The impact of parental leave policy on child-rearing and employment behavior: The case of Germany

Chirkova, Serafima Article The impact of parental leave policy on child-rearing and employment behavior: The case of Germany IZA Journal of Labor Policy Provided in Cooperation with: IZA – Institute of Labor Economics Suggested Citation: Chirkova, Serafima (2019) : The impact of parental leave policy on child-rearing and employment behavior: The case of Germany, IZA Journal of Labor Policy, ISSN 2193-9004, Sciendo, Warsaw, Vol. 9, Iss. 1, pp. 1-27, https://doi.org/10.2478/izajolp-2019-0007 This Version is available at: https://hdl.handle.net/10419/222168 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Serafima Chirkova* The Impact of Parental Leave Policy on Child-Rearing and Employment Behavior: The Case of Germany Abstract Parental leave and child care are important instruments of family policies to improve work– family balance. This paper studies the impact of the substantial change in Germany’s parental leave system on maternal employment. The aim of the reform was to decrease birth-related maternal employment breaks by providing more generous parental benefits for a shorter period of time. Using the German Socio-Economic Panel data for 2002–2015, I exploited quasiexperimental variation in the benefits to estimate the impact of the reform. I incorporated the mother’s decision to substitute her care time with the public child care. To control for the availability of child care, I used spatial and temporal variation in the availability of childcare slots. Overall, I did not find significant changes in maternal employment during the first three years of motherhood after the reform implementation. Only for high-income mothers, the reform produced a significant decrease in the employment participation during the first year of leave and an increase in employment probability after the benefits expired. The empirical findings suggest that the restriction in the childcare availability became an important constraint for the employment effect of the reform. Current version: June 06, 2019 Keywords: employment; child care; parental leave; financial incentives JEL-codes: J13, J18, J22, C25 Corresponding author: Serafima Chirkova serafima.chirkov[email protected] Chirkova. IZA Journal of Labor Policy (2019) 9:7 © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Cite as: Chirkova. IZA Journal of Labor Policy (2019) 9:7. https://doi.org/10.2478/izajolp-2019-0007 Departamento de Economía, Universidad de Santiago de Chile, Av. Libertador O’Higgins 3363, Santiago, Chile Page 2 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 1 Introduction The relatively low labor participation rate of mothers with young children is a particular characteristic of female labor market. Figure 1 reports the significant differences in employment rates of females with and without children in 2012. These differences are significantly large for mothers with children aged zero to two years. Germany has one of the lowest participation rate for this category (53.8%) among the reported countries. At the same time, the German females without children report one of the highest participation rates (8.29%). The comparison of these indicators with the values reported in 2006 (Figure 2) reveals the persistence in such participation distribution. The trade-off faced by women with small children1 and working careers — stay at home to take care of a child or return back to work — is widely discussed in both the academic literature and social policy debates. Mothers may base their decision on different factors: e.g., preferences, social norms, income, and time constraints. In general, family policies seek to relieve economic constraints and maintain the work–family balance. The concerns surrounding such policies are numerous: decreasing fertility rates, demographic burden on the social security system, low employment participation rates of mothers, and child welfare, among others. The implementation of family policy and its effectiveness can vary significantly across countries and institutional contexts. Parental leave is one of the principal components of government policy supporting the reconciliation of family life and careers for women. There is a vast strand of literature analyzing the impact of the parental leave system on mothers’ employment behavior. The empirical 1 Children younger than 6 years. Figure 1 Employment rates of females aged 25–54 years by age of the youngest child, 2012. Source: Calculations based on data from the Organisation for Economic Co-operation and Development (OECD) Family Database 2012. Online at: http://www.oecd.org/els/family/database.htm. Page 3 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 evidence supports a negative relationship between parental leave benefits and maternal employment (Apps and Rees, 2004; Lalive and Zweimuller, 2009; Thevenon and Gauthier, 2011). If a woman with a young child has the option to stay at home on paid leave, she will use it. The longer the job-protected leave and the higher the parental leave payment, the longer a woman’s employment break will last. As a consequence, the more difficult it will be to find employment afterward. Another important instrument of family policy is the system of childcare provision. This type of governmental support tries to ease mothers’ time constraint by subsidizing alternative child care. Childcare policies usually subsidize public childcare institutions or compensate the parents’ costs related to child care. Figure 3 shows the positive association between maternal employment and children enrollment in public child care across European countries in 2012.2 Germany is clearly noted for a low level of childcare provision (17.8%) compared to the average (37%).3 Overall, the literature studying childcare policies finds mixed evidence of the effect of public childcare expansion on mothers’ employment behavior (Blau and Currie, 2006; Fitzpatrick, 2010; Cascio, 2009). At the same time, few studies focus on possible influence of the structure of the childcare market on the effectiveness of other family policy instruments (Geyer etal., 2015; Bick, 2016). The aim of this paper was to address this gap by evaluating the effect of a paid parental leave reform on mothers’ employment decisions, given the existing childcare system. In 2 The childcare systems clearly differ across European countries, depending on the form of provision and funding for it. Nevertheless, formal child care, which includes daily operated childcare centers and registered childminders, remains the most important source of provision. 3 One may think that informal childcare arrangements can offset the lack of formal institutes. Similar plotting does not reflect any correlation between informal care and maternal employment. Since the informal arrangements are an important factor, I discuss them in the following sections. Figure 2 Employment rates of females aged 25–54 years by age of the youngest child, 2006. Source: Calculations based on data from the OECD Family Database 2012. Online at: http:// www.oecd.org/els/family/database.htm. Page 4 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 particular, I analyzed the 2007 German parental leave reform that considerably changed the system of family compensation for time that mothers spent out in the labor market. Under the prereform system, a mother received a means-tested leave benefit of 300 euros per month that was paid out for two years and targeted at lower income families. The reform replaced it with the universal payment of 67% of annual prebirth net labor earnings, paid for just one year and extending the group of eligible families to the entire population. The unanticipated and quick implementation of the new regulation provided a quasi-experimental environment in which to evaluate the reform’s impacts on different outcomes.4 The new policy was designed to meet various goals. The payment was intended to counteract the decline in families’ earnings in the first year after birth. Reducing the payment period from two years to one year was supposed to incentivize the mother’s return to the labor force once the payment was terminated. To increase the father’s involvement in caring for the child, the new policy provided a “father’s quota” of two paid months if a family decided to share childcare responsibilities (14months of paid leave in total). Overall, the new policy intended to make parenthood more attractive to women with careers and to make work more attractive to women who had not been in the labor force. The reform turned to increase total parental leave benefits for middleand high-earning women and to decrease it for low-earning women. Various studies focus on the impact of the 2007 reform on maternal employment. Kluve and Tamm (2013) found evidence of a significant decline in mothers’ employment probability during the first year of motherhood and an increase of that in the second year, once the benefit expires.5 Bergemann and Riphahn (2011) showed the positive effect of the German parental leave reform on mothers’ intentions to return to work. They found a positive response in maternal labor force participation once the benefit is expired, using semi-parametric Cox 4 Refer [18] for a detailed review. 5 The author exploits a “natural” experiment using cross-sectional survey data collected by health insurance funds in two federal states. Figure 3 Relationship between maternal employment and child enrollment in formal child care, 2012. Source: Calculations based on data from European Union Statistics on Income and Living Conditions (EU-SILC) survey 2012. Online at: https://ec.europa.eu/eurostat/web/ income-and-living-conditions/data/database. Page 5 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 hazard models. Overall, these studies concluded that there is a decrease in maternal employment during the first year after childbirth and an increase in return to work in the second year associated with the reform implementation. Kluve and Schmitz (2018) showed that this positive effect persists for the medium run (three to five years after childbirth) for mediumand high-earning mothers. The important drawback of these studies is that they do not consider the availability of child care. Geyer et al. (2015) found a decline in maternal labor supply during the first year and an increase in that in the second year. Using a structural labor supply model, they examined the role of subsidized child care. The simulated results show that significantly larger maternal employment effects might be observed once universal child care is available. A number of studies discussed childcare provision in Germany and its role in maternal employment. Muehler (2010) compared public sector and non-public providers. She emphasized the lack of full-day slots for children under three years of age and a significant regional variation in access. Using microsimulations, Haan and Wrohlich (2011) concluded that childcare subsidies conditional on maternal employment status increase the labor supply. Bick (2016) counterfactual policy analysis showed that insufficient subsidies for childcare provision decrease labor participation by mothers with children younger than three years. Schober (2012) focused on the allocation of childcare time between parents and found that the 2007 parental leave reform increased fathers’ involvement in child care. Following the literature, I studied on the relationship between maternal employment and childcare decisions in each of first three years of motherhood. The reform sought to support families with a newborn through increased parental benefits and to decrease birth-related employment breaks through a shortened paid-leave period. Given that maternal employment is conditional on access to child care, I focused on the reform’s effect on mother’s employment decision in the context of restricted childcare alternatives. To identify the effect of the reform, I exploited temporal variation in parental leave benefits and temporal and spatial variations in childcare provision. Using German Socio-Economic Panel (GSOEP) data for 2002–2015, I estimated a bivariate choice model of maternal employment and childcare decisions in two specifications with and without controlling for a mother’s childcare choice. Without controlling for childcare decision, the findings suggested the positive impact of the reform on the employment after childbirth. These results are in line with the previous findings of Kluve and Tamm (2013), Bergemann and Riphahn (2011), and Geyer et al. (2015). Once the estimations are corrected to the childcare factor, the policy effects turn insignificant. Only the mothers in fulltime employment report the increase in participation associated with policy (6.4% points). The effect is driven by high-income mothers. Estimating the possibility of childcare arrangements and labor force participation jointly, the analysis provides new evidence on the importance of institutional constraints for policy implementation. I contribute to the literature showing that the estimation of previous studies is likely to be biased since the childcare decision is not considered. My empirical evidence confirms the importance of childcare availability for the mothers’ labor market participation. The enrollment in public childcare increases significantly labor force participation. At the same time, I observed that higher availability of public childcare implies higher probability of its usage. The paper is organized as follows. Section 2 focuses on the institutional characteristics of parental leave reform and the childcare system in Germany. Section 3 discusses the Page 6 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 identification strategy and presents the German data used for the analysis. Section 4 describes the main findings, and Section 5 concludes. 2 Institutional background 2.1 Family policy in Germany For a long time, Germany has implemented a complex family policy targeted at maternal protection and child-rearing. A family with a newborn is entitled to maternity leave benefits, parental leave benefits, child benefits, a child tax allowance, and additional household benefits based on the number of children. The benefits are almost universal in Germany. The eligibility criterion is the parent’s German or European citizenship and a settlement permit or a residence permit with at least three years’ past employment in Germany. The maternity leave system consists of a period of 14 paid weeks, typically divided into six weeks before childbirth and eight weeks afterward. During maternity leave, mothers are not allowed to work but are insured against dismissal. They also receive maternity benefits (Mutterschaftsgeld) – a net government transfer of up to 13 euros per day and additional coverage equal to the net income if the woman is eligible for social insurance. Employed women without government insurance receive a one-time maternity allowance of up to 210 euros6 and supplements from the employer. The parents receive an untaxed allowance (Kindergeld) until the child’s 18th birthday. For the first and second child, this consists of 184 euros, with a slight increase to 190 euros for a third child and to 214 euros for each child thereafter. The parents can deduct part of the income necessary to maintain a child from their tax return. Low-income households7 can further apply for an in-work tax credit (Kinderzuschlag) of up to 140 euros per month for six months. Parents may also be eligible for the accommodation allowance (Wohngeld) in the form of home-ownership subsidies for low-income families with a higher number of children. The German paid-leave system was established in the mid-1950s when mothers have been entitled to job-protected paid leave 6weeks before childbirth and 8weeks after childbirth.8 In 1979, the duration was raised to six months, but the payment remained unchanged. In 1986, the duration of the job protection and payment leave was increased to ten months. In addition, fathers became eligible for parental leave. Further reforms in the beginning of the 1990s increased the job-protected period to 36months (1992) and the duration of leave payment to 24months (1993). These reforms defined the structure of the parental leave system for almost 15years, before a new set of family policies came into force.9 Under this original system, a parent could share job-protected parental leave of up to three years (Elternzeit). During parental leave, the parent who stayed at home (or worked fewer than 30hours per week10) received the child-rearing benefit (Erziehungsgeld). It was a universal means-tested transfer with an income cap (30,000 euros for the first six months and 6 http://ec.europa.eu/employment_social/missoc/db/public/compareTables.do?lang=en. 7 Income up to 900 euros per couple or 600 euros per single parent. 8 The following description of changes in the paid leave system is based on Schönberg and Ludsteck (2014). 9 During this period, the parental leave system saw some minor modifications. For instance, since 2001, parents with a part-time contract have been entitled to benefits. 10 It is equivalent to reduced full-time work in Germany. Page 7 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 16,500 euros afterward for a married couple11). The eligible period was up to 24months, and the payment was 300 euros per month.12 If the income threshold was exceeded, the benefit was withdrawn at a rate of 63%. Thus, the child-rearing benefit depended on the gross income of the household and the mother’s labor participation decision. Under the old policy, the parental leave payment was targeted to low-income families, as the income threshold and the amount made the payment negligible for middleand high-income households.13 Even though fathers became eligible for parental leave in 1986, the male take-up ratio of the leave was insignificant. Parental leave was effectively maternal leave, and child-rearing was considered a traditional maternal activity within the “bread-winner” household model. German mothers delayed their return to work, and the availability to take longer leave weakened their labor force attachments (Schönberg and Ludsteck, 2014; Geyer and Steiner, 2007; Ondrich et al., 2003). 2.2 The 2007 parental leave reform Negative trends in reproduction and the low labor market participation rate of mothers with children younger than three years14 led to further reforms to the parental leave system. The new parental leave system was established at the end of 2006 and took effect for children born after January 1, 2007. The main changes in the reform were financial. Under the new system, the means-tested child-rearing benefit (Erziehungsgeld) has been replaced by the parental allowance (Elterngeld), which accounts for 67% of the rearing parent’s net labor income per month. The benefit is calculated on the basis of the average labor earnings 12months prior to the birth. If a parent caring for a child was economically inactive prior to the child’s birth, she (he) receives a flat minimum of 300 euros per month (3,600 euros per year). This minimum is also paid to those whose parents’ benefits would otherwise be less than 300 euros (part-time employees, those with a mini job, low earners). The benefit is truncated at 1,800 euros per month (21,600 euros per year), meaning that a parent with prebirth monthly net labor earnings more than 2,700 euros can get only the maximum payment. The benefit replacement rate increases to 100% if the child-rearing parent’s net income is less than 1,000 euros per month. Another important change in the parental leave legislation is the cutoff in the duration of payment. Parental leave payments are granted for 14months if both parents share the child-rearing and for 12months in a single-carer household. The condition to work no more than 30hours a week remains unchanged, but the reduction in the amount and duration of benefits makes part-time employment less attractive. In 2010, approximately 2%–9% of mothers were employed part-time while receiving benefits. The reform was designed with various goals in mind. First, it aims to provide substantial financial support to families with children younger than one year, targeting middleand high-income households.15 Second, it encourages mothers to return to work after childbirth. 11 These thresholds were applied from 2004 onward, and they had previously stood at 51,130 euros per year per married couple. They were different for single parents and varied with the number of children in the household. 12 The parents could choose a payment of 450 euros per monthly for up to 12 months. This option was chosen by approximately 13% parents. 13 The income cap allowed about 77% of all parents to claim the benefits during the first six months after childbirth. Afterward, only 50% of families were entitled to the benefit. 14 For instance, 48% in Germany compared to 77% in Sweden (OECD 2006). 15 Huebener et al. (2016) provided a detailed discussion of the critics of the political shift in the distribution targets. Page 8 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 Third, it addresses gender equality by enhancing the father’s involvement in child care. Initially, an increased birthrate among working women was not named as an official goal of the reform, but it was added later (Huebener et al., 2016). As Spiess and Wrohlich (2008) point out, the main idea of the 2007 reforms was to move the design of the existing parental system toward the Scandinavian model, in which family policy supports the “dual-earner household.” There was a belief among German politicians that the Scandinavian model, with its high supply of publicly funded daycare and its generous parental leave system, promotes higher employment rates among mothers with young children, as well as higher fertility rates due to policies that reduced family income losses after childbirth. According to administrative statistics, the take-up ratio was about 96% of all mothers between 2007 and 2010, with the average duration of parental leave almost a full period of 12months. The average amount of benefits was also increasing. Comparing the simulated preand postreform benefits, Raute (2019) showed that the financial gain in benefits was continuously increasing in the postreform period. The administrative data also report a significant increase in the take-up ratio of parental leave by fathers. In the first year after the reform, the share jumped from 2.5% to 15%. It increased continuously and was more than 34% in 2014. However, Kluve and Tamm (2013) showed that the higher participation in parental leave did not translate into significant involvement of the father in child-rearing. Most of the fathers received the benefit during the two months that could be added to the mother’s 12months and then returned to work. 2.3 Childcare provision The German childcare system represents a universal mixed-market provision by public institutions and different types of non-state providers, such as religious institutions, private nonprofit organizations, commercial daycare centers, nannies, childminders, and family daycare. While public centers are under municipality or local-authority control, private nonprofit providers belong to welfare organizations. Centers run by churches are part of the public market. The owners of commercial centers are entrepreneurs or companies providing childcare services for employees. According to the survey “Children and Personnel in Child-Care Centres”,16 public centers had 34% of total market share in 2009. The religious centers run by churches covered 36% of the market, while the nonprofit centers operated by nonpublic institutions represented 28% of the market share. The proportion of commercial centers was significantly low – only approximately 2%. Apart from the institutes run by public authorities, all the other categories were eligible for public finance. Therefore, the German childcare market can be defined as highly state dependent. The German childcare market is characterized by price rationing. The local communities that finance and run the majority of public childcare institutions set their fees based on household income and the child’s age.17 Borck and Wrohlich (2011) reported the following numbers for children younger than three years: in 2005, the parents’ monthly fees, on average, were 16 The data are a part of the annual “Child and Youth Welfare Survey” collected by the German Federal Statistical Office. 17 The privately subsidized institutions use a similar scheme. Page 15 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 Table 4 Summary statistics Observational period 2006–2007 2005–2008 2002–2011 Before 2007 After 2007 Mean test Before 2007 After 2007 Mean test Before 2007 After 2007 Mean test Personal characteristics (1) (2) (2)-(1) (3) (4) (4)-(3) (5) (6) (6)-(5) Age 30.34 (4.919) 31.27 (5.109) 0.932 (0.674) 30.91 (5.090) 31.27 (5.173) 0.352 (0.398) 30.93 (5.060) 31.43 (5.174) 0.503* (0.288) Education 12.64 (2.563) 13.11 (2.256) 0.462 (0.329) 12.79 (2.646) 13.28 (2.711) 0.493*** (0.181) 12.81 (2.700) 13.27 (2.717) 0.456*** (0.132) Married 0.715 (0.453) 0.593 (0.494) -0.123** (0.062) 0.687 (0.464) 0.658 (0.475) -0.029 (0.031) 0.722 (0.448) 0.648 (0.478) -0.074*** (0.022) Cohabiting 0.228 (0.421) 0.278 (0.450) 0.050 (0.057) 0.248 (0.433) 0.247 (0.432) -0.002 (0.029) 0.213 (0.410) 0.252 (0.434) 0.039* (0.020) Child gender 0.504 (0.502) 0.491 (0.502) -0.013 (0.066) 0.481 (0.500) 0.524 (0.500) 0.043 (0.033) 0.497 (0.500) 0.509 (0.500) 0.013 (0.024) First child 0.797 (0.404) 0.722 (0.450) -0.075 (0.056) 0.674 (0.469) 0.831 (0.375) 0.157*** (0.028) 0.583 (0.493) 0.844 (0.363) 0.261*** (0.021) Labor characteristics before birth Employment 0.776 (0.419) 0.750 (0.435) -0.026 (0.057) 0.734 (0.442) 0.787 (0.410) 0.053 (0.034) 0.674 (0.469) 0.711 (0.454) 0.038 (0.027) Employment part-time 0.190 (0.394) 0.148 (0.357) -0.042 (0.050) 0.191 (0.394) 0.144 (0.352) -0.047 (0.030) 0.180 (0.384) 0.119 (0.324) -0.061*** (0.021) Employment full-time 0.526 (0.501) 0.463 (0.501) -0.063 (0.067) 0.474 (0.500) 0.555 (0.498) 0.081** (0.039) 0.420 (0.494) 0.511 (0.500) 0.091*** (0.029) Labor income 18,667.4 (15,390.2) 19,154.3 (18,279.5) 486.862 (2,252.418) 18,928.5 (16,312.5) 21,143.3 (16,526.4) 2,214.814* (1,290.415) 17,619.3 (16,733.1) 21,148.2 (16,066.2) 3,528.880*** (964.485) Household labor income 48,008.4 (31,253.0) 47,542.4 (35,309.4) -466.088 (4,448.715) 53,433.7 (35,113.0) 49,731.1 (33,150.1) -3,702.662 (2,705.036) 51,668.8 (34,682.5) 50,556.0 (33,508.7) -1,112.802 (2,003.282) Partner characteristics Age 33.94 (6.496) 34.45 (5.950) 0.510 (0.932) 34.13 (5.836) 34.75 (5.808) 0.615 (0.502) 33.96 (5.750) 34.48 (5.853) 0.519 (0.366) Education 12.55 (2.683) 13.04 (2.661) 0.485 (0.398) 12.77 (2.860) 13.03 (2.826) 0.264 (0.211) 12.82 (2.956) 13.00 (2.849) 0.185 (0.157) Other characteristics Parental leave benefits per month 421.7 (381.5) 712.6 (518.5) 290.913*** (76.575) 402.3 (397.3) 741.3 (486.5) 338.955*** (52.111) 391.7 (290.4) 768.2 (466.0) 376.446*** (24.073) Parental leave benefits per year 2,823.1 (1,629.5) 4,984.3 (4,393.4) 2,161.185*** (491.531) 3,164.1 (2,593.8) 4,978.9 (4,161.6) 1,814.857*** (360.541) 3,293.8 (2,661.2) 4,899.0 (3,792.3) 1,605.207*** (186.716) Childcare coverage 0.178 (0.143) 0.222 (0.145) 0.044** (0.019) 0.171 (0.146) 0.237 (0.137) 0.065*** (0.009) 0.162 (0.144) 0.250 (0.134) 0.087*** (0.007) Sample size 486 913 1,748 Notes: Standard errors in parentheses. *p < 0.10, **p < 0.05, and ***p < 0.01. Page 16 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 in the labor market in full-time jobs, and have higher labor income. At the same time, the households do not report significant differences in household incomes. Other characteristics remain similar. Following Laroque and Salanié (2014), the identification relies on the variation in financial incentives induced by the parental leave reform, the variation in the partner’s labor income, and the regional variation in access to a public childcare slot. The GSOEP provides information on the parental leave benefits received. First, an individual reports the average monthly amount of paid benefits; second, based on the observed socioeconomic characteristics of an individual and a household, the variable of the yearly paid maternal allowance is generated. According to the descriptive statistics for the estimation sample 2002–2011, the parental benefits significantly increased by, on average, 1,600 euros per year (almost 400 euros per month) after the implementation of the reform (see Table 4). Figure 4 shows the distribution of the reported benefits for the preand postreform periods. As expected, for all wage categories, the expected payment shifted to the right. In addition, there is a substantial variation in the introduced benefits for eligible mothers. Descriptive statistics on childcare coverage show significant variation for the treated and control groups. These differences capture the variation in child care between states in the cross-section and over time discussed in Section 2. 3.6 Empirical strategy Following Del Boca and Daniela (2007), I considered a household that derives utility from household net income Yit, leisure time lit , and child care CCit . To avoid the problem of household bargaining, I assumed that in each time period, a woman optimally decides about her employment Lit and child care CCit behavior, i.e., she maximizes the household utility conditioning on her partner’s behavior.25 U = u(cc, l) The goods are assumed to be normal. As a result of the utility maximization problem, we observe the following maternal outcomes: • the mother i’s employment outcome empit at time t (binary variable is equal to 1 if she decides to work at a regular job) and • the mother i’s childcare outcome ccit at time t (binary variable is equal to 1 if she decides to use formal child care). The return-to-work decision is correlated with the decision about child care. On the one hand, a mother needs to substitute for her time in order to work. On the other hand, if she prefers to remain the main childcare provider, she does not return to her job. Therefore, I considered these decisions simultaneously. Summing up, at time t for a household i, the female choice set is represented by the following empit × ccit combinations: 25 Steiner and Wrohlich (2004) showed the insignificance of the cross-elasticities among the partners. Page 17 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 • (0,0): the mother does not work, staying in the household as the principal childcare provider; • (1,1): the mother works, and she engages formal child care; • (0,1): the mother stays at home and takes care of the child, but with additional help from formal child care; and • (1,0): the mother combines work and child care using informal help. Notice that the net household income varies with the set of possible outcomes. To evaluate the policy effect and to account for the potential correlation between employment and childcare decisions, I estimated the bivariate model. I used the existing variation in Figure 4 Maternity benefits in preand postreformed period. Source: Calculations based on data from GSOEP survey v34. Page 18 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 childcare provision at the state level to model the decision of formal childcare usage. Then, I exploited the quasi-random variation in financial benefits at the individual level to identify the effect of the reform on the employment decision given the mother’s childcare decision. The empirical specification of the model is as follows: αα αα α () =+ ++ + 01 Pemp demplbef demplbef cc ijti ij ii ji jt23 4 ∑ α ++ = Xu lijt ij t l k 11 5 , (1) ∑ ββ νβ β () =+ ++ + = Pcccco ccinf Xu ijtjtijt ij ti jt l k 01 2122 3 . (2) where P(empijt) is the employment probability for the mother i at period t in the state j, di is a policy indicator for the postreform period starting in January 2007, and emplbefij is an employment status indicator for the mother i prior to her pregnancy. The variable P(ccijt) represents the probability of formal child care usage for the same mother. cccovjt is the childcare enrollment ratio in region j in period t, ccinfit is an indicator variable for the actual use of informal child care, and X1ijt and X2ijt are vectors of various individual and household sociodemographic characteristics. The general set of individual control variables for the mother includes age, education, marital status, her partners’ characteristics, and household income. I assumed that characteristics such as public childcare availability, use of informal child care, number of children, child’s gender, and whether the child is a firstborn have a direct impact on the childcare decision, but not on the labor participation decision. Notice that the subscript t : t = {1,2,3} refers to the years of parental leave after childbirth. For instance, t = 1 refers to the decisions made during the first year of child leave. Since I focus on the first transition into employment, the probability of being employed in the subsequent period is conditional on the previous nonparticipation decision. The model (2) allows correlation between employment and childcare decisions through two channels. First, there is a direct effect of the childcare decision on the participation decision (as measured by the parameter a4 from equation (1)). Second, the model allows correlation between individual-specific time-invariant effects, corr(u1ijt, u2ijt) ≠ 0. To test the significance of the childcare decision variable, I considered two specifications of equation (1) with and without ccijt. I estimated the system (1, 2) using the bivariate probit model. The parameters of interests are as follows: • a1 that measures the impact of the reform on mothers’ employment participation at extensive margin and • a3 that measures changes in labor supply of females who worked prior to a childbirth. Given the reform’s design, one might expect heterogeneous response with female employment status prior to birth. To evaluate the effect of the 2007 reform on mothers’ return to labor force, I used a quasiexperimental environment generated by the unanticipated changes in the parental leave system. Page 19 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 The treatment group consisted of the mothers who were entitled to the benefit after the reform implementation – i.e., women with children born on or after January 1, 2007.26 The control group consisted of mothers who gave the birth before the reform was implemented. The estimation of the causal effect of the reform has various issues. Since the treatment is universal, every mother is eligible for benefits. Therefore, we could not observe any control group outcome after the reform implementation. Thus, the empirical strategy consists of a before–after comparison of mothers from the control group, who gave birth shortly before a reform, and mothers from the treatment group, who gave birth shortly after the reform.27 The key identification assumption was that, without implementation of the reform, mothers in both groups would have behaved similarly in their employment decision after childbirth. This assumption required that the reform did not influence the fertility patterns in either groups. The empirical studies based on German Microcensus data present mixed evidence on the reform’s impact on the fertility patterns in the short and middle run. Cygan-Rehm (2016) found that the reform had a negative effect on the fertility decisions of low-income mothers through birth postponement. She also documented weak positive effects for other income groups. Raute (2019) pointed to a jump in fertility rates in August 2007. She also found an increase in fertility rates only for highly educated women over the five-year postreform period. Kluve and Tamm (2013) showed that the reform was not anticipated by families. They presented a detailed analysis of the legislative process and news coverage, pointing out that the reform took place quite quickly. The government coalition agreed on the reform in May 2006, and the law was passed in the parliament four months later, in September 2006.28 Then, one might expect that there was no self-selection in treatment for parents who conceived child before October 2006 (children born before June 2007). Another potential threat to the validity of identification was the 2007–2009 financial crisis that could have had consequences for the employment and fertility behavior of the population. The empirical evidence suggests that the German economy, unlike the economies of other European countries, did not suffer long-term consequences of the crisis. On the contrary, Weber and Weber (2013) reported a decrease in unemployment rates during the period 2006– 2011. Raute (2019) found no shift in fertility behavior due to the economic crisis. The parental leave reform coincided with the expansion of child care – a combination of policies that could have affected maternal employment behavior. By construction, however, the two reforms seem not to be systematically related.29 The parental leave system is regulated and financed by the federal government, while the expansion of subsidized child care is supported partially by the federal government and partially by the states. The number of available childcare slots and infrastructural investments vary significantly by the state. Table 1 reports the childcare coverage in terms of the percentage of all children younger than three years who attended child care at the state level. Overall, the average attendance rate for children younger than three years increased from 13.6% in 2006 to 32.9% in 2015. We can see in Table 5 that 26 I focused on mothers’ labor force participation decisions since that was one of the key goals of the reform (see Section (2)). In addition, the “father’s quota” did not increase the fathers’ involvement in child care and significantly changed fathers’ employment pattern (Kluve and Tamm. 2013) 27 A similar identification strategy has been used widely in the literature. In this context, see Schönberg and Ludsteck (2014), Kluve and Tamm (2013), and Geyer et al. (2015). 28 See, also, Kluve and Schmitz (2018). 29 See the discussion of childcare reform in Section 2. Page 20 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 the demand for slots was higher than the supply for all states during the observation period.30 All regions report an upward time trend in number of open slots, and no significant jump is observed around the reform implementation in 2007. This implies that the trends in childcare demand–supply over time were the same for both control and treated groups. 4 Estimation results All estimations of this section are based on the main sample that includes mothers with children aged zero to three years born from 2002 to 2011. Table 5 presents average marginal effects for maternal employment computed using the bivariate probit model (1, 2) estimation. The estimated model controls for a set of individual and household characteristics. The specification includes the mother’s age, education, marital status, prior-to-birth labor income, current household income, received amount of parental leave benefits, and partner’s characteristics. I also controlled for the dwelling location: Eastern or Western Germany. In addition, I included an indicator variable for a woman who gave birth to her first child, gender of the child, and the number of children in the childcare equation. I also controlled for the available help for a newborn from a partner and relatives, including the informal childcare variable. Table 5 is structured as follows. Estimates are reported in three groups of rows, one each for a year of motherhood. I considered three participation decisions of mothers: to work, to work full-time, and to work part-time. For each of these three groups, I reported average marginal effects of the reform with and without childcare decision control. For the first year of motherhood, the data do not report significant changes in mothers’ job participation decisions. Notice that the sign of the coefficient is negative, which supports the hypothesis that labor force participation drops during the first year of motherhood. These results confirm the same pattern identified in the previous research: once we control for covariates, the probability of return to work remains unchanged or drops insignificantly. The childcare decision does not influence the employment decision. The partial explanation might be related to the fact that the number of childcare facilities for a child younger than one year is more limited comparing to the slots for children aged one to three years. Being employed before birth increased the marginal probability of return to work by almost 10% point both for preand postreform mothers. It does not vary after policy implementation and remains both in full-time and part-time participation groups. For the second year of motherhood, when the parental leave payments stop, a positive significant shift in the employment probability by 20% points is observed. However, the effect of the introduced incentives becomes insignificant after controlling for the childcare decision. It implies that the estimate of the policy impact is biased upward and captures partially the possibility of the mother to use a formal childcare. The results also suggest that the mothers who use formal child care increase their marginal probability to return to work by 40% points comparing to the mothers without access to public child care. The policy effect remains statistically significant only for the group of mothers who work full-time. The marginal probability to get full employment is 6.4% points higher for postreform mothers compared to those from the prereform group. Interestingly, this specification reveals 30 The administrative data for the actual enrollment are available from 2006 and onward, and the data for the slots offered are available only from 2012. Page 21 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 Table 5 Average marginal effects for maternal labor outcome Variable Participation Full-time Part-time Panel A: first year of motherhood Policy -0.058 (0.006) -0.067 (0.049) -0.022 (0.040) -0.033 (0.043) -0.041 (0.031) -0.038 (0.037) Employment before birth 0.108*** (0.021) 0.108*** (0.020) 0.064* (0.025) 0.066*** (0.025) 0.047*** (0.005) 0.047*** (0.018) Policy*employment before birth 0.042 (0.047) 0.047 (0.052) 0.026 (0.043) 0.032 (0.046) 0.021 (0.038) 0.020 (0.041) Childcare formal 0.065 (0.155) 0.113* (0.064) -0.024 (0.123) Rho 0.513*** (0.046) 0.378 (0.344) 0.380*** (0.068) -0.126 (0.288) 0.348*** (0.083) -0.432 (0.432) Observations 1,748 1,748 1,748 1,748 1,748 1,748 Panel B: second year of motherhood Policy 0.199*** (0.062) 0.080 (0.061) 0.082*** (0.026) 0.064** (0.027) 0.086 (0.068) 0.033 (0.059) Employment before birth 0.134*** (0.015) 0.063** (0.032) 0.077*** (0.029) 0.069*** (0.025) 0.070** (0.011) 0.046 (0.007) Policy*employment before birth -0.041 (0.071) -0.018 (0.055) -0.070** (0.039) -0.068* (0.039) -0.0005 (0.075) 0.007 (0.070) Childcare formal 0.409*** (0.045) 0.111*** (0.105) 0.258*** (0.069) Rho 0.589*** (0.055) -0.526 (0.274) 0.366*** (0.118) -0.229 (0.440) 0.403*** (0.041) -0.239 (0.202) Observations 1,532 1,532 1,532 1,532 1,532 1,532 Panel C: third year of motherhood Policy -0.038 (0.010) -0.032 (0.067) 0.004 (0.038) -0.006 (0.044) -0.025 (0.072) -0.043 (0.056) Employment before birth 0.135*** (0.034) 0.048** (0.029) 0.026 (0.023) 0.018 (0.027) 0.110*** (0.026) 0.040 (0.035) Policy*employment before birth 0.068 (0.102) 0.068 (0.077) -0.024 (0.041) -0.024 (0.047) 0.054 (0.060) 0.039 (0.043) Childcare formal 0.447*** (0.027) 0.101 (0.099) 0.414*** (0.055) Rho 0.370*** (0.078) -0.938* (0.124) 0.446*** (0.122) -0.223 (0.476) 0.241*** (0.066) -0.909** (0.118) Observations 933 933 933 933 933 933 Controls Yes Yes Yes Yes Yes Yes Notes: Standard errors in parentheses. *p < 0.10, **p < 0.05, and ***p < 0.01. Page 22 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 that the effect is not driven by the mothers who were employed before the childbirth. The access to public child care remains an important factor in participation decision, but the impact value drops significantly. These results partially confirm the empirical findings of previous childcare studies for Germany. For example, Müller and Wrohlich (2018) found that an increase in childcare slots by 1% point increases mothers’ participation by 0.2% points. At the same time, I did not find significant positive policy effects on the labor force participation during the second year, which is in line with Kluve and Tamm (2013), but contradicts the results of Geyer et al. (2015). Interestingly, the positive significant effect of the reform on the mothers with full-time employment is not reported before. The last year of job-protected parental leave is unpaid for both preand postreform groups. The interaction of policy variables with employment status prior to birth still remains insignificant, and the overall policy indicator reports an insignificant negative impact. The impact of other variables discussed earlier remains significant. The possibility of formal child care remains an important factor on the mother’s employment decision, and the employment before childbirth increases the probability to work afterward. Table 6 reports strongly significantly very robust positive associations between childcare coverage rates and the probability to use formal child care. The decision to use formal child care depends positively on the access to public childcare institutions and negatively on the possibility of using a partner and relatives as childcare substitutes. A higher current household labor income has a positive association with the childcare decision. Summing up the preliminary analysis of the 2007 parental leave reform shows a positive association of the policy shift with the labor force participation decision. However, after controlling for childcare decision, the effect remains only for the mothers working full-time during the second year of the motherhood. These findings suggest that if we consider labor force participation and childcare decisions jointly, controlling for a mother’s possibility of finding a substitute for her child care, the 2007 reform does not seem to produce a significant impact on employment behavior during the job-protected parental leave period. The reform design suggests that the new financial scheme of parental leave payment might give low-income females different incentives to return to work. I considered a subsample of women with below-median income (18,000 euros per year). Table 7 reports the reform effect on the marginal probability of employment for low-income mothers. There is no evidence of any significant policy effect on mothers’ participation decision during the first year of parental leave. The estimation results indicate that the employment prior to birth and the availability of formal child care increase the probability of being employed during the first year of Table 6 Average marginal effects for maternal childcare decision Leave year Participation Full-time Part-time First year 0.315*** (0.077) 0.321*** (0.071) 0.257*** (0.062) 0.269*** (0.060) 0.265*** (0.061) 0.262*** (0.065) Second year 1.066*** (0.244) 1.010*** (0.253) 1.295*** (0.244) 1.267*** (0.252) 1.233*** (0.233) 1.287*** (0.236) Third year 0.600 (0.489) 0.626* (0.343) 1.125*** (0.146) 0.832** (0.411) 0.798** (0.381) 0.587 (0.381) Notes: Standard errors in parentheses. *p < 0.10, **p < 0.05, and ***p < 0.01. Page 23 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 motherhood. The estimates presented in Panel B of Table 7 indicate that the reform had no effect on the mothers’ employment for the second year. Only for full-time employment, the policy had some positive and significant effect at the 10% level. Table 8 provides evidence that suggests different behavior patterns for high-income mothers. They report strong reductions in both part-time and full-time employment during the first year of motherhood. This sizable negative effect becomes insignificant for the mothers who were employed prior to childbirth. This finding suggests that career-oriented mothers did not change their employment behavior after the reform. Interestingly, the factor of formal childcare availability becomes insignificant for the employment decision. One possible explanation is that high-income mothers can afford the private childcare. For the second year of motherhood, the behavioral pattern changes: the probability to work increases significantly for the postreform mothers; the childcare availability increases the possibility of labor participation. The effect is driven by part-time employed mothers. I checked whether the baseline estimation results are robust to the different income definitions, such as monthly gross and net wages of the mother before giving birth, her partner’s Table 7 Average marginal effects for maternal childcare decision: low-income category Decision Participation Full-time Part-time Employment decision Panel A: first year of motherhood Policy 0.016 (0.035) 0.025 (0.048) 0.006 (0.022) 0.003 (0.024) 0.005 (0.032) 0.010 (0.043) Employment before birth 0.098*** (0.024) 0.100*** (0.024) 0.040*** (0.015) 0.039*** (0.014) 0.044* (0.024) 0.041 (0.031) Policy*employment before birth 0.001 (0.055) -0.003 (0.054) -0.002 (0.020) -0.0004 (0.020) 0.007 (0.050) 0.029 (0.065) Childcare formal -0.082 (0.207) 0.032 (0.068) -0.195 (0.370) Childcare decision Childcare coverage 0.324*** (0.048) 0.314*** (0.064) 0.303*** (0.056) 0.308*** (0.065) 0.326*** (0.057) 0.196** (0.082) Rho 0.482*** (0.139) 0.687 (0.555) 0.259* (0.135) 0.043 (0.450) 0.406** (0.169) 0.652 (0.432) Observations 909 909 909 909 909 909 Employment decision Panel B: second year of motherhood Policy 0.079*** (0.072) 0.015 (0.064) 0.053*** (0.026) 0.038* (0.023) 0.008 (0.060) -0.017 (0.060) Employment before birth 0.042 (0.039) -0.017 (0.036) 0.041* (0.023) 0.024 (0.018) 0.011 (0.033) -0.009 (0.036) Policy*employment before birth 0.054 (0.092) 0.062 (0.080) -0.008 (0.032) -0.006 (0.035) 0.010 (0.077) 0.013 (0.103) Childcare formal 0.395*** (0.062) 0.180 (0.124) 0.182* (0.103) Childcare decision Childcare coverage 0.672*** (0.244) 0.513** (0.240) 1.023*** (0.178) 0.865*** (0.215) 0.993*** (0.169) 0.985*** (0.180) Rho 0.436*** (0.107) -0.830 (0.246) 0.353** (0.139) -0.743 (0.467) 0.167** (0.079) -0.405 (0.231) Observations 797 797 797 797 797 797 Notes: Standard errors in parentheses. *p < 0.10, **p < 0.05, and ***p < 0.01. Page 24 of 27 Chirkova. IZA Journal of Labor Policy (2019) 9:7 monthly wage, and household income. I also used working hours instead of labor income. Controlling for the different categories, I found the same pattern in terms of policy effect on the probability of being employed: it remains insignificant. I considered two age brackets: 20–30 and 30–40.31 The results are robust to the age specification; the findings still confirm the insignificant impact of the policy on the maternal employment probability. 5 Conclusions This paper studies the impact of the 2007 reform in the German parental leave system on maternal employment. The postreform universal system of benefits is more generous in terms of the amount of the benefit, but it pays for a shorter period of time. I estimated the shortrun effects of the policy focusing on the maternal employment during the first three years of the motherhood. The mothers are protected from the job dismissal during this period. The 31 The sample size does not allow the lower intervals. Table 8 Average marginal effects for maternal childcare decision: high-income category Decision Participation Full-time Part-time Employment decision Panel A: first year of motherhood Policy -0.340*** (0.067) -0.330* (0.183) -0.802*** (0.138) -0.732*** (0.101) -0.191*** (0.042) -0.194*** (0.045) Employment before birth -0.017 (0.070) 0.028 (0.068) 0.041 (0.059) 0.055 (0.069) -0.14 (0.054) 0.012 (0.048) Policy*employment before birth 0.280*** (0.070) 0.269 (0.180) 0.806*** (0.142) 0.716*** (0.106) 0.143 (0.045) 0.149*** (0.049) Childcare formal 0.503*** (0.191) 0.123 (0.279) 0.262** (0.134) Childcare decision Childcare coverage 0.210 (0.161) 0.283* (0.152) 0.195 (0.136) 0.191 (0.152) 0.154 (0.146) 0.196** (0.082) Rho 0.597*** (0.071) 0.615 (0.751) 0.496*** (0.105) 0.055 (1.079) 0.372** (0.146) -0.479 (0.417) Observations 839 839 839 839 839 839 Employment decision Panel B: second year of motherhood Policy 0.501*** (0.139) 0.393** (0.197) 0.627*** (0.216) 0.030 (0.076) 0.258** (0.112) 0.194** (0.096) Employment before birth 0.327** (0.142) 0.339** (0.157) 0.614*** (0.215) 0.035 (0.078) 0.172 (0.114) 0.187* (0.108) Policy*employment before birth -0.357*** (0.120) -0.322** (0.132) -0.649*** (0.219) -0.074 (0.093) -0.142 (0.103) -0.123 (0.088) Childcare formal 0.379** (0.148) 0.222*** (0.39) 0.261 (0.179) Childcare decision Childcare coverage 1.138*** (0.334) 1.447*** (0.356) 1.397*** (0.341) 1.457*** (0.068) 1.276*** (0.325) 1.420*** (0.332) Rho 0.697*** (0.035) -0.013 (0.597) 0.382** (0.163) -0.743 (0.467) 0.526** (0.056) 0.023 (0.437) Observations 735 735 735 735 735 735 Notes: Standard errors in parentheses. *p < 0.10, **p < 0.05, and ***p < 0.01.