She cares and he works — Does paid leave decrease the labour market gap? A Double- and Triple-Differences Analysis

(1)

(2)

The labour market impact of paid leave

2

Abstract

Paid (maternity, parental and paternity) leave enables parents to retain their pre-birth job and income while staying home to take care of their new-born children. Allowing parents to combine caring tasks with a continuation of their careers motivates, in part, the legislation of

paid leave. This thesis investigates the aggregate-level impact of paid (maternity and parental)

leave duration on several labour market outcomes of women in 37 OECD countries between 1970 and 2018. Prior to doing so, the thesis embeds the analysis in a rich review of micro-studies and macro-micro-studies that have analysed this relationship.

The micro-studies evidence has illustrated that paid maternity and parental leave generally leads to delayed return-to-work rates, while they have a positive impact on female employment rates once the benefit period has ended. Yet, when (maternity and parental) leave periods are too long this can adversely impact women’s labour force participation rates. There is no consensus in the micro- and macro-studies evidence what constitutes the tipping point — at what length of paid maternity and parental leave the effects on labour market participation rates become negative. Nonetheless, not all of the micro-studies evidence is significant and the significant positive effects generally dissipate in the long-run. Although only six authors have investigated the relationship, they find evidence for a positive relationship when looking at weekly working hours. The effect on wages is more ambiguous; most support seems to exist for the human capital depreciation argument — that leave decreases female wages due to time out of work. The effect on wages likewise dissipates in the long-term.

Despite a large reduction in the gender gap in female labour market outcomes, a large gap persists regarding who performs the tasks of caring for (new-born) children. The larger share of housework that is typically done by women contributes to persisting gender gaps in labour market outcomes (Hook, 2010). Some authors argue that if paid leave is predominantly focused on mothers, who consequently interrupt their careers, this contributes to the upholding of the gendered specialization of labour where women spend relatively more time at home than at work in order to perform caring tasks. This is underlined by the findings that long periods of

paid maternity and parental leave can adversely impact short-run female labour force outcomes. However, the literature illustrates that these negative impacts for long leave durations do not last in the long-term.

(3)

Eva-Mirre Dam

3 In light of the persisting gender gap in household tasks, the male contribution to caring tasks as a way of closing these gaps has received growing attention. In order to promote active fathering and to reduce sex specialization, many countries have introduced paternity leave and so-called daddy quotas, which reserve some parental leave for fathers (Patnaik, 2019). Micro-studies literature that has investigated this relationship affirms that paternity leave and daddy quotas positively impact female labour force participation rates and wages, whereas the results are insignificant for men. These effects are limited to the introductions of paternity leave and daddy quotas and typically dissipate in the long-run. Nevertheless, the research on paternity leave is still quite scarce and more evidence is needed to confirm these effects.

The literature review illustrates that not all scholars find uniform labour market effects of paid leave. The effect can be influenced by moderating factors, such as other policies (i.e. availability of childcare services), cultural norms and attitudes and/or labour market shocks. Furthermore, the effect may be heterogeneous for different groups of women — i.e. women from different educational backgrounds, income groups and/or whether a woman is working full-/part-time. These facets complicate the external validity of the micro-studies, since the context in which the country is situated can play an important role in explaining the effect that is found. Furthermore, it encumbers the interval validity of cross-country research if the effect of paid leave cannot be isolated from such facets.

Taking the findings of the literature review into consideration, a macro-level empirical analysis is conducted to investigate the relationship between paid parental and maternity leave and female labour market outcomes. The macro-level analysis also looks at differences between within-country groups and at heterogeneities between groups of women and country clusters. To do so, the thesis employs a double-differences model, regressed upon absolute numbers and first-differences, which identifies the treatment effect by studying within-country time variation. Furthermore, the thesis utilises a triple-differences model that studies the differences between comparison groups within countries over time; comparing men and women in particular. Next to updating previous findings the analysis discusses the strengths and weaknesses that are related to each of these models.

The results of the analysis show significant increases in female participation rates and a diminishing gender gap in labour force participation rates. Lower-bound estimate of the double-differences method, regressed upon first-differences, suggests that the introduction or extension 54 weeks of paid leave (the cross-country mean) leads to a 0.6 percentage-point significant increase in female labour force participation rates. Conversely, the higher-bound

(4)

4 estimate posits a significant effect of 2 percentage-point in response to 54 weeks of paid leave. There is no strong evidence for a quadratic paid leave effect, nor for heterogeneous effects for different groups of women or countries. The results also suggest that paid leave does not impact working hours among employed women or the share of employed women that are managers.

The findings in this thesis are helpful for policymakers who are interested in paid leave policies and/or in lowering the labour force participation rate gap between men and women. However, the inconclusive literature evidence, data imperfections and methodological limitations make it difficult to draw inferences regarding the size of the effect. These limitations create avenues for possible future cross-country research. Some examples are: looking more closely at the effects of different lengths of leave and on their labour market outcomes. Furthermore, more micro- and macro-studies research should be conducted regarding the effects of paternity leave.

(5)

Eva-Mirre Dam 5

1. Introduction

Women’s labour force participation rates, earnings and working hours are still behind those of men (United Nations, 2020) and more and more policies are being implemented globally that focus on lowering this gap. Paid leave regulations play a central role in family policies in many OECD countries. They guarantee parents their pre-birth job and offer financial support. Besides fostering the wellbeing and health of mothers and their new-born child, paid leave enables parents to combine caring tasks with a continuation of their careers. By encouraging job continuity after birth, paid leave may help to increase female labour market outcomes, such as their employment rates, earnings and working hours. Recent introductions and/or extensions of paid leave statutes are often installed with the motivation that paid leave legislations help to foster gender equality. However, prolonged periods of career interruptions might also lead to a loss of human capital and deteriorating labour market prospects of women.

Many countries have paid leave arrangements, but their actual systems differ considerably. The impact of these different systems is not well understood. There are multiple elements to a paid leave regime; isolating the effects of these different elements, i.e. their benefit levels, durations or flexibility, is complicated. The aim of this thesis is to shed light on the macro-level effect of the duration of paid leave on several labour market outcomes; female labour force participation rates in particular. The research question that is investigated throughout this thesis constitutes: “How do entitlements to paid leave affect female labour

market outcomes (labour force participation rates, weekly working hours and vertical occupational segregation)?”

The thesis draws lessons from quasi-experimental micro-studies as well as cross-country studies to sketch a rich framework that motivates the empirical cross-cross-country analysis. Many studies find evidence of a positive effect of paid leave on female labour force participation rates that turn negative for longer leave durations and dissipate in the long-run. Yet, the evidence is inconclusive — no evident consensus has arisen in the literature concerning the optimal leave duration and many authors fail to find any significant effects.

In the empirical analysis, the thesis builds upon the findings and methodologies of the literature framework when employing a macro-level double- and triple differences model. This analysis focuses on the impact of paid maternity and parental leave entitlements, which are granted to mothers, on female labour market outcomes. With the use of panel data,

(8)

8 predominantly provided by the OECD, the double-differences model looks at within-country intertemporal changes in paid leave policies and female labour market outcomes for the 37 different OECD countries in the period 1970-2018. The model controls for country and year fixed effects and country-specific time trends to remove unobserved time-invariant confounders. This difference-in-difference model is regressed upon absolute numbers and first-differences, which takes the within-country differences between two subsequent years. The triple-differences framework adds another level to the double-differences model by taking intertemporal differences for the dependent variable between within-country comparison groups in the various countries. The within-country difference between (treated) women and (untreated) men constitutes the main comparison group for the latter model. These frameworks build upon Ruhm’s (1998) prominent study that also investigated this relationship using cross-country data. Ruhm points to a positive relationship between paid leave and female employment rates, while longer leaves (from nine months onwards) negatively affect women’s long-term wages.

The triple- and differences models both find significant evidence that paid leave positively impacts female labour force participation rates, which is in line with previous micro- and macro-studies that have investigated this relationship. In contrast to the literature framework, the analysis does not find significant negative impacts for long paid leave durations (a quadratic negative effect). Additionally, no significant impacts are found for other labour outcome variables, nor for heterogeneity in results across groups of women.

The academic relevance of this thesis is fourfold. Firstly, previous cross-country analyses are updated with more recent data. Secondly, the analysis looks at heterogeneities between country groups and groups of women within countries and adds control variables that have arisen in the micro-level studies. Thirdly, other outcome variables, besides labour force participation rates, are explored. Fourthly, and most importantly, this thesis identifies the different methodological advantages and drawbacks that are related to the different macro-level methodological frameworks. By combining the results of these frameworks, a convincing case can be build that paid leave exerts a significantly positive impact on female labour force participation rates.

This thesis starts by situating the research topic in a theoretical background. This is followed by a rich review of previous micro-studies and cross-country evidence that have investigated the impact of paid maternity, paternity and parental leave on several labour market outcomes. Subsequently, the methodological framework and data, which form the foundation

(9)

Eva-Mirre Dam

9 for the empirical analysis, are discussed. The empirical analysis starts with the double-differences results, which are regressed upon absolute numbers and first-double-differences. The second part of the empirical analysis concerns the triple-differences estimations. Finally, the results are synthesised. As there are several caveats when methodologically investigating this relationship, the different double- and triple-differences estimates are presented together to provide a convincing answer to the research question.

(10)

10

2. Theoretical Background

2.1. Defining leave

Leave policies provide new parents with time off from work to care for a new-born child, often while protecting their job and replacing some extent of their income. Leave policy statutes vary across countries, but a general distinction can be made between the following four types of leave (Blum et al., 2018):

Maternity leave constitutes a job-protected type of leave that is reserved for mothers

and is paired with payments in almost all OECD countries. It is typically provided shortly before, during and after pregnancy to guarantee the health of the mother and child. In some countries, a short period of pre-birth leave is compulsory. On average, mothers are entitled to 18 weeks of paid maternity leave on average across OECD countries, with 77% of the OECD countries offering payments that replace 100% of previous earnings (OECD, 2016).

Paternity leave denotes a job-protected type of leave that is solely reserved for the father

or a partner shortly after the birth of a child and is most often paid. This type of leave is typically shorter than maternity leave; OECD countries offer 1.4 weeks of paternity leave on average. 69% of OECD countries offer paternity leave payments that replace 100% of previous earnings (OECD, 2016).

Parental leave is a job-protected type of leave that is complementary to maternity and

paternity leave and is paid in some countries1_{. In a number of countries certain periods of} parental leave are reserved for one parent and cannot be transferred. In some countries parents are obliged to both take up a share of the parental leave period. On average, parental leave periods have a duration of 36 weeks. 48% of the OECD countries offer payments that replace 100% of previous earnings and most of the remaining countries replace between one-third to two-thirds of previous earnings (OECD, 2016).

Leave to care for children who are ill, which differs widely between countries with

respect to age of the child, payments, length and whether it includes caring for an adult who has fallen ill. Due to its heterogeneous nature, this type of leave is not included in this thesis.

Besides distinguishing between these four types of leave, there are numerous other ways in which leave constellations differ across countries. The biggest differences in leave

1_{This list concerns statutory paid leave statutes that are determined by law. In several sectors in OECD countries parental}

(11)

Eva-Mirre Dam

11 Table 1 — Duration and average payment rates of paid maternity leave, paid paternity leave and paid

parental leave in 2018

Data is extracted from: OECD (2018)

2_{On January 1}st_{, 2020 paid paternity leave in the Netherlands has been expanded to 1 week and on July 1}st_{, 2020 this will be expanded to 6}

weeks (with benefits equalling 70% of pre-birth wage levels during week 2-6) (Rijksoverheid, n.d.)

Paid maternity leave Paid parental and home care leave available to mothers

Paid paternity leave Paid parental and home care leave reserved for fathers

Length (weeks) payment rate Average

(%) Length (weeks) Average payment rate (%) Length (weeks) Average payment rate (%) Length (weeks) Average payment rate (%) Australia 18,0 42,9 0,0 0,0 2,0 42,9 0,0 0,0 Austria 16,0 100,0 44,0 75,8 0,0 0,0 8,7 75,8 Belgium 15,0 63,7 17,3 20,3 2,0 73,0 17,3 20,3 Canada 16,0 49,9 35,0 53,2 0,0 0,0 0,0 0,0 Chile 18,0 100,0 12,0 100,0 1,0 100,0 0,0 0,0 Czech Republic 28,0 61,0 35,3 84,5 1,0 61,0 0,0 0,0 Denmark 18,0 53,0 32,0 53,0 2,0 53,0 0,0 0,0 Estonia 20,0 100,0 146,0 44,1 2,0 100,0 0,0 0,0 Finland 17,5 74,4 143,5 19,1 3,0 62,9 6,0 62,9 France 16,0 90,4 26,0 13,7 2,0 90,4 26,0 13,7 Germany 14,0 100,0 44,0 65,0 0,0 0,0 8,7 65,0 Greece 43,0 49,5 0,0 0,0 0,4 100,0 0,0 0,0 Hungary 24,0 70,0 136,0 37,8 1,0 100,0 0,0 0,0 Iceland 13,0 68,2 13,0 68,2 0,0 0,0 13,0 68,2 Ireland 26,0 26,7 0,0 0,0 2,0 26,7 0,0 0,0 Israel 15,0 100,0 0,0 0,0 0,0 0,0 0,0 0,0 Italy 21,7 80,0 26,0 30,0 0,8 100,0 0,0 0,0 Japan 14,0 67,0 44,0 59,9 0,0 0,0 52,0 58,4 Korea 12,9 80,2 52,0 28,5 0,6 100,0 52,0 28,5 Latvia 16,0 80,0 78,0 49,8 1,4 80,0 0,0 0,0 Lithuania 18,0 100,0 44,0 100,0 4,0 100,0 0,0 0,0 Luxembourg 20,0 100,0 17,3 67,2 2,0 100,0 17,3 67,2 Mexico 12,0 100,0 0,0 0,0 1,0 100,0 0,0 0,0 Netherlands 16,0 100,0 0,0 0,0 0,42 _100,0 _0,0 _0,0 New Zealand 18,0 46,8 0,0 0,0 0,0 0,0 0,0 0,0 Norway 13,0 94,2 78,0 39,4 0,0 0,0 10,0 94,2 Poland 20,0 100,0 32,0 67,5 2,0 100,0 0,0 0,0 Portugal 6,0 100,0 24,1 59,6 5,0 100,0 17,3 43,6 Slovak Republic 34,0 75,0 130,0 21,2 0,0 0,0 0,0 0,0 Slovenia 15,0 100,0 37,1 90,0 4,3 90,0 0,0 0,0 Spain 16,0 100,0 0,0 0,0 4,3 100,0 0,0 0,0 Sweden 12,9 77,6 42,9 57,4 1,4 58,4 12,9 77,6 Switzerland 14,0 58,4 0,0 0,0 0,0 0,0 0,0 0,0 Turkey 16,0 66,7 0,0 0,0 1,0 100,0 0,0 0,0 United Kingdom 39,0 30,1 0,0 0,0 2,0 19,2 0,0 0,0 United States 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 OECD average 18,1 - 35,8 - 1,4 - 6,7 -

(12)

12 regulations regard the length and the height of the payments (table 1). Some patterns can be identified from this table. In general, Eastern European (i.e. Czech Republic, Hungary, Poland, Slovakia) and Northern European countries (Finland, Sweden, Norway, Denmark) provide the longest leave entitlement coupled with the highest payments. Most Anglo-Saxon countries (Australia, Ireland, New Zealand, United Kingdom, United States) provide the least generous leave entitlements. Continental European countries (Austria, Belgium, France, Germany, Luxembourg, Netherlands), Southern European (Greece, Italy, Portugal, Spain) and Asian countries (Japan, South Korea) provide leave entitlements that are close to the OECD-average (Thévenon, 2011).

The leave constellations can also differ with respect to flexibility — leave can for instance be provided as a non-transferable individual right as well as a transferable entitlement. The former is often referred to as a daddy quota, a part of the leave period that is solely reserved for fathers. Besides, in some countries leave can be flexibly taken-up in combination with part-time work. Eligibility for paid leave may also differ per country (i.e. how long an individual is required to be employed, or whether there is an income cap for leave benefits). Lastly, countries construct the financing of leave in different ways — i.e. whether the employers contribute or whether the benefits are collected via general taxation (Dearing, 2015).

Employer-provided leave payments are not included in table 1 and figure 1. Practices differ across companies and countries, but discretionary employer-provided benefits can be substantial. The thesis only considers statutory leave entitlements because information on collective agreements and employer initiatives is lacking. Consequently, the indicators that this thesis refers to will in some countries underestimate the actual amount that parents receive. Although the thesis at some points refers back to these distinctions in the literature review, the thesis uses the umbrella term ‘leave’ that subsumes the different strands of leave. The empirical analysis focuses solely on paid maternity and parental leave, while the literature review also discusses paternity leave.

2.2. Justifications for paid leave policies

Most developed countries have installed leave entitlements, although the path to adoption has varied considerably across countries (figure 1). Early adoptions of leave legislation mostly focused on protecting the health of mothers and children; family policy designs were often suited to protect women in their role as wives and mothers. The surge in female labour market participation in most OECD countries over the past few decades ignited higher demand for

(13)

Eva-Mirre Dam

13 Figure 1 — Evolution of paid leave across OECD countries

Panel A — Length of paid maternity, parental and home care leave available to mothers in 1970, 1990 and 2018

Panel B — Length of paid paternity, parental and home care leave available to fathers in 1970, 1990, 20003 and

2018

Data is extracted from: OECD (2018)

leave provisions. In this light, leave entitlements were regarded as a way to assist mothers in managing employment obligations when young children are present. Subsequently, many high-income countries also introduced job protection and high-income support during career interruptions. Recently, OECD countries were incentivised to introduce and widen parental and paternity leave legislation in order to enable parents to more equally share formal labour and household responsibilities (Olivetti & Petrongolo, 2017).

3_{2000 is included in panel B (and not included in panel A) because most changes in paternity leave are more recent than}

changes in maternity and parental leave available to mothers.

0 20 40 60 80 100 120 140 160 180 Weeks 1970 1990 2018 0 5 10 15 20 25 30 35 40 45 50 55 60 Weeks 1970 1990 2000 2018

(14)

14 The evolution of leave legislations illustrates that there are multiple justifications for introducing or extending paid leave, which are discussed below. It is important to note that paid leave legislation reduces overall economic efficiency in a static supply-and-demand framework, with perfect information and an absence of externalities, as it limits the ability of employers and employees to reach an optimal voluntary leave agreement. An economic rationale for paid leave policies in a market economy can be based upon the argument that there are market failures if paid leave is left to the market (Baum, 2003). This rationale is also revisited in the following list of justifications that are often linked to paid leave legislation:

Potential externalities: Health, well-being, development and fertility — A prominent

argument for paid leave constitutes that paid leave mandates can improve the health of the mother and child (Staehelin et al., 2007; Rossin, 2011; Stearns, 2015) and even of the siblings (Beuchert et al., 2016). Paid leave can improve maternal mental health (Avendano et al., 2015) and increase breastfeeding rates (Huang & Yang, 2015).

Another motivation is grounded in the belief that the first months in a child’s life are crucial for future cognitive and emotional development (Dustmann & Schönberg, 2012; Huebener et al., 2019; Ruhm & Waldfogel, 2012). Some studies for instance suggest that it can affect children’s schooling outcomes in the long run (Rasmussen, 2010; Danzer & Lavy, 2013; Carneiro et al., 2015; Cools et al., 2015; Ginja et al., 2020).

Furthermore, parental leave can improve the opportunities for parents to care for new-born children, which may influence fertility decisions (Averett & Whittington, 2001; Lalive & Zweimüller, 2009; Luci-Greulich & Thévenon, 2013; Ang, 2015; Cygan-Rehm, 2016; Raute, 2019).

The health and wellbeing of children and working parents and the development of children can be perceived as externalities that are not fully accounted for in the wage that is determined on the market. The presence of such externalities can be an argument in favour of leave legislation. In the presence of imperfect information, if employers and/or employees do not aptly internalise such aspects when negotiating a contract, a suboptimal paid leave package may be selected. Furthermore, behavioural economists argue that people have higher than socially optimal discount rates. When negotiating a contract, this may hinder (prospective) parents from foreseeing the true value of paid leave at later points in their lives (Ruhm, 1998).

Adverse selection: Moreover, if paid leave is left to the market, asymmetric information

and adverse selection may cause companies with generous leave constructions to attract employees that have a ‘high risk’ of taking-up leave. This can result in a race to the bottom,

(15)

Eva-Mirre Dam

15 with no company wishing to provide a generous leave package (Ruhm, 1998). It is important to note that if parts of the costs of the paid leave legislation are shifted onto employers — i.e. if employers are directly responsible for contributing to the benefits, or if they face the costs of temporary replacing and training workers — this effect may persist. Due to asymmetric information and the risk of adverse selection, employers may then engage in statistical discrimination against women of childbearing age as a group (Kunze, 2016).

Search frictions — There may also be search frictions on the labour market associated

with employees facing difficulties to find a job that provides the same utility as the pre-birth job, or with employers that struggle with replacing an employee. In the presence of search frictions there may be a surplus that the firm and the worker share when they are matched; these jobs leave both the firm and the worker better off in the job than would their outside options. The job surplus is higher if the outside options for both parties (i.e. unemployment, working for another firm, hiring another worker) are less valuable. Such a surplus can only exist in labour markets with frictions, because otherwise the worker would be paid his full marginal product and the outside option would be equally valuable. In case of search frictions and employment surplus, job tenure is valuable (Acemoglu & Autor, 2011).

Search frictions and match surpluses can exist in the presence of firm-, occupation- or task-specific human capital or because of match-specific search capital (Schönberg & Ludsteck, 2014). Paid leave enables a continuous labour market attachment and retainment of work-specific human capital; thereby it prevents employers and employees from having to search for and/or train a match if parents temporarily interrupt their careers. Legislating paid leave can therefore reduce job losses and thereby lowers the negative externality of parents falling into social security.

Gender equality: Promoting gender equality is another rationale for legislating paid leave,

as they can support mothers’ paid work rights and enable mothers to enter higher employment positions (Korpi et al., 2013). Section 2.4. goes into more detail regarding the ways in which paid leave can either promote or hinder economic gender equality, by drawing upon Becker’s (1985) theory of specialisation within the household.

2.3. Mechanisms driving potential labour market outcomes

The labour market effects of paid leave can be illustrated in a supply-and-demand framework (figure 2). The presence of a new-born child may decrease the worker surplus and lead to job

(16)

16 separations, if it makes the outside option of caregiving more valuable. When it offsets these effects, paid leave may cause the labour supply curve to shift to the right (from S1 to S2a) as it enables a continuing labour market attachment of parents who would exit the labour market if leave would not be provided. Additionally, parents’ labour supply is increasing with their valuation of paid leave; the availability of paid leave (conditional on working) makes being attached to the labour market more attractive for prospective parents (Akgundez & Plantenga, 2013).

On the other hand, the labour supply curve is likely shift to the left (from S1 to S2b) in the short-run, due to a rise in reservation wages. In this framework, parents can ‘buy’ an hour of time to care for their new-born child for the price of their hourly wage rate minus the hourly price of childcare purchased outside the home. Paid leave changes a parent’s budget constraint as it guarantees a higher temporary non-working income while maintaining job tenure (Cascio et al., 2015). Hence, paid leave improves the bargaining power of the parent by making the outside option of temporarily interrupting one’s career more attractive. Paid leave may therefore incentivise people to interrupt their careers and/or may lengthen the time that parents stay away from work in the short-run. Moreover, the chance to return to one’s previous career in the long-run diminishes if the leave period is too long, as the work ties are loosened and existing knowledge and skills deteriorate (Kluve & Schmitz, 2018). Parents may also ‘learn’ about the enjoyability of caring for a child during the leave period, diminishing their likely return to work (Schönberg & Ludsteck, 2014).

The demand for labour may shift to the right (D1 to D2a) due to the retainment of human capital. Paid leave may facilitate swifter returns to work, as it prevents parents from exiting the labour market. A positive effect on labour demand is expected if a continuous labour market attachment or prior experience is highly valuable (Olivetti & Petrongolo, 2017). Job tenure is seen as valuable by employers in case of employer surplus that is determined by the human capital endowments of the employee (Kunze, 2016). High investments borne by the firm in the worker’s (both general and firm-, occupation, or task-specific) human capital adds to the employee surplus and thereby to the value of job tenure for the employee. Furthermore, some workers are a better fit to some companies than others, and the other way around, which also adds to the job surplus (Acemoglu & Autor, 2011). In case of such employer surplus, paid leave may cause the demand to rise. This can be attributed to the mechanism that paid leave raises job tenure, which makes it more valuable for employees to hire and invest in individuals with a high chance of bearing children.

(17)

Eva-Mirre Dam

17 Figure 2 — Economic Consequences of Leave

D1, D2a & b represent the labour demand curves and S1, S2a & b represent the labour supply curves. In the short-run the labour supply may shift from S1 to S2b, since paid leave induces parents to interrupt their careers (for longer times). The demand curve can simultaneously shift from D1 to D2b, if employers face costs of temporarily replacing workers or of if they are held responsible for providing leave benefits. The human capital depreciation, due to career interruptions, can also hit employers. The labour supply can also shift from S1 to S2a if paid leave is valued by workers. The job protection aspect of paid leave also enables swifter returns to the labour market. The demand curve can also shift from D1 to D2a as it enables a longer labour market attachment of workers and thereby a retainment of company-specific human capital (Akgundez & Plantenga, 2013).

On the other hand, when paid leave confronts employers with costs this lowers the employer surplus of being matched with the worker. The demand curve is then expected to shift to the left (from D1 to D2b). The extent in which the demand shifts leftwards depends on whether employers, employment sectors or the government is responsible for the leave payments. Insofar as employers have to contribute to the paid leave benefits the demand for labour is expected to be more negatively impacted. Employers may also face indirect costs, i.e. temporary replacement and training costs and a decline of leave-takers’ human capital due to the time out of work (Akgundez & Plantenga, 2013). To prevent the risk of adverse selection, employers may even engage in statistical discrimination against women if they expect them to benefit relatively more from those policies (Kunze, 2016). Besides, (multiple) career interruptions and time-out-of-work of mothers may ‘signal’ to employers that parents on leave are less career-oriented, which may damage their career opportunities (Olivetti & Petrongolo, 2017).

(18)

18 This section demonstrates that the theoretical effects of paid leave are ambiguous, pointing to the importance of empirically investigating this relationship. The labour demand and supply curves can move up- as well as downwards. The extent to which these curves are adjusted, and in what way, determines the effects of paid leave. Section 2.5 integrates these facets in the hypotheses.

2.4. Becker’s theory of intra-household specialisation

2.4.1. Supply of labour within the household

In order to shed more light on how paid leave may impact gender equality outcomes, this section shifts the focus from aggregate supply of labour to the division of labour within the household. When zooming in upon the impact of paid leave on the labour supply within the household many scholars refer to Becker's theory on the economics of marriage. In his theoretical work, Becker (1985) explains why household members allocate their time and effort spent on either market labour or household chores. In line with the supply-and-demand framework discussed above, Becker highlights the importance of human capital in explaining the supply of (household and market) labour.

According to Becker, household members are encouraged to allocate time to activities where they have a comparative advantage in doing so. An intra-household allocation can be regarded as a trade-off between tasks at home and in the labour market. The trade-off depends on one’s relative productivity in each area, which in turn depends upon one’s human capital. Individuals’ human capital specific to a particular activity is positively related to the time and effort spent at that activity. Spouses within a household exploit such human capital endowments by specializing in the specific tasks they are most productive in (their comparative advantage); consequently reinforcing their respective human-capital investments. This entails that once spouses specialise in either housework or market labour, the gains from division of labour and trade are reinforced as the spouses will built up more skills in these fields (Baker & Jacobsen, 2007).

Paid maternity leave is often regarded by the government as a way of protecting and securing women’s rights, as it enables mothers to stay attached to the labour market during temporary employment interruptions and to retain firm-specific or occupation-specific human capital. The retainment of firm-specific human capital during the leave period is beneficial to the employee who is prevented from having to train and replace workers, as well as to the

(19)

Eva-Mirre Dam

19 parent who will have better opportunities to continue their careers (Akgundez & Plantenga, 2013).

Nonetheless, paid maternity leave may also impede women’s labour market achievements by inducing women to stay out of work for long periods preserving women’s responsibilities for housework. By highlighting the reinforcing role of human capital investments on comparative advantages, Becker describes that small changes in the short-run can be pivotal for explaining time allocation decisions and career developments in the long-run (Becker, 1985). Prolonged periods of absence from the workplace may therefore have longer-term implications for the gender division of labour beyond the leave period due to depreciating labour market skills and improved domestic work skills (Mandel & Semyonov, 2006; Olivetti & Petrongolo, 2017).

2.4.2. Welfare state paradox

In light of the theoretical framework sketched above, leave policies may create a welfare state paradox. On the one hand it increases the labour market attachment of mothers, but it may simultaneously limit their job opportunities and wages (Mandel & Semyonov, 2006). This can be linked to the literature on the insider/outsider problem, which entails that collective negotiations and policies perpetuate the division between those with good versus bad employment opportunities. If paid leave is predominantly focused on mothers interrupting their careers, this contributes to the upholding of the male breadwinner model in which women spend relatively more time at home to perform household chores (Rovny, 2011). This is the case for maternity leave, but typically applies to parental leave too — as most of the parental leave period that can be shared between spouses is often taken-up by mothers (Patnaik, 2019).

Empirical evidence on women’s specialisation within the household underlines this paradox. Some research finds that the increase in female labour market participation rates has induced women to spend less time on household tasks — either resulting from a more equal distribution of household tasks between partners (Grossbard-Shechtman, 2005), or because housework has reduced overall since it could be more easily consumed on the market (Bianchi et al., 2000). Other research, on the other hand, finds that the relative comparative advantages between partners has not dramatically changed, as women often remain responsible for a large fraction of child care and other housework (Blair & Lichter, 1991; Hochschild, 1997; Bianchi et al., 2000; Bianchi, 2011; Patnaik, 2019).

(20)

20 2.4.3. Fostering father’s take-up rates

More recently, paternity leave has been introduced and extended in several countries (Patnaik, 2019). If paternity leave causes men and women to make more equal investments in housework-specific human capital in the first few weeks/months, it might reduce sex specialisation in the long-run. Besides the idea that an increase in fathers who take up leave may benefit child development and improves the connection between father and child, paternity leave may therefore also promote gender equality. Father’s take-up of leave may strengthen women’s labour market positions as it enables mothers to return to work earlier and may diminish statistical discrimination against women in the labour market. Furthermore, paternity leave can equalise the initial division of household tasks, which can have long-term specialisation consequences (Patnaik, 2019), consistent with the theory of Becker (1985).

2.5. Hypotheses

The economic mechanisms described in this section show that the theorised effects of paid leave on labour market outcomes can be positive as well as negative. Increasing the length of maternity leave has a direct beneficial effect on the utility of parents as it lowers their opportunity costs of caring for a new born child. Parents are therefore incentivised to take-up leave and interrupt their careers. After the leave period has ended, however, paid leave mandates are expected to positively influence labour market outcomes as it enables parents to stay attached to the labour market during temporary interruptions of employment and to retain work-specific human capital.

However, as the length of paid leave increases it raises the costs imposed on the employer — either via direct payments of benefits, temporarily replacing workers and a loss of human capital of the parent. Furthermore, if mothers take up most of the leave this can reinforce the gendered division of household and market labour tasks. This, in turn, harms the parent on leave by reducing his/her wage and employment opportunities. One can assume that the marginal benefit of paid leave is positive and decreasing for parents, while the employer’s marginal cost of paid leave is positive and increasing. Therefore it is likely that the positive labour supply effect dominates at short periods of leave, while the negative labour demand effect dominates for long leave durations. At the tipping point, the marginal gain and marginal loss of longer paid leave are equal (Givati & Troiano, 2012).

The theoretical framework can be summarised in the hypotheses listed below. Nonetheless, as mentioned before, some ambiguity remains in hypothesizing how the effects

(21)

Eva-Mirre Dam

21 of paid leave unfold. This points to the importance of empirically investigating this relationship. Section 3 explores these hypotheses when reviewing previous micro- and macro-studies evidence. Subsequently, section 4 empirically investigates hypotheses 2 and 345 within a double- and triple-differences framework.

Figure 3 — List of hypotheses derived from the theoretical framework

4_{Hypothesis 1 is difficult to investigate empirically, because women that are on leave are often counted as ‘employed’. This}

hypothesis is therefore explored in the literature review, but is not tested in the empirical analysis.

5_{Hypothesis 3 is an empirical question, for which the theoretical effect is ambiguous, and which is therefore investigated in}

the literature review and empirical analysis. During the benefit period:

Hypothesis 1 Declining employment during the benefit period4 — taking-up leave is more attractive for parents due to a rise in their reservation wages

In the post-benefit period:

Hypothesis 2 The effect of paid leave on women’s labour outcomes follows an inverted u-shape curve: 2a Paid leave positively impacts female labour force participation rates — due to a higher labour supply, swifter returns and retainment of work-specific human capital;

2b The positive effect on female labour force participation rates diminishes for longer leave durations — due to a likely fall in labour demand, human capital depreciation and specialisation within the household

Hypothesis 3 — The effect of paid leave varies for different labour market outcomes5 (i.e. the effects on working hours and vertical segregation) — The effect may be positive: labour market attachment and retainment of work-specific human capital, or negative: due to a human capital depreciation, specialisation within the household and a negative signalling effect

(22)

22

3. Literature Review

There is a large body of literature analysing the empirical impact of paid leave policies on labour force outcomes; within-country as well as cross-country studies. Studying this relationship is not straightforward. Since one can never observe the counterfactual situation of what the world would have looked like without the paid leave policy changes, it is complicated to empirically decipher whether labour market outcomes are the direct consequences of paid leave regulations. The cause-and-effect relationship can be overstated or understated in the case of endogeneity or confounding variables — if the causal labour market effect is entirely attributed to paid leave while there are other unobserved characteristics that influence both of these variables. Furthermore, since paid leave regulations are studied in their natural environment, there is no random selection of the sample of interest (individuals eligible for paid leave). Without random selection, it is more difficult to isolate the causal effects of a certain policy regulation from those of other characteristics (Buligescu et al., 2009). Micro-studies and cross-country Micro-studies use different methods to deal with these empirical challenges. This section reviews both strands of literature, which forms the foundation for the cross-country analysis that follows.

3.1. Micro-Studies

3.1.1. Characteristics

Micro-studies aim to solve the aforementioned empirical issues by studying the impact of specific policies by using the exogenous variation from natural experiments in combination with rich microdata. Typically, a policy change is studied — in this case the introduction of paid leave or the extension of such existing programs — that naturally creates a treatment and control group of individuals who are highly similar in their characteristics, apart from paid leave eligibility. These studies take advantage variation across time (before and after a cut-off), in age groups, in regions, and of sexes to study the effect of a policy change that affects one group and leaves the other (comparable) group unaffected. If the treatment and control groups are sufficiently similar, apart from whether or not they are affected by a paid leave policy change, the comparison gives a consistent estimate of the direct effect of paid leave on labour market outcomes (Kunze, 2016).

(23)

Eva-Mirre Dam

23 Tables 2-4 summarise the micro-studies evidence that is discussed in this section. Most of the studies use difference-in-difference or regression discontinuity methods. Difference-in-difference methods use a paid leave policy change that affects a share of the observations in the sample, but not all of them, or at least not all at the same time. First, one takes the difference across time within the treatment (with a change at a certain point in time) and control group (no change). Afterwards, one takes the difference of these two differences. When the treated and untreated group follow the same time trend in absence of the policy change, this approach helps to eliminate a time trend that confounds the relationship between the policy change and the effect. On the other hand, the regression discontinuity approach compares treated and untreated individuals who just fall below and just above the threshold for paid leave eligibility. These groups are fairly comparable as they are close to the threshold (Angrist & Pischke, 2008). When interpreting micro-studies’ evidence, it is important to be aware of the fact that paid leave policies do not exist in isolation. When studying the effects of paid leave, the effect is partly determined by other institutional characteristics such as the availability of childcare services. Alongside policy changes related to paid leave, such institutional characteristics may also change during the period studied. Furthermore, there may be spillover effects that lead to a biased estimate of the policy impact. This may be the case if the behaviour of treated individuals influences individuals in the control group. Finally, if individuals can adapt their behaviour when anticipating that a beneficial policy change is going to take place the sample selection will be biased. For example, if mothers can time the birth of their child as a response to changes or if parents join the labour force to be eligible for anticipated paid leave benefits (Neugart & Ohlsson, 2013). Thanks to the use of the microdata this type of study can be evaluated within a detailed characterisation of the context (Olivetti & Petrongolo, 2017), which improves their internal validity: the process of determining cause-and-effect relationships. However, the contextual situation of these studies lowers their external validity: the ability to generalize the results of one study to other countries. These considerations are revisited throughout this thesis.

While parental leave can be shared between mothers and fathers, most of the period is taken-up by mothers. Maternity and parental leave are therefore discussed together. Subsequently, paternity leave and the weeks of parental leave reserved for fathers are discussed.

(24)

24 3.1.2. Maternity and parental leave

3.1.2.1. — Leave take-up

Before reviewing the micro-studies evidence on the impact of paid leave on labour market outcomes, it is important to examine how paid leave policy changes affect take-up rates. With regards to the leave-taking rates, several factors may limit the take-up of leave among mothers and fathers, such as a lack of awareness, eligibility requirements, low replacement rates or the absence of job protection (Rossin-Slater, 2017). Moreover, parents may not take up leave to prevent the risk of giving off a negative ‘signalling effect’ — the phenomenon that taking-up leave signals a low career commitment that may be damaging to future career opportunities (Buligescu et al., 2009).

The empirical literature generally shows that both the introduction of paid leave and extensions of existing programs increase take-up rates (Rossin-Slater, 2017). Maternity leave reforms in Germany (Schönberg & Ludsteck, 2014) and Canada (Baker & Milligan, 2008) and parental leave reforms in California (Rossin-Slater et al., 2013; Baum & Ruhm, 2016; Bartel et al., 2018; Bailey et al., 2019), the U.S. (Han et al., 2009), Korea (Kim, 2018) and France (Joseph et al., 2013) have led to increases in female take-up rates. The average treatment effect on take-up rates is a positive 18.4 percentage-point (with a standard deviation of 9.6; ranging from 5.8 to 30.5 point increases) for maternity leave and a positive 10.5 percentage-point (with a standard deviation of 12.1; ranging from 0 to 38 percentage-percentage-point increases) for parental leave

When take-up rates were unchanged, this was because there were generous programs already in place. Namely, extensions in Austria (from 12 to 30 months) in Austria (Lalive et al., 2013) and in the Czech Republic (from 3 to 4 years, exceeding the 3-year job-protection period) (Bicáková & Kalísková, 2016) did both not change the female take-up rates.

3.1.2.2.— Delayed return to work (take-up duration)

Many policy changes have also impacted the duration of leave take-up. Micro-studies evidence points out that the leave duration take-up lengthens with introductions and extensions of paid leave, although the extent in which duration is lengthened differs across policy contexts. The changes in take-up duration are often proportional to the extensions for shorter leave periods, while the duration does not increase by the entire amount for longer leave periods. For instance, a Canadian maternity leave extension from 15 to 18 weeks extended leave duration by the same amount (Dunatchik & Özcan, 2019). Similarly, Schönberg & Ludsteck (2014) find that a

(25)

Eva-Mirre Dam

25 considerable fraction of mothers returns to the labour market exactly at the end of the maternity leave period in Germany. The authors study six different reforms and find a stronger duration extension effect for the shorter leave periods (2, 6 or 10 months) than for the longer leave periods (18, 24 or 36 months).

When generous leave periods are already in place mothers also increase the leave duration period in response to leave extensions, but not by the entire period (Baker & Milligan, 2008; Lalive et al., 2013; Kim, 2018). These findings underline the hypothesis that the marginal value of leave may decrease with the length of leave.

Multiple authors also find that the duration effects are heterogeneous for different groups of women. This may be due to the fact that the reservation wages differ for advantaged versus disadvantaged mothers. For instance, disadvantaged mothers with lower incomes, lower educational levels and without access to employer-funded leave respond most to the Australian Paid Parental Leave Scheme that introduced paid leave for 18 weeks increased leave duration (Broadway et al., 2016). Similarly, the take-up duration impacts of Canadian and Czech Republic parental leave extensions are found to be larger for women from more economically disadvantaged groups — i.e. low-educated mothers (Hanratty & Trzcinsky, 2009; Bicáková & Kalísková, 2016).

3.1.2.3. — Labour force participation

One important goal of paid leave is that is enables parents to stay attached to the workplace. Therefore, it is important to assess what happens to labour market outcomes once the leave period has ended. The average treatment effect, of the articles discussed in tables 2 and 3, on female labour force participation rates is small and the standard deviation is large — the average treatment effect is -0.3 percentage-point for maternity leave, which ranges from -4.5 to 3 percentage-point, and 0.19 percentage-point for parental leave, which ranges from -11.5 to 5.5 percentage-point (with standard deviations of 2.2 and 3.7 respectively).

An adverse impact of paid leave on employment is typically found if the leave period is long. An Austrian increase in parental leave from one to two years detected a small decrease in mothers’ return to work probabilities in the first three years after birth (significant -2.7 percentage-point with an 84.7% baseline) (Lalive & Zweimuller, 2009). Furthermore, a reform in the Czech Republic that extended parental leave from 3 to 4 years impacted non-employment by 11.5 percentage-point (significant) one year after the leave period, with the baseline of 16%

(26)

Maternity Leave

Table 2 — Outcomes of Maternity Leave Policies — Columns 6–12 give respectively: (6) the pre-reform (PRE) or counterfactual female take-up rate (FTU), (7) the treatment effect (TE) on the female take-up rate (in percentage-points), (8) the pre-reform (PRE) or counterfactual female labour force participation rate (FPR), (9) the treatment effect (TE) on the female labour force participation rate (FPR) (in percentage-points), (10) the treatment effect on the long-run (LR TE) female labour force participation rate (FPR) (in percentage-points), (11) treatment effect (TE) on female wages (FWG) (in %), (12 ) long-run treatment effect (LR TE) on female wages (FWG) (in %) Column 9-12 concern the post-leave period and the distinction between short- and long-run effects are based upon the distinctions made by the authors.

Study Country Reform Method Sample

Schönberg & Ludsteck (2007) Germany DD 1975-2001 31,7 +12,7

2 to 6 months (1979) 37,9 -0,2 -0,3 -0,4 -0,5

6 to 10 months (1986) _42,5 _+0,6 _+0,4 ₀ ₀

18 to 36 months (1992) 37,7 +1,6 +1,6 +0,48 0

Buligescu et al., (2009) Germany Paid leave period of 36 months (1992) 2SLS 1994-2005

-10 to -14 -1

Schönberg & Ludsteck (2014) Germany DD 1979-1994 31,5 +12,7 38

2 to 6 months (1979) +30,5 -1 -0,9 +0,3

6 to 10 months (1986) +27,5 -1,2 -0,9 -0,4

18 to 24 months +5,8 -1,9 -1,2 -1,4

6 to 22 months (longer than job protection period in

Baden-Württenberg (1986) +30,3 -4,5 -3,8 -8,5

Baker & Milligan (2008) Canada Introductions of paid leave up to 18 weeks (1976-1990) DDD 1976-2003 43 +12 12,6 +3

Hanel (2013) Australia 12 months unpaid maternity leave (1996)

Matching DD

(27)

27

Study Country Reform Method Sample Period

Naz (2004) Norway Provide families that do not utilise state- subsidised day-care centers benefits up to 400 euros per month for up to 3 years

(1998) DD 1989-1999

TE | FH/W(%): -11,5

Dahl et al., (2016) Norway

Several extensions — i.e. 18 to 20 weeks (1987) _RD 1986-2006

76 +1,3 +2,4 TE | MWG(%): +0,8 i.e. 24 to 28 weeks (1990) 85 +0,9 +2 TE | MWG(%): +2 i.e. 32 to 35 weeks (1992) 87 _-0,7 +2,2 TE | MWG(%): +1,5

Lequien (2012) France _{Introduction of 3 years paid leave benefits for parents of a}

second-born (1994) DD; DDD 1976-2005 _-2,8

Joseph et al. (2013) France Supplementary 6 months of benefits (2004) _DD 2000-2008

17 +38 83 -1,5 -2 to -4

Propensity

Score Matching +4,1

Canaan (2019) _France _{Introduction of 3 years paid leave benefits for parents of a} second-born (1994)

RD 1990-2002 ₅₀ -10,6 to +13,7 _-0,3

TE | MWG(%): +0,12

Lalive & Zweimuller (2009) Austria _{1 to 2 years (1990)}

RD 1985-2000

84,7 _-2,7 _+4.3

Lalive et al. (2013) Austria 12 to 24 months (1990) _{DD; RD} 1990-2005

100 0 78,6; LR:51,3 -7,1; LR: +0,2

24 to 18 months (shorter than job protection period) (1996) _{82,6; LR:51,3}

-2,3; LR: 0 18 months to 30 months (longer than job protection period)

(2000) 79,5; LR:48,6 _{-0,9; LR: +0,9}

Kluve & Tamm (2013) Germany 12 to 14 months and more generous benefits (2007)

RD 2006-2008

34,8; LR: 48 +4,8; LR: -2,8 Pre & TE | MPR(%): 87,6 & _+1,6

Geyer et al. (2015) Germany _{12 to 14 months and more generous benefits (2007)} _{RD; DD} 2006-2007 ₃₀ ₊₂

Kluve & Schmitz (2018) Germany 12 to 14 months and more generous benefits (20067)

RD; DD 2006-2011 33; LR: 42,8 +0,95; LR: +2,9

Bicákova & Kalísková (2016)

Czech Republic 3 to 4 years (1995) DD 1993-2013 95 0 47 -11.5; LR: -1.1

Parental Leave

Table 3 (1/2) — Outcomes of Parental Leave Policies — Columns 6–12 give respectively: (6) the pre-reform (PRE) or counterfactual female take-up rate (FTU), (7) the treatment effect (TE) on the female take-up rate (FTU) (in points), (8) the pre-reform (PRE) or counterfactual female labour force participation rate (FPR), (9) the treatment effect (TE) on the female labour force participation rate (FPR) (in percentage-points), (11) treatment effect (TE) on female wages (FWG) (in %), (12) other, which includes: the counterfactual (PRE) and the treatment effect (TE) of/on the male take-up rate (MTU) (in percentage-percentage-points), the treatment effect (TE) on male wages (MWG) (in%), and the treatment effect (TE) on female hours worked per week (FH/W) (in %)

(28)

28

Study Country Reform Method Sample Period Pre | FTU (%) _{TE | FTU}_(PP) Pre | FPR (%) _{TE | FPR}_(PP) _{TE | FWG}

(%) Other

Baum (2003) U.S. Unpaid FMLA of 6 weeks and 6 weeks paid TDI (1993) DD; DDD 1986-1994 _65,2 _{+1 to +1,8} _+0,49

Han et al. (2009) U.S. Unpaid FMLA of 6 weeks and 6 weeks paid TDI (1993) DD 1987-2004 _55,3 _+8,7 ₆₆ _+4,5

Pre & TE | MTU(pp): 7,2 & 3,9 Pre & TE | MPR(pp)): 93 & 1,2

Goodpaster (2010) U.S. Unpaid FMLA of 6 weeks and 6 weeks paid TDI (1993) DD; DDD 1989-2003 _63,9 _-2,7

Rossin-Slater et al. (2013) U.S. (California) _{Introduction of paid CA-PFL of 6 weeks (2004)} DD; DDD 1986-1994 5,4 +6 to +7 +2 to +11 TE | FH/W(%): +10 to +17

Das & Polacheck (2015) U.S. (California) Introduction of paid CA-PFL of 6 weeks (2004) _DD 1996-2006

71 +1,37

Reed & Vandegrift (2016)

U.S. (New

Jersey) Introduction of 6 weeks of paid leave (2009) DD 2003-2015 _63,2 _-2,4

Pre & TE | MPR(pp): 60,2 & -1,6

Baum & Ruhm (2016) U.S. (California) Introduction of paid CA-PFL of 6 weeks (2004) DDD 2000-2010 ₄₉ _+15,5 ₉₁ _+5,5 ₊₅

TE | FH/W(%): +16

Curtis et al. (2016) U.S. (California) Introduction of paid CA-PFL of 6 weeks (2004) _{DD; DDD} 2002-2004

-1.3 to -2

Bartel et al. (2018) U.S. (California) Introduction of paid CA-PFL of 6 weeks (2004) DD; DDD 2000-2013 _6,4 _+2,3 Pre & TE | MTU (pp): 2,4 & +0,9

Bailey et al. (2019) U.S. (California) Introduction of paid CA-PFL of 6 weeks (2004)

DD 2001-2015

8 +18 76 -1,2 +0,5

Hanratty & Trzcinsky (2009) Canada 15 to 35 weeks (1990-2000) DD 1998-2003 ₅₈ _+0,6

Ang (2015) Canada (Quebec)

QPIP with more generous payments and lower cut-off for eligibility based on earnings (2006)

DD 2000-2008 _81,4 _+1,5

TE | FH/W(%): +2,9

Asai (2015) Japan 12 months at low payments (20%) (1995) DD 1997-2002 _29,6 ₀

Broadway et al., (2016) Australia Australian Paid Parental Leave Scheme of flat rate benefits

of 18 weeks (minimum wage) (2011)

Hazard rates & propensity score matching 2009-2012 69 +4 Kim (2018) Korea 12 to 15 months (2008) RD 2007-2008 ₃₇ _+5,4 _80,9 +1,6 LR: +1,04

Table 3 (2/2) — Outcomes of Parental Leave Policies — Columns 6–12 give respectively: (6) the pre-reform (PRE) or counterfactual female take-up rate (FTU), (7) the treatment effect (TE) on the female take-up rate (FTU) (in points), (8) the pre-reform (PRE) or counterfactual female labour force participation rate (FPR), (9) the treatment effect (TE) on the female labour force participation rate (FPR) (in percentage-points), (11) treatment effect (TE) on female wages (FWG) (in %), (12) other, which includes: the counterfactual (PRE) and the treatment effect (TE) of/on the male take-up rate (MTU) (in percentage-percentage-points), the treatment effect (TE) on male wages (MWG) (in%), and the treatment effect (TE) on female hours worked per week (FH/W) (in %)

(29)

29

Study Country Reform Method

Johansson (2010) Sweden 1-month daddy quota (1995) DDD 1993-2008 SR: +6,1 SR: -2,5

2-month daddy quota (2002) SR: +1 SR: -2,2

Rege & Solli (2013) Norway Introduction of a 1-month daddy quota (1993)

DD 1999-2002 SR: -1 to -3

LR: -1

Cools et al. (2015) Norway Extension of parental leave of 7 weeks with 4-week daddy quota (1993)

DD 1992-2006 ₇₁ _-2,2 ₈₂ _-0,6 _-7,8 _-1,4

Earlier reform (1992) that added 3

weeks of parental leave (comparison) 70 +0,7 81 -0,1 +5,2 -2,4

Hart et al., (2016) Norway Daddy quota from 6 to 10 weeks (2009) RD 2008-2013 _76,6 SR: 0 LR: -0,1 86,7 SR: 0 LR: +0,1 SR: +6,6 LR: +2,4 SR: 0,6 LR: -0,36

Abrahamsen (2018) Norway 2 additional parental leave weeks;

4-week daddy quota (2009) RD 2009-2014 88,5 SR: +0,42

LR: -0,9 91,7 SR: +0,34 LR: -0,11 SR: +2,8 LR: -0,3 SR: -0,5 LR: -0,8 SR: +1,7 LR: +2,4 SR: -1,6 LR: -1,1

Ekberg et al. (2013) Sweden

1-month daddy quota (1995) _RD _1993-2003

87.8 +0.9

Druedahl et al. (2019) Denmark 2 to 13-day daddy quota (1998) DD 1980-2013 _{SR: +3,1} _{SR: -0,64}

Farré & Gonzalez (2007) Spain 2-week paternity leave extension (2007)

RD 2006-2007 _52,5 _{SR: +6,1}

LR: -0,14 93 SR: +1,4 _{LR: +0,1} +3 -2,6

Tamm (2019) Germany Shorter, more generous parental leave period; 2-month daddy quota (2007)

DD 2000-2015 _{SR: -25}a

LR: -4

Dunatchik & Özcan (2019) Canada (Quebec) 5-week (or 3 with higher RR) daddy-quota; extended maternity leave (by 3 weeks); 3 week shorter parental leave; more generous replacement rate (2006)

DD 2007-2014 75 +5 85

-1 _{SR: +3}

LR: +6

Patnaik (2019) Canada

(Quebec) 5-week (or 3 with higher RR) daddy-_{quota; extended maternity leave (by} 3 weeks); 3 week shorter parental leave; more generous replacement rate (2006) RD; DD; DDD 2005-2010 57,2 +4,6 83,6 +2 LR: DD: +4,3 DDD: +13,8 LR: DD: +0,03 DDD+ 5

Table 4 — Outcomes of Paternity Leave Policies — Columns 6–12 give respectively: (6) the pre-reform (PRE) or counterfactual female (8: male (MPR)) labour force participation rate (FPR), (7) the treatment effect (TE) on the female (9: male (MPR)) labour force participation rate (FPR) (in percentage-points), (10 ) treatment effect (TE) on female (11: male (MH/W)) working hours (FH/W) (in %), (12) treatment effect (TE) on female (13: male (MWG)) wages FWG) (in %)

Column 7,8 and 10 concern the post-leave period (except a

) and the distinction between short- and long-run effects are based upon the distinctions made by the authors.

a

This estimate refers to the period when the child is 0-14 months (“during leave”). Paternity Leave

(30)

The labour market impact of paid leave (Bicáková & Kalísková, 2016). A significant negative effect is, however, also for the relatively short 6-week parental leave reform in New Jersey (a reduction of -2.4 percentage-point with a 63.2% baseline). Yet, this is probably due to the fact that the reform does not provide job protection (Reed & Vandegrift, 2016).

Many authors, however, fail to find significant effects (Baum, 2003; Goodpaster, 2010; Baker & Milligan, 2013; Han et al., 2013; Lalive et al., 2013; Geyer, et al., 2015; Dahl, et al., 2016; Kluve & Schmitz, 2017). There may be a lack of effect because there are no search frictions on the labour market, which reduces the added value of a job-protected leave period as employers and employees can easily find a ‘match’ on the labour market after career interruptions. Conversely, Asai (2015) attributes the lack of finding effect of leave policy changes in Japan to the limited availability of childcare, inflexible working hours and gender norms, which shows that the institutional context also matters. If workers could already make use of paid leave arrangements, in case employers-provided leave benefits were already widespread, this can also explain why no effect is found.

Furthermore, there seems to be a consensus that the effects of maternity and parental leave on employment rates dissipate in the long-run. Almost all the long-run estimates in tables 2 and 3 are smaller than their short-run counterparts. Negligible long-run effects are found for the CA-PFL at 5 to 11 years after birth (Bailey et al., 2019), for several Austrian parental leave reforms (Lalive & Zweimüller, 2009; Lalive et al., 2013), the German Elterngeld reform (Kluve & Tamm, 2013) and a Canadian extension from 15 to 35 weeks (Hanratty et al., 2009).

Although the findings are not fully concordant, most of the quasi-experimental evidence underlines the image that the effect of maternity and parental leave on female labour market participation is small. The effect is unlikely to be substantial as most authors fail to find any significant effect, whereas the significant effects typically dissipate in the long-run. Most of the negative estimates concern relatively longer leave durations (over 2 years) or for paid leave durations that exceed the job-protection period (Schönberg & Ludsteck, 2014). Moreover, multiple authors highlight that the effect depends upon the context that the paid leave statute is situated in; section 3.1.4.2-4 revisits some of these aspects.

3.1.2.4. — Weekly working hours

Besides studying whether mothers and fathers participate in the labour force it is also of interest to study the hours worked in the labour market when studying couples’ specialisation between household and market labour, in relation to Becker’s theory described above. Unfortunately,

She cares and he works — Does paid leave decrease the labour market gap? A Double- and Triple-Differences Analysis

Abstract

Table of contents

1. Introduction

2. Theoretical Background

2.1. Defining leave

2.2. Justifications for paid leave policies

2.3. Mechanisms driving potential labour market outcomes

2.4. Becker’s theory of intra-household specialisation

2.5. Hypotheses

3. Literature Review

3.1. Micro-Studies