in Australia

ABSTRACT

The introduction of the Australian Paid Parental Leave scheme in 2011 provides a rare opportunity to estimate the impacts of publicly funded paid leave on mothers in the first year postpartum. The almost universal coverage of the scheme, coupled with detailed survey data collected specifically for the scheme’s evaluation, means that eligibility for paid leave under the scheme can be plausibly taken as exogenous, following a standard propensity score-matching exercise. Consistent with much of the existing literature, the study finds a positive impact on mothers’ taking leave in the first half year and on mothers’ probability of returning to work in the first year. The paper provides new evidence of a positive impact on continuing in the same job under the same conditions, where previous conclusions have been mixed. Further, it shows that disadvantaged mothers – low income, less educated, without access to employer-funded leave – respond most.

HIGHLIGHTS

• Studying the effects of introducing paid parental leave (PPL) in an advanced industrial country is important for the US, which is considering PPL.

• PPL was introduced in Australia in 201 l. Previously only 57 percent of 20–45-year-old women had access to paid parental leave provided by employers.

• Post-PPL, mothers initially return to work from leave more slowly than before, but after about six months of leave they return to work at a faster rate than pre-PPL.

• Post-PPL, the probability of returning to work within a year is higher than pre-PPL.

• PPL helps mothers balance paid work and family life and improves workplace attachment. Both effects are, on average, stronger among more disadvantaged groups.

KEYWORDS:

• Labor supply
• parental leave
• Australia

JEL: Codes:

• J13
• J18
• J22

ACKNOWLEDGMENTS

This paper is based on commissioned research for the Australian Government Department of Social Services (DSS). We gratefully acknowledge funding by DSS and the opportunity to evaluate the newly introduced Paid Parental Leave. We also acknowledge support for Duncan McVicar’s 2015 visit to Melbourne from the Visiting Research Scholar scheme of the Faculty of Business and Economics, University of Melbourne. Neither DSS nor the University of Melbourne is responsible for the views in this paper. We thank Katrien Stevens and participants at the 2015 Labor Econometrics Workshop in Sydney, the 2016 European Society for Population Economics Conference in Berlin, the 2016 Work and Family Researchers Network Conference in Washington DC, and seminars at the Tax and Transfer Policy Institute at the Australian National University, the Business School at the University of Western Australia, Queen’s University Belfast, University College Dublin, the Economic and Social Research Institute, and Cardiff University, as well as three anonymous referees for their helpful comments. The views expressed in this paper and any remaining errors are those of the authors solely.

SUPPLEMENTAL DATA

Supplemental data for this article can be accessed at https://doi.org/10.1080/13545701.2020.1718175.

Notes

1 For details of the proposed Family and Medical Leave Act, introduced in the US Senate in 2015, see https://www.congress.gov/bill/114th-congress/senate-bill/786/text. In more recent developments, Senator Kirsten Gillibrand reintroduced a similar bill in February 2019, see https://www.congress.gov/bill/116th-congress/senate-bill/463, while an identical bill was also introduced in House of Representatives, see https://www.congress.gov/bill/116th-congress/house-bill/1185/related-bills. Other, mainly Republican, bills were also introduced in 2019, see https://www.congress.gov/bill/116th-congress/house-bill/1940.

2 Marit Rønsen and Ragni Hege Kitterød (2015) find evidence from Norway that this effect can be counteracted by policies that decrease childcare prices or by family leave arrangement targeted at fathers, while Carmen Castro-García and Maria Pazos-Moran (2016) add that this is only effective if leave for both parents is equally long, non-transferable, and provides a relatively high payment rate.

3 See Joseph et al. (2013); Rossin-Slater, Ruhm, and Waldfogel (2013); Lalive et al. (2014); Schönberg and Ludsteck (2014); Dahl et al. (2016); Baum and Ruhm (2016).

4 Selective eligibility for employer-provided paid parental leave is also likely to be an issue here, although neither Han, Ruhm, and Waldfogel (2009) nor Rossin-Slater, Ruhm, and Waldfogel (2013) discusses this.

5 See Baum and Ruhm (2016) on this point.

6 The term “casual employment” is understood differently in Australia compared to many other countries (where it is typically associated with irregular working hours, and no expectation of continued employment). In Australia, so-called casual employment is often of long duration, and working hours and earnings may or may not be steady over time. Instead, it is commonly understood that casual employees are those who are not entitled to paid annual leave and paid sick leave; this is also the definition used in surveys collected by the Australian Bureau of Statistics (Buddelmeyer, Wooden, and Ghantous 2006). In contrast, permanent and fixed-term contracts provide paid leave entitlements, with fixed-term contracts having a specified end date for the employment contract and permanent contracts having no specified end date. The survey used in this study asks respondents to self-identify as being employed on a (i) permanent basis, (ii) fixed-term basis, (iii) casual basis, or (iv) other.

7 Age at birth was not asked directly but derived from information on the mother’s age at time of the interview (recorded in full years) minus the child’s age at the time of interview (recorded in days); it is thus subject to some minor measurement error.

8 The number of other children in the household is not recorded at the time of the study child’s birth, but at the time of the interview. To provide a better approximation of the number of children at the time of birth, we disregard any other children who were less than 1 year old at the time of interview. This affected three families in the post-PPL cohort and sixteen families in the pre-PPL-cohort. These are likely to be children who were born after the study child.

9 The Australian Socioeconomic Index 2006 (AUSEI06) assigns a “status score” to each occupation, coded according to the Australian and New Zealand Standard Classification of Occupations (ANZSCO). The scale is a continuous measure that ranges from 0 to 100.

10 We define an industry as female dominated if more than 50 percent of the employed persons in that industry are women, regardless of full-time or part-time status. We used Table Builder Basic to retrieve that information from the 2011 Census (ABS 2011b). Female-dominated industries are (i) retail trade, (ii) accommodation and food services, (iii) financial and insurance services, (iv) rental hiring and real estate services, (v) administrative and support services, (vi) education and training, and (vii) healthcare and social assistance.

11 The main advantage of the Cox estimate over the KM estimate is that the former provides us with a hypothesis over the signs of two parameters that is easily testable and makes the effect of PPL easy to interpret. However, the KM-estimator is more flexible in how the impact of PPL is allowed to vary with the time that has passed since birth, as no functional form for the hazard rate is imposed.

12 We also estimate a version of the model that includes an additional quadratic effect of PPL, and a version with only the constant impact of PPL, ${\text{β}}_1$

13 Not included in the matching are the child’s age at time of the interview and the indicator for female-dominated industry (since the finer measurement of industry is included). Income is included in 2012 $AUD only.

14 The interpretation as “average treatment effect” in this context holds only if we interpret having access to PPL as a treatment. However, if we interpret taking PPL as the treatment, introducing access to PPL in January 2011 is called the “intention to treat,” and the procedure described above yields the so-called “intention-to-treat effect” (ITT). From a policy perspective this ITT effect is the more interesting result since the non-use is not due to a lack of information or other hurdles preventing parents from accessing PPL; it is a conscious choice.

15 Figure 1 shows the distribution of propensity scores for the full set of mothers; this is the sample we use for the analysis of the duration of return to paid work. The analysis of the job characteristics upon return relies on the subset of mothers who have returned to paid work, and another subset of mothers is used for those who have returned and changed some of their job characteristics. Figures analogous to Figure 1 for these two subsets are shown in Appendix A.1 (Supplemental Data tab, online).

16 Note that any “untreated individual” can be a matching partner for several different “treated” individuals.

17 However, a sensitivity analysis with the “off support” mothers included does not change the results. They are presented in Appendix A2.

18 Sensitivity tests with alternative approaches are carried out. Nearest neighbor matching with one, two and five neighbors, and radius caliper matching with caliper values of 0.1, 1, 5, and 10 percent are applied. All point estimates remain very similar. The results are presented in Appendix A.2.

19 For the full sample, the matching quality is shown in Table A.2. For the analysis of job characteristics upon return to paid work, the sample is restricted to include only those mothers who actually returned to paid work. The propensity score is reestimated for the restricted sample, and the same matching procedure is applied afterward. The matching quality for the subsample of returned mothers is reported in Table A.3 of Appendix A.3. Similarly, for the analysis by subgroups, the matching procedure is applied after restricting the sample to the subgroup in question, which is equivalent to exact matching on the group indicator, combined with on all other characteristics.

20 This is because mothers can receive PPL for children born on 1 January 2011 at the earliest, and they must have had some employment in the ten months before that, that is, from March 1, 2010 onward. The date from which employment histories could be relevant for future eligibility also coincides with the announcement of the policy.

21 The survivor function based on estimation without applying the propensity score-matching approach, showing a very similar pattern, is provided in Appendix A.4.

22 The equivalent estimates without matching are given in Appendix A.4.

23 The simplest model, restricting the impact of PPL to be duration-invariant, gives very different and misleading results, because the initial negative and eventual positive impacts, which are likely to continue beyond the first year, approximately average out within the first year after birth.

24 In order to see whether the main source of disadvantage in this context is mothers’ low income or the lack of access to employer-paid leave, we repeated the analysis comparing four groups: low-income and high-income mothers, each with and without access to employer-paid leave. It appears that the initial slowdown in mother’s return to paid work is driven primarily (although not exclusively) by low-income status (with PPL these women can now afford longer leave), while the increased return to paid work later on is primarily (but not exclusively) driven by a lack of access to employer-paid leave (with PPL there appears to be a stronger connection to the employer). The results are presented in Appendix A.5.

25 Single mothers are defined here as those not living with a partner at the time they gave birth. Although a few of these mothers had partners (whom they were not living with), the vast majority did not have a partner.

26 The job under consideration is the first job upon return after childbirth.

27 As mentioned earlier, we test the robustness of our results using the uncensored sample of mothers who had their interview on or after the child’s first birthday, instead of the full sample. Appendix A.6 shows very similar statistics for destination states (return to different types of jobs or no job) for the both samples. We find similar results if we repeat the analysis in Table 6 unconditional on having returned to any job. That is, 49 percent of all post-PPL mothers returned to the same job within one year versus 47 percent of all pre-PPL mothers, and 21 percent of all post-PPL mothers versus 18 percent of all pre-PPL mothers returned to the same job and same conditions within one year (see Appendix A.6).

Additional information

Funding

This work was supported by the Department of Social Services, Australian Government; University of Melbourne [a Visiting Research Scholar scheme grant from the Faculty of Business and Economics].

Notes on contributors

Barbara Broadway

Barbara Broadway is Research Fellow in the Labor Economics and Social Policy program at the Melbourne Institute of Applied Economic and Social Research (University of Melbourne). She is an applied econometrician who is interested in labor economics, in particular, female and maternal labor supply, its impact on families, and its interplay with social policy. Recent publications include a study on the impact of paid parental leave on children’s health in the Economic Record and on gender differences in doctors’ labor supply in Health Economics.

Guyonne Kalb

Guyonne Kalb is Professorial Research Fellow and Director of the Labor Economics and Social Policy program at the Melbourne Institute of Applied Economic and Social Research (University of Melbourne). She is an applied econometrician with a research interest in labor supply and childcare and child development, and in the interactions of all three with social policy. Recent publications include the impact of parental reading to children on child outcomes in Economics of Education Review and the intergenerational correlation of labor market outcomes in Review of Economics of the Household.

Duncan McVicar

Duncan McVicar is Professor of Economics at Queen’s Management School, Queen’s University Belfast. His research interests include program evaluation and welfare reform; disability, unemployment, and inactivity; education, peer effects, and substance use; and sequence analysis of labor market data. Recent publications include a study of the links between substance use and homelessness in Social Science and Medicine, a study of the impact of early marijuana use on educational outcomes in the Economic Record, and a study of job satisfaction in non-standard employment in Industrial Relations.

Bill Martin

Bill Martin was Honorary Professor at the Institute for Social Science Research (University of Queensland). His research interests included the sociology of work and employment, including work and family, work quality, skills utilization, retirement pathways and institutions, labor markets, and inequality. His applied research produced important reports on many topics, including the impact of paid parental leave and the structure and dynamics of social care workforces. Sadly, Bill passed away in April 2016. This paper would not have been possible without his inspired leadership of the Paid Parental Leave project.