Parental leave length and mothers’ careers: what can be inferred from occupational allocation?

ABSTRACT

This paper shows that the time spent on parental leave affects mothers’ careers several years after childbirth. It also shows that policy-relevant conclusions can be drawn from occupational allocation data even in the absence of individual wage or earnings information. I take advantage of a legislative reform that changed the duration of parental benefit payments effective in the Czech Republic in 2008. Introducing the possibility of the flexible timing of benefit collection over the period of 2–4 years, as opposed to the only option of 4 years before, the reform significantly increased the share of mothers returning to work before their youngest child turns four. This further translates to increased representation of employed mothers in high-skilled occupations and increases the average occupation wage of the affected mothers 6 to 8 years after childbirth. These findings indicate that shorter leaves are beneficial for mothers’ careers, at least in the medium run.

KEYWORDS:

• Parental leave length
• parental benefits
• occupational allocation
• mothers’ careers

JEL CLASSIFICATION:

• J13
• J18
• J24

Acknowledgments

Michal Šoltés provided excellent research assistance. The author gratefully acknowledges support from the Grant Agency of the Czech Republic grant No. 14-26574P. The Labor Force Survey data used in this paper were provided by the Czech Statistical Office. The author bears the sole responsibility for all the remining errors and conclusions.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹ There is large literature documenting occupational gender segregation (Blau and Kahn 2017). One of the explanations of occupational segregation is childbearing. Nevertheless, there is no research directly analyzing the relationship between the time women spend off the labour market after childbirth and occupational segregation.

² A similar challenge is faced by studies analyzing wage effects of time spent on parental leave.

³ Employment rates among Czech women with school-age children are among the highest in Europe.

⁴ Pre-leave earnings correspond linearly to the level of parental benefits for mothers earning below a specific threshold, and there is a cap on the daily amount of maternity leave.

⁵ This is a replication of the findings of Bicakova and Kaliskova (2016) who analyse the effects of the 2008 reform (and of a 1995 reform) on mothers’ non-employment, inactivity and unemployment rates.

⁶ Education is coded using the ISCED system. I recode it into three education groups: Low (ISCED 1,2,3c) corresponding to primary, lower secondary and secondary without graduation exam, Medium (ISCED 3a, 3b, 4) corresponding to high school education, and High (ISCED 5,6) corresponding to college or higher levels of education.

⁷ Age increases by one after one’s birthday. Thus, an individual born in May 2000 is recorded as 9 years old in the first quarter of 2010, and as 10 years old in the third quarter of 2010. Taking advantage of the rotating panel structure of the data I can thus retrieve each child’s age with precision up to a quarter.

⁸ Robustness check using data prior to 2012 and direct definition of parental leave provide almost identical results as the analysis using the adapted definition of childcare leave.

⁹ In 2011 there was a significant change in the system of occupational classification used in the Czech Labor Force Survey. Prior to this year information on occupations was recorded using a Czech system of occupational classification, KZAM, while in 2011 the Czech Statistical Office officially started to code occupations according to ISCO. In the dataset used for analysis presented in this paper, all occupations are coded using the ISCO classification. Data prior to 2011 were recoded from KZAM to ISCO by the statistical office.

¹⁰ 2-digit ISCO occupation classification is outlined in the Appendix.

¹¹ http://ispv.cz/en/homepage.aspx.

¹² A similar approach was taken by Bicakova and Kaliskova (2016) who analysed the effect of the 2008 reform on mothers’ non-employment, unemployment and inactivity rates.

¹³ See the Appendix for detailed definition and discussion of treatment and control group choices.

¹⁴ See Appendix B for a detailed explanation of the identification strategy and discussion of treatment and control group choices.

¹⁵ See the Appendix for maternal employment rates by the age of their youngest child.

¹⁶ Graphical representations of unconditional employment rates for mothers with their youngest children in respective age groups are presented in Figure C1 in the Appendix. On top of visualizing the effect of the reform on maternal employment rates, these graphs also document the existence of parallel trends in the evolution of employment rates between the treatment and control groups prior to the reform. Appendix Table D6 presents simple before-after estimates of the effects of the reform on employment rates of mothers at different distances from childbirth.

¹⁷ It might appear striking that employment effects of the reform on mothers with 3-years-old children do not differ across education levels, while the before-after analysis identified significant heterogeneity in parental leave length adjustments in response to the reform. This discrepancy might be explained by the fact that lower-educated mothers’ response to the reform was indeed stronger, but this group of mothers often entered unemployment right after terminating the parental leave (Bicakova and Kaliskova 2019) and this is why their response in terms of employment rate is similar to other education groups.

¹⁸ The period of job protection was 3 years since childbirth both before and after the reform. Motivating mothers to enter the labour market around their child’s third birthday might have increased the share of mothers taking advantage of job protection. This would be another reason why employment rates of mothers with 4-years-old children increased in response to the reform.

¹⁹ The analysis presented in this paper basically replicates the results of Bicakova and Kaliskova (2016) who estimate the effects of the 2008 reform (and of another earlier reform) on non-employment, inactivity and unemployment rates of mothers with 2- to 6-years-old children. Their results might slightly differ from the ones presented in this paper due to different time span of the data used and different definitions of control groups. Nevertheless, the overall picture presented in their and in my study is the same pointing towards strong employment effects of the reform for mothers of 3- and 4-years-old children and fading out of these effects with increasing child age.

²⁰ Lower precision of estimates using the average occupation wage as the dependent variable may be driven by the fact that in some occupations wages are more compressed, while in others wages are highly dispersed and their distribution is possibly skewed, what limits the representativeness of the average wage.

²¹ Increased representation of the employed mothers in high-skill occupations might be also a consequence of better positions held before childbirth. When the expected time spent out of employment after childbirth is shortened, more future mothers might sort to skill-demanding occupations. However, for this effect to be observed in the data a longer observation period would be needed. Currently, available data only allows to test whether the reform has affected the occupational allocation of young childless women. A difference-in-differences analysis using either young childless men or older women as control groups does not identify any change in the occupational allocation of young women in response to the reform (results available on request).

²² Observing more mothers of school-aged children in high-skill occupations might be also driven by increased pre-leave sorting to high-skill occupations.

²³ Example: When analyzing the effects of the reform on mothers of 3-years-olds, the treatment group is composed of mothers of three-year-olds and control group is composed of mothers of 6–7-years-olds in any year. The dummy $afte{r}_t$ takes value 1 for observation years 2011 and 2012, and value 0 for observation years 2008 and earlier. Years 2009 and 2010 are omitted from the analysis because this is when 3-years-old children born in 2006 and 2007, who were partially affected by the reform, are observed.

²⁴ Children born in 2006, i.e. in the first birth cohort that was partially affected by the reform, were 10-years old in 2016. Mothers whose youngest child is 10 in 2016 are excluded from the control group in 2016..

Additional information

Funding

This work was supported by the Grant Agency of the Czech Republic [14-26574P].