Policy Research Working Paper 10492 Filling the Gaps Childcare Laws for Women’s Economic Empowerment S Anukriti Lelys Dinarte-Diaz Marina Elefante Maria Montoya-Aguirre Alena Sakhonchik Development Economics Development Research Group June 2023 Policy Research Working Paper 10492 Abstract This paper aims to provide global evidence on whether and by 2 percent, on average. Moreover, the effect increases over what attributes of laws governing the provision of childcare time, reaching up to 4 percent five years after an enactment. services affect women’s labor market outcomes. It merges This effect is driven by women who are married, have com- country-year-level data from the World Bank’s Women, pleted less than primary education, and are between the Business and the Law database, which documents childcare ages of 35 and 44. Lastly, regulation of the availability and laws across countries, with data on women’s labor force par- affordability of childcare has a similar impact on female ticipation from ILOSTAT. Using a difference-in-difference labor force participation, whereas the effect of quality reg- estimation framework, the analysis finds that the enactment ulation is smaller. of childcare laws increases women’s labor force participation This paper is a product of the Development Research Group, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at sanukriti@worldbank.org and ldinartediaz@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Filling the Gaps: Childcare Laws for Women’s Economic Empowerment* S Anukriti† Lelys Dinarte-Diaz‡ Marina Elefante§ Maria Montoya-Aguirre¶ Alena Sakhonchik|| Keywords: Childcare Laws, Female Labor Force Participation, Women’s Empowerment. JEL Classification: J16, J31, J71, J78, K31, K38. * We are thankful to Kathleen Beegle, Tea Trumbic, Norman Loayza, and seminar participants at the World Bank Indicators Group Seminar for their useful comments and feedback. Funding for this work was provided via the World Bank’s Knowledge for Change Program and the Multi-Donor Trust Fund for Jobs. The authors have no conflicts of interest to report. The findings, interpretations, and conclusions expressed in this report are entirely those of the authors. They do not necessarily represent the views of the World Bank and its affiliated organizations, its Executive Directors, or the governments they represent. † The World Bank. Email: anukriti@worldbank.org ‡ The World Bank. Email: ldinartediaz@worldbank.org § The World Bank. Email: melefante@worldbank.org ¶ Paris School of Economics. Email: maria.montoya@psemai.eu || The World Bank. Email: asakhonchik@worldbank.org 1 Introduction The current global female labor force participation (FLFP) rate for women is roughly 47 percent, which is 25 percentage points (pp) lower than the participation rate for men (ILO- STAT). In some regions, the gender gap in LFP is as high as 50 pp. Gender inequality in labor market outcomes is problematic for several reasons. Not only does it imply a mis- allocation of talent, which has negative consequences for GDP (Hsieh et al., 2019), it also disempowers women within their families and in society because paid employment typi- cally confers more autonomy for women than unpaid labor within the household (Kabeer, 2008). Thus, improving FLFP is not only valuable per se, but can also offer instrumental returns in other dimensions of gender equality and economic development. Women face multiple barriers, such as lower educational attainment than men and gender discrimination in the labor market, that translate into limited access to the labor market, fewer job opportunities, lower returns to work experience, and pay gaps (Frut- tero et al., 2020). In this paper, we focus on the role of childcare responsibilities as a constraint to women’s participation in market work. In almost all countries, women are designated as the primary caregivers for their households and are more likely than men to stay at home, work part-time, or pause their careers in order to provide care (Staab, 2015). Women account for about three-quarters of the total time spent in unpaid care work glob- ally. In fact, women spend more time than men on unpaid care work activities in almost all countries, irrespective of a country’s level of development. The gendered division of household care responsibilities has significant negative implications for women’s labor 1 market outcomes.1 In fact, countries with greater gender gaps in unpaid care responsibil- ities tend to have lower FLFP. The COVID-19 pandemic has brought the issue of childcare and its impact on FLFP to the forefront of policy discussions. Between 2019 and 2020, global female employment declined by 4.2 percent compared to a 3 percent decrease in male employment (ILO, 2021). Women were not only over-represented in the hardest hit sectors, but also bore the brunt of the lockdowns with additional childcare and housework responsibilities, worsening the gender gaps in labor market outcomes. As the world emerges from the pandemic and policymakers think about tackling women’s increased domestic care burden, access to affordable childcare for families, especially working mothers, has been gaining momentum. An extensive literature has shown that access to affordable childcare services is a key component in improving maternal labor market outcomes. Olivetti and Petrongolo (2017) conduct a systematic review of both macro- and micro-level evidence from high-income countries on the impacts of family policies and finds that overall spending on early childcare is strongly associated with better labor market outcomes for women.2 Empirical evidence on the benefits of access to childcare for maternal labor market outcomes in lower- and middle-income countries (LMICs) is also growing and aligns with the findings from developed countries. Halim et al. (2023) present a summary of 22 causal studies from LMICs that rigorously test if an 1 Virginia et al. (2015) show that discriminatory gender norms related to the provision of care decrease FLFP by 12 percent. Conversely, addressing primary caregivers’ childcare needs could lead to a 10 pp increase in the LFP rate in countries such as Kenya, an 8 pp increase in Nigeria and South Africa, and a 5 pp increase in India. 2 See Givord and Marbot (2015), Bettendorf et al. (2015), Geyer et al. (2015), Nollenberger and Rodríguez- Planas (2015), and Herbst and Barnow (2008) for evidence from France, the Netherlands, Germany, Spain, and the United States, respectively. 2 increase in access to childcare improves maternal labor force outcomes.3 In this paper, we specifically examine whether enactment of laws governing the provi- sion of childcare in a country influences women’s labor market outcomes in that country using a difference-in-difference estimation strategy. We utilize the World Bank’s Women, Business, and the Law (WBL) database to obtain information on the dates of enactment and commencement of childcare laws and other indicators related to accessibility, afford- ability, and quality of childcare services across countries. We merge this data with panel data on women’s labor market outcomes from ILOSTAT and other socio-demographic data from the Our World In Data database and the Barro-Lee Educational Attainment dataset. Our estimation strategy relies upon temporal and country-level variation in the presence of childcare laws to causally estimate their effect on FLFP. The effect of childcare laws on women’s participation in market work is, a priori, un- clear. On the one hand, availability of childcare services can enable women to reallocate their time from unpaid care activities at home to paid market work, increase working hours, productivity, wages and income, and influence the type of employment. Women may also take advantage of work opportunities in the childcare industry to increase their LFP. On the other hand, women’s willingness to do so will depend on the cost of child- care services relative to their potential labor market income, which in turn depends on the structure of the economy and local labor market conditions, among other things. In a 2016 household survey in the European Union (EU), over 40 percent of families cited 3 See Ajayi et al. (2022), Sanfelice (2018), Calderón (2014), Dang et al. (2022), and Clark et al. (2019) for evidence from Burkina Faso, Brazil, Mexico, Vietnam, and Kenya, respectively. 3 cost as the main barrier to using formal childcare (European Commission 2016).4 The average net childcare costs amount to 17 percent of middle-income two-earner couples’ median full-time earnings (OECD 2021). Relatively less information is available on the costs of childcare in LMICs, but existing studies suggest that childcare expenses are sig- nificant relative to families’ earnings in many countries.5 Clark et al. (2019) show that high childcare costs discourage households in Kenya from using formal childcare facili- ties, which negatively affects mothers’ participation in the labor market. Similar findings exist for Mozambique and Liberia (Bhatkal, 2014; Lokshin et al., 2000). Conversely, in rural Colombia, small monthly fees for community-based childcare centers more than tripled women’s employment and increased their work hours by 75 hours per month (Attanasio and Vera-Hernandez, 2004). The willingness to use formal childcare services and the potential for childcare regu- lation to impact FLFP also depend on the strength of social norms related to gender roles and women’s work, and on the quality of childcare. According to the latest wave of the World Value Survey, 41 percent of respondents globally agree or strongly agree with the statement that “A pre-school child suffers with a working mother” and 51 percent agree or strongly agree with the statement that “On the whole, family life suffers when a woman works full time.” Although existing studies on the impact of childcare quality on women’s 4 Another study from 27 EU member states highlights that approximately 25 percent of women who do not work or work part-time report that childcare is unaffordable (Mills et al., 2014). 5 For example, a survey conducted in Edo state, Nigeria, in 2005 found that the average cost of a private childcare center kindergarten for one child was more than half the salary of someone earning the minimum wage (Olubor, 2009). A study of peri-urban areas in Accra (Ghana), Johannesburg (South Africa), Lagos (Nigeria), and Nairobi (Kenya) found that the average cost of a preschool child was one-quarter to nearly half of a person’s monthly spending (Bidwell et al., 2014). In contrast, Latin American countries with broad public sector involvement in childcare have programs that are more affordable. 4 labor market outcomes are mostly qualitative in nature, they suggest that low quality dis- courages women from enrolling their children in childcare services (Halim et al., 2023; Zaid and Amin, 2021). Additionally, building parents’ trust that children will be in a safe and stimulating environment could be a challenge, as regulations around quality of for- mal childcare arrangements are often either absent or insufficient to ensure child safety and inspire parent confidence (Devercelli and Beaton-Day, 2020). We find that the enactment of childcare laws increases FLFP by 2 percent, on aver- age. Moreover, the effect increases over time, reaching up to 4 percent five years after an enactment. This effect is driven by women who are married, have completed less than primary education, and are between the ages of 35 and 44 – these are subgroups for which a childcare law is more likely to be relevant and that are more “treatable” in terms of their LFP rate in the absence of the law. Lastly, the regulation of availability and affordability of childcare has a similar impact on FLFP, whereas the effect of quality regulation is smaller. This paper contributes to the literature in two respects. First, while most of the existing studies focuse on one country at a time and examine small-scale childcare programs, this paper is global in nature and evaluates the effect of childcare provision in formal center- based settings in countries spanning seven regions of the world. Second, we examine the associations between different pillars of childcare laws (availability, affordability, and quality) and women’s labor market outcomes. The literature typically focuses on the im- pact of a single determinant, and no prior studies have been conducted around the impact of multiple determinants. The remainder of the paper is organized as follows. Section 2 describes our data 5 sources. Section 3 presents the empirical approach used to estimate the effects of childcare laws on FLFP. Section 4 presents the main results and is followed by robustness checks in Section 5. Section 6 concludes the paper. 2 Data Our analysis is based on data from 80 countries spanning a 20-year period (2000-2020). We combine information on childcare regulations, labor statistics, and socio-demographic indicators from the following sources to construct a country-year panel for our analysis. 2.1 Data sources Childcare regulations. The World Bank’s WBL database collects information on the legal en- vironment pertaining to women’s economic participation and opportunity in 95 countries (Hyland et al., 2020). In particular, the dataset covers laws and regulations governing the supply and parental demand for childcare services. The dataset includes information on the dates of enactment and commencement of a law and other indicators related to ac- cessibility, affordability, and quality of childcare services applicable to children below the formal pre-primary school starting age, as this is where the largest legal policy gap exists. The WBL database maps current legal and regulatory measures adopted by each country to ensure or strengthen the following three pillars of childcare (see list of indicators within each pillar in Table A1): • Availability: The childcare law expands access to childcare by supporting different types of childcare provision and its convenience. 6 • Affordability: The childcare regulation improves childcare service provision espe- cially for low-income or vulnerable families through government support (financial or non-financial) to parents, private childcare centers, employers, or cost-regulating structures. • Quality: The childcare regulation ensures a safe environment for children, con- tributes to healthier nutrition and school readiness, and promotes uptake. Data collection builds on established WBL methodology and expertise and relies on desk research through the review and assessment of official laws and regulations as well as consultations with WBL experts in labor legislation. Both federal and local legislation applicable to a country’s main business city are considered. For federal systems, where provision of childcare is not established at the national level, legislation applicable to the main business city is explored. Non-binding documents and instruments – typically re- ferred to as policy notes, national strategies, guidelines, recommendations, declarations, and opinions – are not considered for the purposes of data collection. Official ministe- rial websites specifying or explaining certain regulatory aspects covered within the estab- lished three-pillar framework are cited in limited circumstances. To measure the separate effect of childcare availability, affordability, and quality on FLFP, we construct a score for each pillar as a simple average over all variables included in each pillar, each of which is either a dummy variable (e.g., “Does the law mandate employers to provide or support childcare services?”) or is normalized between 0 and 1 (e.g., “What are the mandated minimum hours of operation of public childcare centers if regulated in the law?”). The score takes a value of 0 before the enactment of the law and 7 also when the law does not include any of the elements that are tracked in that pillar. A score equal to one means that the country’s law has been enacted and includes all possible elements that are tracked for that pillar. We then standardize each score across the full sample. Labor Market Outcomes We use ILOSTAT indicators on FLFP rates and other labor mar- ket indicators reported by the International Labor Organization. The data is available on a yearly basis for 80 of the 95 countries covered in the WBL dataset but its time coverage varies across countries, resulting in an unbalanced panel. The LFP rate is reported for 10-year age groups as well as by two aggregate categories of marital status where women are categorized as single or married.6 Our sample comprises women aged 25 or above. Socio-demographic variables. Our analysis also requires country-year-level information on marriage rates and educational attainment. We obtain data on marriage rate per 1000 people and women’s average age at marriage from the Our World In Data database. The marriage rate corresponds to the ratio of the number of marriages during a given year to total population. The average age at marriage for women corresponds to the mean age at first marriage for OECD countries. For other countries, figures are estimated indirectly from the proportions of single women by age. The education data comes from the 2021 update of the Barro-Lee Educational Attain- ment dataset (Barro and Lee, 2013), which covers the years between 1995 and 2015. We obtain the share of women with primary education, secondary education, and tertiary 6 Women in a civil union or cohabiting women are categorized as married; and widowed or divorced women are considered single. 8 education by 10-year-age bands. These variables are only available for the years that cor- respond to a population census. Thus, for the years with missing information, we assign the value of the previous closest year. 2.2 Descriptive Statistics Table A2 summarizes the characteristics of the resulting sample with and without con- trols, which differ in size due to the availability of marriage and education variables. The sample without controls, which results from merging the WBL dataset with the labor indi- cators from ILOSTAT, covers 80 countries and is an unbalanced panel of 918 country-year observations, where each country is observed for 11.5 years on average. Among these countries, 79 percent had enacted a law regulating childcare by 2022, and we observe 60 percent of them both before and after the law enactment in our data. The full estima- tion sample, which includes controls related to marriage and education, has 54 countries and 760 country-year observations. This smaller panel is more balanced, however, and countries appear, on average, for 14 years in the panel. Tables A3-A5 detail the countries included and show some characteristics of the sample by country. As explained in the next section, we also construct cohort-specific datasets where we include countries that enact a childcare law in a given year and its appropriate control units. The resulting stacked dataset has 11,899 and 9,497 country-year-stack observa- tions for the baseline (without socio-demographic controls) and the full specification (with socio-demographic controls), respectively. Summary statistics of key variables used in our analysis are reported in Table 1. The 9 FLFP rate ranges from 14 percent to 93 percent, with an average of 54 percent. For roughly half (55 percent) of the country-year observations in our data, a childcare law has been enacted, with the year of enactment ranging from 1976 to 2021 and commencing one year later on average. The average age at first marriage for women is 25.7 years. The ratio of marriages per 1,000 people ranges between 2.2 and 15.7, and is 5.8 on average. The mean share of women with primary, secondary, and tertiary education is 15.9 percent, 48.2 percent, and 29.2 percent, respectively. 3 Estimation Strategy A simple comparison of countries where a childcare law has been enacted with those with- out such a law will yield biased estimates of the effects of childcare laws because countries that enact such laws may be different in terms of unobservables relevant for FLFP from those that do not. For example, if countries that have more progressive gender norms are more likely to pass childcare legislation, then we would expect to see differences in their FLFP rates even in the absence of such laws. Therefore, to causally identify the effects of childcare laws on women’s labor market outcomes, we leverage the staggered enactment of childcare laws across countries and use a difference-in-difference estimation strategy. This approach compares the before-after differences in women’s labor market outcomes between countries where a childcare law has been enacted and those where such a law has not been enacted yet. Specifically, we estimate the following two-way fixed-effect (TWFE) 10 model for a country c in year t: Yct = α + βP ostct + γc + θt + πc t + Xct + ϵct (1) where Yct corresponds to the outcome of interest, e.g., FLFP rate. P ostct is an indica- tor that equals one if t >= the year of childcare law enactment in country c; if a country has never enacted a childcare law during our study period, we define P ostct to be equal to zero throughout. The country and year fixed effects are denoted by γc and θt , respec- tively. Lastly, πc t denotes country-specific linear time trends and Xct represents a vector of country-year-level controls that includes the population marriage rate, average age at marriage for women, and the share of women with primary, secondary, and tertiary edu- cation. We use standard errors clustered at the country level for inference. The advantage of using TWFE is that we can rule out two concerns related to causal identification of treatment effects. The first concern is that the results may be driven by time-invariant differences in FLFP across countries, such as time-invariant social norms about participation of women in the labor market. By including country fixed effects, we can rule out such concerns. The second concern is that women’s labor market outcomes in countries that are included in our sample may evolve over time in a way that is common across countries. For example, trends in FLFP are likely to be correlated with trends in women’s educational attainment and fertility, or macroeconomic shocks that are common across countries. The year fixed effects help us rule out this latter concern. In a standard difference-in-difference model, the key identifying assumption is that the average outcome among the treated and comparison populations would have followed 11 “parallel trends” in the absence of treatment, conditional on Xct . In our case, this implies that, in the absence of the enactment of the childcare law, women’s labor market outcomes for different countries in our sample would have evolved along parallel paths. We relax this assumption with the inclusion of country-specific linear time trends, that capture pre- existing linear differences in trends across countries. In addition, we explicitly test the assumption of parallel trends, as well as estimate the dynamic effects of childcare laws, by estimating an event-study version of equation (1) using the following specification: 5 Yct = α + βk P ostc,t+k + γc + θt + πc t + Xct + ϵct (2) k=−4 where k = 0 is the year of enactment and P ostc,t+k equals one if k >= 0. The βk coefficients capture the evolution of women’s labor market outcomes before and after the enactment of the childcare laws over a 10-year period in countries where the law has been enacted relative to countries where the law has never been enacted. For countries where a childcare law is never enacted during our study period, there is no unique reference year that can be used to split the sample period into pre- and post-treatment years since the enactment year varies across “treated” countries. Therefore, for never-treated countries, we follow the standard approach and define P ostc,t+k = 1 for k = −1 , and 0 for all other t so that βk can be interpreted as the effect of the childcare law t years from its enactment. Although TWFE regressions similar to equation (1) are commonly used for staggered research designs, they deliver consistent estimates only under relatively strong assump- tions about homogeneity in treatment effects. However, if treatment effects of the child- care law are heterogeneous across countries or time, the TWFE estimation does not deliver 12 consistent estimates of the true average treatment effect (Goodman-Bacon, 2021). There- fore, we estimate (1) on a stacked dataset as in Cengiz et al. (2019) to alleviate concerns related to the staggered nature of our treatment. We create a specific dataset for each treat- ment cohort (defined as a group of countries that enacted their childcare law in a given year), which includes all observations for those countries after the law enactment, and the appropriate controls, i.e., never-treated units, and not-yet treated units. This means that for a given treatment cohort k the dataset includes all periods for the countries that enact the law in year k , only the periods before law enactment for countries that get the law in a year different from k , and all periods for countries that never enact a childcare law in our period of analysis. Then, we stack the cohort-specific datasets and estimate equation (1) with country-cohort and year-cohort fixed effects. Sun and Abraham (2021) show that when dynamic treatment effects are heterogeneous across adoption cohorts, the coefficients from equation (2) become difficult to interpret. Thus, for example, problems may arise if the average treatment effect in the third year after adoption of the law is different for countries that adopted the childcare law in 2010 as it is for countries that adopted in 2015. Therefore, for robustness checks, we also ex- amine alternative estimators presented in Callaway and Sant’Anna (2021) and Sun and Abraham (2021), that deliver consistent estimates even in the presence of heterogeneity in treatment effects. Additionally, we present an estimator proposed by De Chaisemartin and d’Haultfoeuille (2020), though given the unbalanced nature of our panel, these esti- mates are considerably noisier. 13 4 Results We first present our estimates for the effect of childcare laws on FLFP and then analyse if the treatment effects differ by women’s age, education level, and marital status. 4.1 Main Results We begin our analysis by estimating the effect of childcare law enactment on FLFP us- ing our main specification (equation 1). In Table 2, column (1) presents estimates of β from our baseline specification, which includes only the childcare law enactment indica- tor and country and year fixed effects. Column (2) controls for a vector of country-year characteristics (Xit ) that includes the population marriage rate, average age at marriage for women, and the share of women with primary, secondary, and tertiary education. In column (3), we also include country-specific linear time trends. In columns (4) - (6), we use the same specification as in columns (1) - (3), but we replace the indicator for childcare law enactment with an indicator for childcare law commencement (i.e., the period when the law becomes effective).7 Our preferred estimates are those presented in column (3) for childcare law enactment and in column (6) for childcare law commencement. In Panel A of Table 2, the estimated effect size of childcare law enactment or com- mencement on the FLFP rate in our preferred specification is approximately 1 pp or be- tween 1.75 and 1.9 percent relative to the average FLFP rate for countries where the child- 7 As we show in Table 1, our sample includes countries where a childcare law was enacted or became effective between 1976 and 2021. For the average country in our sample, the childcare law was enacted in 2007 and it commenced a year later, in 2008. 14 care law was not enacted or effective, i.e., for those with P ostct = 0. To understand the magnitude of these estimated effects, we compare them with the estimated effects of child- care interventions implemented in LMICs on mothers’ labor market outcomes. We rely on a recent evidence review of 22 studies conducted by Halim et al. (2023). Out of these stud- ies, half evaluate the impacts of childcare services targeted to children aged three years or younger on women’s labor market outcomes. In sum, the effects of childcare services on FLFP are between 11 percent to 29 percent in urban China (Du and Dong, 2013; Du et al., 2019) and 37 percent in Colombia (Attanasio and Vera-Hernandez, 2004).8 Thus, our esti- mated effects of childcare laws on FLFP are one-fifth of the estimated effects of childcare services in these studies. This smaller effect of the childcare laws, relative to the provision of childcare services, is not surprising as the regulation of childcare precedes the actual provision of these services in most contexts. Moreover, enactment of legislation does not always translate into effective implementation. In order to test for parallel trends and to study the dynamic effect of the childcare law enactment on FLFP rate, we estimate an event-study version of the TWFE model with indicators for different periods before and after the enactment of the law. The estimates from specification (2) are presented in Figure 1. First, the estimates are consistent with the parallel trend assumption: the coefficients for the years prior to the enactment of the childcare law in each country are all close to zero and exhibit no discernible differential pre-trends. Second, the treatment effect of childcare laws on FLFP increases over time. In 8 One study finds no effects of public daycare on women’s labor market outcomes in Chile (Medrano Vera, 2009). Other studies included in the review find effects of childcare services on women’s employment that range from 10 percent (Barros et al., 2013) to 22 percent (Rosero and Oosterbeek, 2011). 15 fact, the effects are close to zero during the year of enactment (period 0) and one year after enactment (period 1), likely because the law becomes effective, on average, 1 year after enactment in our sample countries. Then, starting in the second year after the enactment, the estimated effect on FLFP is close to 1 pp and increases up to 2.2 pp (or 4 percent relative to the reference group, P ostct = 0) five years after the enactment. This increase in treatment effect over time can be explained by the fact that once the law has been enacted, its adoption by firms and its dissemination to women may not take place immediately but rather in a gradual manner. Considering the findings in Figure 1, we estimate specification (1) to produce two ad- ditional sets of results: a) excluding the year of enactment (or commencement) and b) excluding both the year of and one year after enactment. These results are presented in Panels B and C of Table 2. As expected, the magnitude of the estimated effects is larger in Panels B and C than in Panel A. If we exclude the enactment year or the commencement year, the estimated effect of the childcare law on FLFP is 1.27 pp and 1.64 pp, respectively. Similarly, if we exclude both the year of law enactment (or commencement) and the sub- sequent period, then the estimated effect is 1.67 pp and 1.30 pp, respectively. These results suggest that the enactment of the childcare law may have increased parents’ willingness to use and utilization of formal childcare services for their children, which in turn increased the LFP of women, who tend to be the primary caregivers of children. 16 4.2 Heterogeneity Next we examine whether the enactment of childcare laws had differential effects on FLFP by women’s marital status, age, and education level. First, to study differences by marital status, we estimate specification (1) separately for married and single women. These re- sults are presented in Figure 2. The estimated effect of childcare law enactment on FLFP of married women (2.2 pp) is larger than the effect on single women (1.7 pp). This is not surprising given that married women experience a heavier childcare burden, on average, than single women, making the childcare law more relevant for the former.9 Moreover, on average, the FLFP rate of single women (66.27 percent) is already higher than that for married women (54.74 percent) even in the absence of the law, making them less “treat- able” by the law. These findings are consistent with previous literature on the differential effects of childcare policies on FLFP by women’s marital status. For example, Bick (2016) examine women who are married (or in a long-term relationship) in West Germany, and find that increasing the provision of subsidized childcare increases the maternal LFP rate. Similar results are found by Baker et al. (2008) on the effect of subsidized childcare services on the labor supply of married (and cohabiting) women. We also explore heterogeneity by age of women. We present three main results from this analysis in Figure 3. First, the effects are smaller and statistically insignificant for younger women (25-34 years old). Second, the aggregate effects appear to be driven by women aged 35-44. Lastly, although the effects on women aged 45 years or older are sta- 9 The average number of children is typically higher for married women relative to single women. 17 tistically insignificant, the magnitude of the estimated effect is similar to that for women aged between 35 and 44 years. These results can be explained by several factors. First, younger women are more likely to already be in the labor force, which makes them less “treatable” compared to older women. Moreover, women between 35 and 44 years of age are more likely to have young children, making the need for childcare services more relevant to them. Simultaneously, women older than 45 years of age are more likely to be grandmothers who offer their working offspring a free, flexible, and reliable source of childcare. Once the law is enacted, their assistance with young children is no longer required, which may lead them to decide to participate in the labor force. Lastly, we also test for heterogeneity by education level. We estimate the childcare enactment effects separating the sample by level of education completed. We create four groups: less than primary education, primary education, secondary education, and ter- tiary education. We present the results in Figure 4. We find that the effects of the enact- ment of childcare regulation on FLFP are driven by women with less than primary edu- cation, that is, women that are more likely to work in low-skilled jobs, where childcare services are not available. Studies from higher-income countries suggest that low-skilled women (usually poorer) often are most responsive to changes in costs and availability of childcare (Morrissey, 2017). However, impacts on maternal labor outcomes from childcare provision in LMICs are not concentrated among the most disadvantaged (Halim et al., 2023). A potential explanation could be the lack of regulation of access, affordability, and quality of childcare services in LMICs. In this sense, promoting regulation of childcare services can be an starting point to promote labor market outcomes of the most disadvan- 18 taged women in these countries. 4.3 Decomposing the Childcare Law Effect To this point, we have presented the positive effect of childcare enactment on FLFP rate. However, what specific element of the childcare attributes regulated in the law is driv- ing this positive impact? Existing evidence documenting the positive impact of childcare availability and affordability on women’s labor market outcomes is compelling and broad for both higher and lower-income countries. Moreover, although the literature highlights childcare quality as a key factor that determines parental decision to use childcare ser- vices, this has not been rigorously evaluated yet. Exploiting the information on childcare availability, affordability, and quality addressed by the childcare laws in the WBL dataset, we assess the contribution of each of these three pillars to women’s labor market outcomes. Specifically, we create standardized scores for each of the pillars and replace P ostct with these scores one by one in specification (1), which we estimate on the stacked dataset by law enactment cohort. The results on the impacts of each childcare legislation pillar on FLFP rate are pre- sented in Table 3. Columns (1) to (3) present the results including all years; columns (4) to (6) exclude the year of law enactment, and columns (7) to (9) exclude the year of enact- ment and the year after. Consistent with previous literature, the effect of the regulation of childcare availability on FLFP rate is positive and statistically significant across different samples, ranging from 0.53 pp (when we include all years) to 0.86 pp (when we exclude the year of law enactment and one year after). The estimated effects of regulating child- 19 care affordability (and quality) on FLFP rate are also positive but statistically significant only after excluding the year of enactment (and one period after). The magnitudes of the estimated effects of childcare availability and affordability are similar across samples, whereas the estimated impact of childcare quality is somewhat smaller. The variations in the estimated effects could stem from both preference and supply factors. Firstly, it is possible that legislation aligns with existing evidence, with an em- phasis on ensuring accessible and affordable childcare, while placing less emphasis on regulating quality. Secondly, quantifying quality is more challenging compared to mea- suring accessibility and affordability. As a result, the perceived quality of childcare may have a lesser impact on women’s decision to participate in the labor force compared to accessibility and affordability. 5 Robustness Checks We conduct additional exercises to probe the robustness of our estimates that include es- timating modified versions of specification (2) to test the dynamic effects of the child care enactment and conducting placebo tests. First, the fully dynamic version of the TWFE model estimated using OLS yields consistent estimates only under the strong assump- tion of treatment effect homogeneity (Roth et al., 2023). To allow for heterogeneity in treatment effects across time and treated units, we estimate the event study estimates fol- lowing Callaway and Sant’Anna (2021) and Sun and Abraham (2021), and stacked TWFE as in Cengiz et al. (2019). The Callaway and Sant’Anna (2021) approach provides sensible estimates even under arbitrary heterogeneity in treatment effects and it makes transpar- 20 ent exactly which units are being used as a control group to infer the unobserved potential outcomes. Moreover, Sun and Abraham (2021) accounts for heterogeneous dynamic treat- ment effects across enactment groups and imposes parallel trends only for groups that are eventually treated and not for the never-treated groups. These results are presented in Figure 5. Overall, we show that, independently of the estimation method used, the coeffi- cients for years prior to the enactment of the law are close to zero and do not exhibit any differential pre-trend. Moreover, all estimators yield treatment effects that increase over time in the post-law-enactment years. In our setting, stacked TWFE is more appropriate than the alternative new difference-in-difference estimators because the latter rely either on balanced panels or on unbalanced panels whose observations are evenly spaced out. In a second robustness check, we conduct a placebo test in two steps. First, we ran- domly assign the year of the law enactment for each country 50 times and create a new stacked dataset for each draw. Next, we use specification (1) to estimate the main results, excluding the year of enactment and the subsequent year. The results are presented in Fig- ure 6. We find that the positive effects of childcare law enactment on FLFP are statistically significant in only 2 our of 50 draws, boosting our confidence in our findings. 6 Conclusion Our findings demonstrate that the enactment of childcare laws increases women’s likeli- hood of participating in the labor force. Additionally, the impact on FLFP is similar when the law regulates access and affordability of childcare services, and higher compared to quality regulations. 21 Unlike previous studies that establish positive correlations between legal reforms and labor market outcomes, our study is the first to provide causal estimates of the impacts of childcare regulation on women’s economic outcomes in a cross-country setting. Another unique contribution of this paper is the quantification of the effects of each aspect of child- care laws, namely accessibility, affordability, and quality, on FLFP. However, our study also opens up opportunities for further inquiry to deepen our understanding of the influ- ence of laws on women’s economic opportunities. For instance, what specific mechanisms can explain our results? What factors contribute to the establishment of regulations aimed at promoting childcare services? Is it driven by increasing awareness, pressure from the international community, labor demands, or a combination of these factors? What are the effects of childcare laws on women’s employment rates and wages? Do firms respond by altering their recruitment and hiring policies, thereby introducing ambiguity into the final effect of the childcare law? Due to data limitations, we cannot explore these questions in our setting and leave them for further research. Despite well-established empirical evidence on the positive effects of access to afford- able childcare services for children’s development and women’s labor market participa- tion, a substantial gap remains between the demand and supply sides of the childcare market. A recent World Bank study revealed that 40 percent of all children (nearly 350 million) who are below primary-school-entry age worldwide need childcare but do not have access to it, with LMICs accounting for nearly 80 percent of the gap (Devercelli and Beaton-Day, 2020). Our results highlight the role played by legislative changes in improv- ing women’s and children’s access to childcare and other outcomes. 22 References Ajayi, K., Dao, A., and Koussoubé, E. (2022). The effects of childcare on women and children: Evidence from a randomized experiment in burkina faso. Forthcoming in Policy Research Working Paper Series, World Bank. Attanasio, O. and Vera-Hernandez, M. (2004). Medium-and long run effects of nutrition and child care: evaluation of a community nursery programme in rural colombia. Insti- tute for Fiscal Studies. Baker, M., Gruber, J., and Milligan, K. (2008). Universal child care, maternal labor supply, and family well-being. Journal of political Economy, 116(4):709–745. Barro, R. J. and Lee, J. W. (2013). A new data set of educational attainment in the world, 1950–2010. Journal of Development Economics, 104:184–198. Barros, R. P. d., Olinto, P., Lunde, T., and Caralho, M. (2013). The impact of access to free childcare on women’s labor market outcomes: evidence from a randomized trial in low-income neighborhoods of rio de janeiro. In World Bank Economists’ Forum. Bettendorf, L. J., Jongen, E. L., and Muller, P. (2015). Childcare subsidies and labour sup- ply—evidence from a large dutch reform. Labour Economics, 36:112–123. Bhatkal, T. (2014). What do women want? gender, perceptions data and development priorities. Bick, A. (2016). The quantitative role of child care for female labor force participation and fertility. Journal of the European Economic Association, 14(3):639–668. 23 Bidwell, K., Watine, L., and Perry, K. (2014). Exploring early education programs in peri- urban settings in africa. Innovations for Poverty Action: Accra, Ghana. Calderón, G. (2014). The effects of child care provision in mexico. Technical report, Work- ing Papers. Callaway, B. and Sant’Anna, P. H. (2021). Difference-in-differences with multiple time periods. Journal of Econometrics, 225(2):200–230. Cengiz, D., Dube, A., Lindner, A., and Zipperer, B. (2019). The effect of minimum wages on low-wage jobs. The Quarterly Journal of Economics, 134(3):1405–1454. Clark, S., Kabiru, C. W., Laszlo, S., and Muthuri, S. (2019). The impact of childcare on poor urban women’s economic empowerment in africa. Demography, 56(4):1247–1272. Dang, H.-A. H., Hiraga, M., and Nguyen, C. V. (2022). Childcare and maternal employ- ment: evidence from vietnam. World Development, 159:106022. De Chaisemartin, C. and d’Haultfoeuille, X. (2020). Two-way fixed effects estimators with heterogeneous treatment effects. American Economic Review, 110(9):2964–2996. Devercelli, A. E. and Beaton-Day, F. (2020). Better jobs and brighter futures. Du, F. and Dong, X.-y. (2013). Women’s employment and child care choices in urban china during the economic transition. Economic Development and Cultural Change, 62(1):131– 155. Du, F., Dong, X.-y., and Zhang, Y. (2019). Grandparent-provided childcare and labor force 24 participation of mothers with preschool children in urban china. China Population and Development Studies, 2(4):347–368. Fruttero, A., Gurara, D., Kolovich, M. L. L., Malta, V., Tavares, M. M. M., Tchelishvili, N., and Fabrizio, M. S. (2020). Women in the Labor Force: The Role of Fiscal Policies. Interna- tional Monetary Fund. Geyer, J., Haan, P., and Wrohlich, K. (2015). The effects of family policy on maternal labor supply: Combining evidence from a structural model and a quasi-experimental approach. Labour Economics, 36:84–98. Givord, P. and Marbot, C. (2015). Does the cost of child care affect female labor market participation? an evaluation of a french reform of childcare subsidies. Labour Economics, 36:99–111. Goodman-Bacon, A. (2021). Difference-in-differences with variation in treatment timing. Journal of Econometrics, 225(2):254–277. Halim, D., Perova, E., and Reynolds, S. (2023). Childcare and mothers’ labor market outcomes in lower-and middle-income countries. The World Bank Research Observer, 38(1):73–114. Herbst, C. M. and Barnow, B. S. (2008). Close to home: A simultaneous equations model of the relationship between child care accessibility and female labor force participation. Journal of Family and Economic Issues, 29:128–151. Hsieh, C.-T., Hurst, E., Jones, C. I., and Klenow, P. J. (2019). The allocation of talent and us economic growth. Econometrica, 87(5):1439–1474. 25 Hyland, M., Djankov, S., and Goldberg, P. K. (2020). Gendered laws and women in the workforce. American Economic Review: Insights, 2(4):475–490. ILO (2021). Building forward fairer: Women’s right to work and at work at the core of the covid-19 recovery. Kabeer, N. (2008). Paid work, women’s empowerment and gender justice: critical path- ways of social change. Lokshin, M., Glinskaya, E. E., and Garcia, M. (2000). The effect of early childhood develop- ment programs on women’s labor force participation and older children’s schooling in Kenya, volume 15. World Bank Publications. Medrano Vera, P. (2009). Public day care and female labor force participation: evidence from chile. Universidad de Chile. Mills, M., Präg, P., Tsang, F., Begall, K., Derbyshire, J., Kohle, L., Miani, C., and Hoorens, S. (2014). Use of childcare in the eu member states and progress towards the barcelona targets. Morrissey, T. W. (2017). Child care and parent labor force participation: a review of the research literature. Review of Economics of the Household, 15(1):1–24. Nollenberger, N. and Rodríguez-Planas, N. (2015). Full-time universal childcare in a con- text of low maternal employment: Quasi-experimental evidence from spain. Labour Economics, 36:124–136. Olivetti, C. and Petrongolo, B. (2017). The economic consequences of family policies: 26 lessons from a century of legislation in high-income countries. Journal of Economic Per- spectives, 31(1):205–230. Olubor, R. O. (2009). Private cost analysis of pre-school education in nigerian private schools. Journal of Social Sciences, 19(2):141–148. Rosero, J. and Oosterbeek, H. (2011). Trade-offs between different early childhood inter- ventions: Evidence from ecuador. UvA-DARE (Digital Academic Repository) Universiteit van Amsterdam. Roth, J., Sant’Anna, P. H., Bilinski, A., and Poe, J. (2023). What’s trending in difference-in- differences? a synthesis of the recent econometrics literature. Journal of Econometrics. Sanfelice, V. (2018). Universal public childcare and labor force participation of mothers in brazil. Staab, S. (2015). Gender Equality, Child Development and Job Creation: How to Reap the’Triple Dividend’From Early Childhood Education and Care Services. UN Women. Sun, L. and Abraham, S. (2021). Estimating dynamic treatment effects in event studies with heterogeneous treatment effects. Journal of Econometrics, 225(2):175–199. Virginia, R. D. L. V. et al. (2015). Unpaid care work: The missing link in the analysis of. Zaid, Z. B. and Amin, S. B. M. (2021). The influence of the quality and availability of childcare on mothers’participation in the labour force. International Journal of Accounting, 6(33):164–173. 27 7 Tables and Figures Table 1: Summary statistics N Mean St. Dev. Min Max Female labor force participation rate 918 53.91 13.29 14.21 92.79 Childcare law enacted (P ostc t) 918 0.55 0.50 0 1 Year of law enactment 815 2007.84 8.93 1976 2021 Year of law commencement 815 2008.04 9.03 1976 2021 Childcare availability score 918 0.30 0.29 0 0.84 Childcare affordability score 918 0.18 .22 0 .91 Childcare quality score 918 0.32 0.34 0 1 Average age at first marriage among women 892 25.68 3.33 18.40 33.20 Marriage rate (per 1,000 people) 837 5.81 2.58 2.20 15.70 Women with primary education (%) 858 15.89 16.85 0.02 73.16 Women with secondary education (%) 858 48.26 17.46 2.65 97.16 Women with tertiary education (%) 858 29.20 19.27 0.11 81.59 Notes: This table presents summary statistics for our country-year panel dataset covering 80 countries for the period 2000-2020. The data is obtained by merging the WBL dataset on childcare regulations with the ILOSTAT indicators of yearly female labor force participation rates. Data on mean age at first marriage and marriage rates come from Our World in Data, while education data comes from Barro and Lee (2013). 28 Table 2: Childcare laws and female labor force participation Dependent variable: Female labor force participation (1) (2) (3) (4) (5) (6) Panel A: All periods Childcare law enacted 1.206 0.386 0.976* (0.956) (0.922) (0.521) Childcare law commenced 0.840 -0.0824 1.073* (1.127) (1.222) (0.543) Observations 11,899 9,497 9,497 12,171 9,715 9,715 Control mean of dep var 41.301 55.661 55.661 43.742 55.905 55.905 Controls ✓ ✓ ✓ ✓ Country time trend ✓ ✓ Panel B: Excluding year of enactment/commencement Childcare law enacted 1.278 0.401 1.271* (1.051) (1.059) (0.633) Childcare law commenced 1.040 0.0246 1.723** (1.309) (1.478) (0.794) Observations 11,866 9,467 9,467 12,104 9,657 9,657 Controls ✓ ✓ ✓ ✓ Country time trend ✓ ✓ Panel C: Excluding year of enactment/commencement and one year after Childcare law enacted 1.350 0.357 1.665** (1.159) (1.244) (0.771) Childcare law commenced 0.942 -0.0120 1.374** (1.228) (1.352) (0.662) Observations 11,828 9,436 9,436 12,139 9,687 9,687 Controls ✓ ✓ ✓ ✓ Country time trend ✓ ✓ Notes: Each column represents a different regression estimated in an event-by-event stacked dataset as in Cengiz et al. (2019), where the dependent variable is the female labor force participation rate and the panel in- dicates the estimation sample. Panel A is estimated on the full sample, Panel B excludes the year of enactment of the law, and Panel C excludes the year of enactment and one year after. Childcare law enacted/commenced is a dummy indicator equal to one after the childcare law has been enacted/commenced. All regressions in- clude year-cohort and country-cohort fixed effects. Columns 2, 3, 5, and 6 include the following controls: population marriage rate, average age at marriage for women, and share of women with primary, secondary and tertiary education. Columns 3 and 6 also include country-specific linear time trends. Standard errors clustered at the country level are shown in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01 29 Table 3: Childcare law pillars and female labor force participation Dependent variable: Female labor force participation (1) (2) (3) (4) (5) (6) (7) (8) (9) All periods Excl. enactment year Excl. enactment year + 1 Childcare law pillar (standardized score) Availability 0.532* 0.675* 0.860* (0.301) (0.361) (0.434) Affordability 0.492 0.635* 0.876* (0.310) (0.374) (0.471) Quality 0.460 0.586 0.773* (0.302) (0.354) (0.421) Observations 9,497 9,497 9,497 9,467 9,467 9,467 9,436 9,436 9,436 Notes: Each column represents a different regression estimated in an event-by-event stacked dataset as in Cengiz et al. (2019) (see equation 1) where the dependent variable is the female labor force participation rate. Columns 1-3 are estimated on the full sample, columns 4-6 exclude the year of enactment of the law, and columns 7-9 exclude the year of enactment and one year after. Treatment variables of childcare availability, affordability, and quality are standardized scores. The pillar scores are constructed as a simple average over all the variables included in each pillar, which are either dummy indicators or normalized between 0 and 1. The score takes a value of 0 before the enactment of the law or when the law has been enacted and does not include any of the elements that were tracked in that pillar. A score equal to one means that the country’s law has been enacted and includes all possible elements that are tracked for that pillar. Then, these scores are standardized across the full sample. All regressions include year-cohort and country-cohort fixed effects as well as the following controls: population marriage rate, average age at marriage for women, and share of women with primary, secondary, and tertiary education, and country-specific linear time trends. Standard errors clustered at the country level are shown in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01 30 Figure 1: Dynamic effects of childcare law on female labor force participation Notes: The figure shows the estimates of the stacked event-study (Equation 2). The outcome variable is the female labor force participation rate. The specification includes controls for the population marriage rate, the average age at marriage for women, and the share of women with primary, secondary and tertiary education, as well as country-specific linear time trends. The bars represent 95 percent confidence intervals. Standard errors are clustered at the country level. 31 Figure 2: Heterogeneous effects by marital status Notes: This figure shows the coefficient of the childcare law enactment on a regression estimated in an event-cohort stacked dataset where the dependent variable is female labor force participation by marital status specified on the y-axis. The year of enactment of the law is excluded from the estimation sample and the specification includes the following controls: population marriage rate, average age at marriage for women, and the share of women with primary, secondary, and tertiary education in the corresponding age group as well as country-specific linear time trends. The specification is equivalent to column (6) in Panel B of 2. The bars represent 95 percent confidence intervals. 32 Figure 3: Heterogeneous effects by age Notes: This figure shows the coefficient of the childcare law enactment on a regression estimated in an event-cohort stacked dataset where the dependent variable is female labor force participation in the age group specified on the y axis. The year of enactment of the law is excluded from the estimation sample and the specification includes the following controls: population marriage rate, average age at marriage for women, and the share of women with primary, secondary, and tertiary education in the corresponding age group as well as country-specific linear time trends. The specification is equivalent to column (6) in Panel B of 2. The bars represent 95 percent confidence intervals. 33 Figure 4: Heterogeneous effects by education Notes: This figure shows the coefficient of the childcare law enactment on a regression estimated in an event- cohort stacked dataset where the dependent variable is female labor force participation in the education group specified on the y axis. The year of enactment of the law is excluded from the estimation sample and the specification includes the following controls: population marriage rate, average age at marriage for women, and the share of women with primary, secondary, and tertiary education in the corresponding age group as well as country-specific linear time trends. The specification is equivalent to column (6) in Panel B of 2. The bars represent 95 percent confidence intervals. 34 Figure 5: Alternative estimators of the dynamic effect of childcare law on female labor force participation Notes: This figure overlays the event-study plots constructed using four different estimators: a dynamic version of the TWFE model, equation (2), estimated using OLS (in red with line markers); Cengiz et al. (2019), estimated on a stacked dataset with OLS (in black with square markers); Sun and Abraham (2021) (in blue with diamond markers); and Callaway and Sant’Anna (2021) (in green with triangle markers). Estimates according to De Chaisemartin and d’Haultfoeuille (2020) are relatively noisier and are excluded from this figure to preserve the scale (see Figure A1). The treatment group variable is given by the year in which the country enacted the childcare law. The bars represent 95 percent confidence intervals. Standard errors are clustered at the country level. All estimates include the following controls: population marriage rate, average age at marriage for women, and the share of women with primary, secondary, and tertiary education. All estimates include country-specific linear time trends except for those based on Callaway and Sant’Anna (2021). 35 Figure 6: Placebo test Notes: The figure shows point estimates and 95 percent confidence intervals of the coefficient of childcare law enactment of the main specification over 50 random assignments of a placebo year of enactment. The sample only includes years before the true year of childcare law enactment and excludes the placebo year of enactment from the estimation. The specification is equivalent to column 6 of panel B of Table 1. 36 Appendix Table A1: WBL variables by childcare pillar Availability Does the law regulate public provision of childcare services? Does the law regulate private provision of childcare services? Does the law mandate employers to provide/support childcare services? Does the law provide for flexible hours at public childcare centers? Does the law provide for flexible hours at private childcare centers? Does the law mandate min. operation hours at public childcare centers? How many min. hours for public childcare centers? (normalized) Does the law mandate min. operation hours at private childcare centers? How many min. hours for private childcare centers? (normalized) Affordability Does the law establish free provision of childcare services? When fees are required by public care providers, are they regulated? When fees are required by private care providers, are they regulated? Does public provision of childcare services cover meals? Does the law establish meals at no additional cost to parents? Does the government provide financial support to parents for childcare? Does the government have programs to support low-income families? Do parents receive tax benefits for using childcare services? Does the government provide private childcare with financial support? Do private childcare centers receive tax benefits? Does providing childcare for employees result in support for employees? For example, support in form of tax benefits? Quality Are public childcare centers required to obtain licenses? Are private childcare centers required to obtain licenses? Does the law mandate a teacher-to-child ratio for public childcare? Does the law mandate a teacher-to-child ratio for private childcare? Does the law mandate a max. group size at public childcare centers? Does the law mandate a max. group size at private childcare centers? Does the law require educators at public centers to undergo training? Does the law require educators at private centers to undergo training? Does the law require inspections for compliance with laws and others? Does the government require reporting of information by public centers? Does the government require reporting of information by private centers? Are there penalties imposed for noncompliance on public centers? Are there penalties imposed for noncompliance on private centers? 37 Table A2: Sample characteristics Sample No controls With controls Number of countries 80 54 Number of observations (country-year) 918 760 Mean obs. per country 11.47 14.07 Median obs. per country 11.00 16.00 Max obs. per country 21.00 21.00 Countries where law is enacted (%) 0.79 0.85 Countries w/only pre-law outcomes (%) 0.10 0.07 Countries w/only post-law outcomes (%) 0.30 0.30 Countries w/both pre and post law (%) 0.60 0.63 Countries where law is never enacted (%) 0.21 0.15 Number of observations (stacked dataset) 11,899 9,497 Notes: The table summarizes the main characteristics of the dataset by estimation sample. The first column shows the characteristics of the estimation sample for the baseline spec- ification, and the second column shows those for the full specification with marriage and education controls. 38 Table A3: Coverage by country Childcare law Observations Years covered Country Enacted Year All Pre Post First Last Angola Yes 2016 5 4 1 2004 2019 Argentina Yes 2006 14 2 12 2004 2020 Australia Yes 2010 11 0 11 2010 2020 Bangladesh Yes 2021 4 4 0 2006 2017 Barbados Yes 1983 5 0 5 2015 2019 Belgium Yes 2019 21 19 2 2000 2020 Bhutan No - 3 - - 2018 2020 Bolivia No - 19 - - 2000 2020 Botswana No - 4 - - 2006 2020 Brazil Yes 1996 6 0 6 2009 2015 Bulgaria Yes 2008 21 8 13 2000 2020 Cabo Verde Yes 2018 1 1 0 2015 2015 Canada Yes 2015 21 15 6 2000 2020 Chile Yes 2009 15 3 12 2000 2020 Colombia Yes 2004 17 2 15 2002 2019 Croatia Yes 1997 19 0 19 2002 2020 Czech Republic Yes 2014 21 14 7 2000 2020 Côte d’Ivoire Yes 2015 5 2 3 2012 2019 Denmark Yes 2016 21 16 5 2000 2020 Djibouti Yes 2019 1 1 0 2017 2017 Ecuador Yes 2014 16 9 7 2005 2020 Egypt, Arab Rep. Yes 1996 13 0 13 2008 2020 Ethiopia No - 2 - - 2005 2013 Fiji No - 3 - - 2005 2016 France Yes 2003 16 0 16 2005 2020 Georgia Yes 2016 11 7 4 2009 2020 Ghana Yes 1998 6 0 6 2000 2017 Guinea Yes 1984 2 0 2 2002 2019 Notes: The table summarizes the main characteristics of the baseline esti- mation sample by country. Columns 1 and 2 indicate whether the country has enacted a childcare law by 2022 and when it was enacted according to the WDL dataset. Columns 4-6 display the number of country-year obser- vations, overall and before and after the law enactment. Finally, the last two columns show the first and last year observed in the dataset for the country. 39 Table A4: Coverage by country (continued) Childcare law Observations Years covered Country Enacted Year All Pre Post First Last Hong Kong SAR, China Yes 1976 11 0 11 2010 2020 India Yes 2017 7 4 3 2000 2020 Indonesia Yes 2003 21 3 18 2000 2020 Iran, Islamic Rep. Yes 2011 16 6 10 2005 2020 Jordan Yes 2018 4 1 3 2017 2020 Kenya Yes 2017 4 3 1 2005 2019 Korea, Rep. Yes 1991 21 0 21 2000 2020 Lithuania Yes 1991 21 0 21 2000 2020 Malawi No - 4 - - 2005 2020 Malta Yes 2019 12 10 2 2009 2020 Mauritania No - 3 - - 2012 2019 Mauritius Yes 2000 20 0 20 2001 2020 Mexico Yes 2019 19 17 2 2002 2020 Moldova Yes 2014 21 14 7 2000 2020 Mongolia No - 12 - - 2009 2020 Namibia Yes 2015 6 4 2 2010 2018 Nepal Yes 2002 2 0 2 2008 2017 Nicaragua No - 4 - - 2001 2014 Norway Yes 2005 21 5 16 2000 2020 Pakistan No - 11 - - 2006 2019 Panama Yes 2016 12 8 4 2003 2019 Paraguay Yes 1998 20 0 20 2001 2020 Peru Yes 2003 19 1 18 2002 2020 Philippines Yes 2013 19 11 8 2001 2020 Poland Yes 2011 21 11 10 2000 2020 Portugal Yes 2011 21 11 10 2000 2020 Romania Yes 2011 21 11 10 2000 2020 Russian Federation Yes 2012 11 2 9 2010 2020 Rwanda No - 5 - - 2014 2020 Notes: The table summarizes the main characteristics of the baseline estimation sample by country. Columns 1 and 2 indicate whether the country has enacted a childcare law by 2022 and when it was enacted according to the WDL dataset. Columns 4-6 display the number of country-year observations, overall and be- fore and after the law enactment. Finally, the last two columns show the first and last year observed in the dataset for the country. 40 Table A5: Coverage by country (continued) Childcare law Observations Years covered Country Enacted Year All Pre Post First Last Senegal No - 5 - - 2015 2019 Serbia Yes 2010 13 2 11 2008 2020 Sierra Leone No - 3 - - 2003 2018 Slovenia Yes 1996 1 0 1 2000 2000 South Africa Yes 2004 21 4 17 2000 2020 Spain Yes 2008 21 8 13 2000 2020 Sri Lanka No - 10 - - 2010 2020 Suriname Yes 2017 1 1 0 2016 2016 Switzerland Yes 2008 21 8 13 2000 2020 Tajikistan Yes 2013 3 3 0 2003 2009 Tanzania Yes 2009 6 2 4 2001 2020 Thailand Yes 1999 11 0 11 2000 2020 Togo No - 4 - - 2006 2017 Trinidad and Tobago No - 9 - - 2010 2020 Tunisia Yes 2006 12 1 11 2005 2017 Turkey Yes 2015 21 15 6 2000 2020 Uganda No - 2 - - 2012 2017 Ukraine Yes 2001 3 0 3 2018 2020 United Arab Emirates Yes 1983 3 0 3 2017 2019 United Kingdom Yes 2006 20 6 14 2000 2019 United States Yes 2021 21 21 0 2000 2020 Vietnam Yes 2019 11 9 2 2010 2020 Zambia Yes 2011 4 0 4 2017 2020 Notes: The table summarizes the main characteristics of the baseline estimation sample by country. Columns 1 and 2 indicate whether the country has enacted a childcare law by 2022 and when it was enacted according to the WDL dataset. Columns 4-6 display the number of country-year observations, overall and be- fore and after the law enactment. Finally, the last two columns show the first and last year observed in the dataset for the country. 41 Figure A1: Dynamic effects of childcare law on female labor force participation Notes: This figure overlays the event-study plots constructed using four different estimators: a dynamic ver- sion of the TWFE model, equation (2), estimated using OLS (in red with line markers); Cengiz et al. (2019), estimated on a stacked dataset with OLS (in black with square markers); Sun and Abraham (2021) (in blue with diamond markers); and Callaway and Sant’Anna (2021) (in green with triangle markers); De Chaise- martin and d’Haultfoeuille (2020) (in yellow with square markers). The treatment group variable is given by the year in which the country enacted the childcare law. The bars represent 90 percent confidence intervals. Standard errors are clustered at the country level. All estimates include the following controls: population marriage rate, average age at marriage for women, and the share of women with primary, secondary, and higher education. All estimates include country-specific linear time trends except for those based on Call- away and Sant’Anna (2021). 42