Morgan is a new parent who takes a year of job-protected parental leave following the birth of their child and later returns to their pre-birth job. Imagine that Morgan receives six months of full pay from their employer and an additional six months at 50% pay from the government. If we are to estimate the impact of having children on labour market outcomes, a key question arises: should Morgan be classified as employed or non-employed during this leave, and how should their earnings be measured? Zero per cent of their usual level? Or would one of 50%, 75%, or 100% be more appropriate?
The answer to these questions will depend on the research objective. If the aim is to understand the contribution of productive work to the formal labour market, it may be appropriate to categorise Morgan as non-employed, with zero earnings during leave. Conversely, if the aim is to analyse labour market attachment and labour supply adjustments post-childbirth, Morgan should be counted as employed. This is also in line with International Labour Organization guidelines, which count individuals on temporary leave as employed if the leave is job protected.
The approach taken can significantly affect perceptions of the ‘child penalty’ – the degree to which maternal earnings fall behind paternal earnings due to childbirth. The study of the career costs of children has burgeoned over the past decade (see, among others, Adda et al. 2017, Kleven et al. 2019, Zweimüller et al. 2019). However, a recent review of the literature reveals that the approach adopted by the papers is often unclear or inconsistent regarding the measurement of labour market outcomes over spells of parental leave. In fact, most of the main studies do not provide sufficient details to verify how outcomes during parental leave are measured. Moreover, the institutional design of parental leave benefits complicates the consistent classification of outcomes during leave, as part of income replacement is often processed through payroll (thus assimilated to labour earnings), while part is paid directly by the social security system and assimilated to benefits.
In our new research using comprehensive Danish administrative data (Adams et al. 2024), we delve into the intricacies of how parental leave dynamics relate to the timing and spacing of births and subsequent labour market outcomes. Our findings reveal notable disparities in birth timing and spacing based on maternal education levels. These dynamics complicate the estimation of the child penalty and its effects on gender gaps in labour market outcomes.
In line with existing evidence (Tertilt et al. 2022), we find that more-educated mothers tend to have their first child later in life. On average, women who do not complete high school give birth to their first child at 25 years, while those with a university degree do so at 31 years. The spacing between subsequent births also varies significantly: over 80% of the most-educated mothers have at least one additional child within five years of their first birth, compared to just 60% of their less-educated counterparts. This variance in family planning directly impacts the likelihood of being on parental leave: on average, the highest-educated mothers spend 25% of the first six years after birth on parental leave, compared to 11% for the least-educated group. We find that Danish men only take very little parental leave over the same time window.
Our findings call for simple modifications to the main event-study specification for child penalties pioneered by Kleven et al. (2019). Systematic variation in age at first birth by skill level and discontinuous arrival rates of first birth around graduation require a fully interacted model with skill fixed effects and additional controls for years since graduation. We show that this simple modification reduces the estimated child penalty in earnings five years after first birth by 13% on average, but this reduction ranges from 24% among college graduates to being negligible among the least-educated group, relative to that implied by the standard specification. This reflects the fact that age-earnings profiles are relatively steep upon graduation for all groups except the least educated, and correlations between age, graduation, and time of first birth differ substantially between women and men.
In our next exercise, we reconsider individual outcomes during spells of parental leave, treating them as employed if working before going on leave and setting their earnings at pre-leave levels during spells of leave. This exercise reduces estimates of women’s child penalties in earnings five years after first birth by 45% on average, ranging from 50% among university graduates to 12% among the least-educated group, relative to that implied by the standard specification.
Our findings suggest that researchers should be careful with the measurement of labour market status during periods of parental leave. Institutional features will affect how employment and earnings are measured over a spell of parental leave, and whether parental leave is job protected. The approaches we advocate for in our paper, namely either to set earnings to zero or at their pre-leave level (which an individual is entitled to return to in countries with job-protected leave), permit a consistent treatment of periods of parental leave.
These adjustments also provide information on the source of estimated child penalties, that is, changes in labour supply behaviour when not on parental leave versus the impact of subsequent births. Our results are also important for interpreting changes in child penalties over time and across countries when returns to skill and parental leave policies vary. In other words, our results highlight that the institutional setting is vitally important to consider when estimating child penalties in labour market outcomes.
References
Adams, A, M Fjællegaard Jensen, and B Petrongolo (2024), “Birth timing and spacing: Implication for parental leave dynamics and child penalties”, CEPR Discussion Paper 19324.
Adda, J, C Dustmann, and K Stevens (2017), “The career costs of children”, Journal of Political Economy 125(2): 293–337.
Kleven, H, C Landais, and J Søgaard (2019), “Children and gender inequality: Evidence from Denmark”, American Economic Journal: Applied Economics 11(4): 181–209.
Tertilt, M, A Hannusch, F Kimdermann, and M Doepke (2022), “A new era in the economics of fertility”, VoxEU.org, 11 June.
Zweimüller, J, A Steinhauer, C Landais, J Posch, and H Kleven (2019), “Child penalties across countries: Evidence and explanations”, VoxEU.org, 14 May.
