Policy Research Working Paper 10188 The Impact of Paid Maternity Leave on Women’s Employment Evidence Using Firm-Level Survey Data from Developing Countries Mohammad Amin Asif M. Islam Global Indicators Group & Middle East and North Africa Region September 2022 Policy Research Working Paper 10188 Abstract The relationship between paid maternity leave and the share the number of days of paid maternity leave from its smallest of female workers in registered private firms is analyzed to highest value (0 to 410 days) increases the share of female using firm-level survey data for 111 developing and emerg- workers by 12.5 percentage points. The positive relation- ing countries. Theoretically, the relationship can be either ship between female workers and maternity leave is much positive or negative. Higher maternity leave raises the cost larger when maternity leave is fully funded by the govern- of female workers to the employer, discouraging female ment than when paid for by the employer, and in countries employment. However, higher maternity leave encourages where there is a higher share of females in the childbearing more females to enter the labor market, implying greater age group of 20–29 years. These heterogeneities highlight female employment. The results show that the latter effect channels that accentuate the relationship, thereby serving dominates. That is, a positive, large, and statistically sig- as checks against endogeneity concerns with the estimation. nificant relationship is found between maternity leave and The distributional implications of paid maternity leave are female employment. A conservative baseline estimate is also analyzed by estimating its impact on low-skilled versus that the share of female workers in a firm increases by 2.08 high-skilled employment. Important policy implications of percentage points for each log point increase in the number the findings are discussed. of days of paid maternity leave. Alternatively, an increase in This paper is a product of the Global Indicators Group, Development Economics and the Office of the Chief Economist, Middle East and North Africa Region. 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 mamin@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 The Impact of Paid Maternity Leave on Women’s Employment: Evidence Using Firm-Level Survey Data from Developing Countries By: Mohammad Amina and Asif M. Islamb Keywords: Maternity leave, Women employment, Gender, Labor market JEL Codes: J13, J16, J18, J22 a Corresponding author. Senior Economist, Enterprise Analysis Unit, DECEA, World Bank, Washington, DC. Email: mamin@worldbank.org b Senior Economist, Office of the Chief Economist, Middle East and North Africa Region, World Bank, Washington DC. Email: aislam@worldbank.org. Phone: 202-458- 5246. 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. We thank the Enterprise Analysis Unit of the Development Economics Global Indicators Department of the World Bank Group for making the data available. We also thank Jorge Luis Rodriguez Meza and Norman Loayza for providing useful comments. All remaining errors are our own. 1. Introduction Most developing and emerging countries have implemented maternity leave policies to help women balance careers with childbearing and childrearing activities (World Bank 2022). Maternity leave policies have the potential to contribute significantly to the development process by improving women’s labor market participation. However, the efficacy of these policies is yet to be properly established. Theory suggests competing effects of maternity leave on female employment. The available empirical evidence is inconclusive, confined mostly to developed countries, and often fails to distinguish between women in the formal vs. informal sector, even though maternity leave laws only apply to the formal sector workers. The present paper contributes to the literature using firm-level survey data on private registered or formal firms in 111 developing and emerging countries. We estimate the impact of the duration of paid maternity leave (henceforth, maternity leave) on the share of workers in a formal private sector firm that are females (henceforth, female workers). The results indicate a large, positive, and statistically significant relationship between the two. We uncover important heterogeneities in this relationship depending on who bears the cost of maternity leave (government or employer), and the share of females in the country who are in the childbearing age group of 20-29 years. We argue that these heterogeneities serve the additional purpose of raising our confidence against endogeneity concerns. We also find strong distributional implications of maternity leave in that maternity leave benefits more women at the relatively low end of the skills and wage spectrum. The existing empirical studies on the impact of maternity leave on women’s labor market outcomes are largely restricted to the developed countries (see Albagli and Rau 2019, Dahl et al. 2016, Baum and Ruhm 2016, Zvegligh and Rodgers 2003, Ruhm 1998). With a few exceptions, studies on developing countries tend to focus on a single country or a handful of countries with no 2 indication if the results extend to other developing countries (see Uribe et al. 2019, Ilkkaracan 2012, Chang 2004). Some studies such as Fallon et al. (2017), Aslim et al. (2021), and Del Rey et al. (2021) do focus on a large cross-section of developing or emerging countries but employ macro- level data and so cannot distinguish beween female workers in the formal vs. the informal sector. This is an important limitation because maternity leave policies apply only to the formal and not informal sector workers. Given large levels of informality in most developing countries, a failure to isolate formal from informal workers implies that the estimated effect of maternity leave on female labor market outcomes will be biased towards zero. Fallon et al. (2017) acknowledge the problem and note that the lack of any significant relationship between maternity leave and female labor force particiaption in their sample could be attributed to the inclusion of both formal and informal workers in their labor force participation measure.1 Even so, the existing evidence on the impact of maternity leave for developed and developing countries is mixed and inconclusive (see Olivetti and Petrongolo 2017). Thus, there is room for further debate in this area. Establishing a causal effect of maternity leave on women’s labor market outcomes has been a challenge (see the discussion in Aslim et al. 2021, Olivetti and Petrongolo 2017, Berger and Wladfogel 2004). One issue is reverse causaltiy. An increase in female labor market participation may create demand and political pressure for maternity leave (see Ruhm 1998, Blau and Kahn 2013, Olivetti and Petrongolo 2017, Del Rey et al. 2021). The liteature suggests a few ways of mitigating this problem which are discussed in section 2. We follow and go beyond these suggestions. For instance, reverse causality is unlikely to be an issue when considering single industries that account for a small share of formal sector female employees in the country. Thus, 1 Fallon et al. (2017) note that: “Ideally, we would have liked to include a variable that measures the informal labor force independently of female labor force participation. Unfortunately, limited data availability prevents us from doing so.” 3 we test if the female workers and maternity leave relationship holds within industries such as manufacture of basic metals that typically accounts for a miniscule proportion of female formal sector workers. Another challenge is that maternity leave may be correlated with other drivers of female employment (omitted variable bias problem). We employ several checks to guard against this problem including using of a large number of controls, falsification test that involves comparing the impact of maternity leave on female employment versus female entrperenuerhsip, and in the spirit of Rajan and Zingales (1998), testing for some sensible predictions of the ways in which maternity leave affects female employment. The predictions relate to how the impact of maternity leave on female employment varies with who pays for maternity leave (government or employer), and the share of females in the childbearing age group in the country. The predictions are specific to the way maternity leave affects female employment and so unlikely to hold in the data if maternity leave were a mere proxy for other determinants of female employment. Part of the interest in women’s labor market activity stems from the fact that women tend to be less privileged than men. So, more jobs for women can empower them and lead to a more egalitarian society. However, not all women are the same and the distributional effects of maternity leave in part depend on whether maternity leave benefits more women at the low or high end of the wage or skills spectrum. There are arguments for a larger impact at either end, and the existing emprical evidence on the issue is scant, limited to developed countries, and mixed (discussed in section 2). We explore how the impact of maternity leave on female employment varies with the wage rate and skill level of women workers. We note that data limitations restrict us to only using the firm’s labor productivity and the average wage rate of all workers in a firm as proxies for women workers’ wage and skill level. Regardless, to the best of our knowledge, our paper is the first to explore the issue for developing and emerging countries. We go beyond the existing 4 literature by providing suggestive evidence that factors associated with the demand for female workers rather than just supply side factors highlighted in the literature may be driving the differential effect of maternity leave on low vs. high wage/skilled women workers. Our paper is also related to the broader literature on the drivers of female labor force participation (reviewed in section 2). Increase in women’s labor market activity over the last few decades has been a notable feature of the developed countries. However, the performance of developing and emerging countries in this regard has been modest (see Klasen 2019). We contribute to the borader literature on drivers of female employment, and specifically to the literature on the impact of maternity leave in several ways. First, we cover 111 developing and emerging countries, ensuring that our findings apply to almost all developing economies. As mentioned above, most existing studies on the impact of maternity leave are restricted to the developed countries. Second, we focus on women workers in the registered or formal private sector. This is important because maternity leave policies are only relevant for the formal sector. Third, most previous work is based on macro- or country-level data which fails to account for heterogeneities within a country. For instance, we find below that female employment is significantly higher in firms that have a female top manager and/or a female owner. Failure to account for such factors can bias the estimation results. Our empirical exercise effectively removes these and other such sources of bias. Fourth, we pay due attention to endogeneity concerns. We do so by using lagged values of maternity leave, employing a large number of controls, using a falsification test, checking if maternity leave affects female employment in industries that account for a small share of formal sector female workers, and most importantly, testing some predictions about the ways in which maternity leave impacts female employment. Fifth, we check how maternity leave affects women’s employment at the low versus high level of wage rate and skills. 5 Figure 1 shows that there is a positive relationship between the share of female workers in a typical firm in the country and maternity leave. Our empirical results confirm that this positive relationship is large, statistically significant, and robust. A conservative baseline estimate is that for each log point increase in maternity leave duration the share of female workers in a typical firm increases by 2.08 percentage points (against the mean share of 33.7 percent). In terms of the number of days of materntiy leave (without logs), an increase in maternity leave days from the smallest to the largest level (0 to 410 days) implies an increase of 0.91 percentage points in female workers on average for each additional month of maternity leave. Further, the positive relationship between maternity leave and female workers is much stronger (more positive) in countries where maternity leave is fully paid by the government compared to when it is paid (partly or fully) by the employer, and in countries that have more females in the childbdearing age group of 20-29 years as a proportion of the country’s total population. The relationship is also stronger among firms that have lower productivity and have a lower wage bill per worker, suggesting that maternity leave benefits more the relatively less skilled and low-wage women workers. 2. Conceptual framework 2.1 Maternity leave and female employment realtionship Maternity leave aims to provide better opportunities for women to balance their role as mothers and workers. It seeks to enable mothers to stay involved in the labor market during temporary interruptions of employment, while retaining firm-specific or occupation-specific human capital. Thus, women are encouraged to join the labor force both before and after childbirth. In addition to this supply side effect, if maternity leave makes continuity of employment for mothers easier, it may reduce the extent of statistical discrimination by employers against women and thereby 6 increase the demand for women workers. Countering these positive effects, maternity leave may depress demand for female workers if part of the costs of maternity leave directly or indirectly falls on employers. The direct costs may include employer’s monetary contribution (if any) to maternity leave benefits, while the indirect costs involve short-term work disruption and loss of skills due to the increased absence during childbirth. A negative supply side effect could be that longer maternity leave may delay women’s return to work. The net impact of these competing effects of maternity leave is essentially an empirical issue. Our analysis reveals that the positive effects dominate. We summarize this in the following hypothesis. Hypothesis 1: Higher duration of maternity leave is associated with a higher share of female workers in a firm. 2.2 Other determinants of female employment There are several other determinants of maternity leave highlighted in the literature. If these determinants happen to vary systematically with maternity leave, our results for the impact of maternity leave on female workers may suffer from the omitted variable bias problem. Thus, as a first defense against the problem, we draw on the exitsting literatue to control for several determinants of female employment. The motivation for the controls is as follows. Several studies have documented differences across industries in the share of female and male workers (see Borrowman and Klasen 2020, Olivetti and Petrongolo 2014, World Bank 2011, Goldin 2006). Factors such as “brain vs. brawn” content of jobs, culture and norms that view some industries and jobs as more “female oriented” than others, flexibility of working hours and so forth, are some of the underlying factors for the industry-wise segregation of men and women. Job 7 prospects of women relative to men may also be affected by annual global shocks such as those stemming from changes in technology, world trading system, and international capital flows. Regarding firm-level factors, firm-size and age could impact female employment directly by influencing firms’ preference and costs for discrimination (see Vo and Ha 2021, Zhang et al. 2021, and Bratti et al. 2005). Firm-size and age may also influence skill requirement, labor productivty, export orientation, and foreign ownership, which may in turn affect female employment. Firm size may also affect application of law. Several countries have stricter employment protection laws for large firms than small firms suggesting that larger firms may find it more difficult to discriminate against female workers. On the other hand, smaller and younger firms tend to face greater competition in the product market. More competition is often linked with lower discrimination and therfore higher women’s employment (see Cooke et al. 2019). There is also evidence to suggest that lack of proper skills or labor market discrimination may force women to work in firms that pay lower wages (Card et al. 2016). Thus, we may expect more female workers in firms that are less productive and dynamic in terms of growth rate and R&D activity. Exporting status has been linked with greater demand for female workers in several studies (see World Bank 2020, Juhn et al. 2014). There is also a small but growing body of work showing that foreign direct investment may improve the share of female workers through different channels including transplantation of more female-friendly foreign cultural norms (see Kodama et al. 2018, Ouedraogo and Marlet 2018). We also consider the gender of the firm’s owners and top manager. Female owners can be expected to discriminate less against female workers than their male counterparts, although the evidence on the issue is mixed (see Bossler et al. 2020, Bertrand et al. 2019, Nelson and Bridges 1999). Women may be more affected than men by crime as women are more likely to be the targets of criminal activities (Mishra et al. 2021, Islam 2013). In many 8 developing countries, women lag men in technical skills. Thus, provision of training by firms may be especially important for women’s labor market outcomes (see Da Mata et al. 2021, Card et al. 2018, Acevedo et al. 2017). Tight labor markets reflected in shortage of skilled and unskilled workers could be an important driver of female employment. Skill shortage can force employers to fall back on the relatively less skilled women workers or discriminate less against them. Another factor benefiting women relative to men is the rapid growth and adoption of ICT and computers (See Jain 2021, Viollaz and Winkler 2020). It offers greater flexbility in terms of work location that is especially benefitcial for women. Informality is pervasive in most developing countries. Some formal sector firms start off in the informal sector which may impact their subsequent growth and inputs choice. Competition that formal sector firms face from informal sector firms can have significant effects on formal firms’ product and inputs choice. For one, greater informal competition could reduce formal firms’ profit margins, making discrimination against women workers more costly. On the other hand, the presence of a large informal sector may encourage women workers to shift from formal to informal sector as the latter offers better work-life balance (see Schmieder 2021). Other factors such as easy access to finance, lower taxes, and lower regulatry burden on private firms can also improve market efficiency and ease entry of new firms thereby increasing the level of competition in the product market. As mentioned above, higher competition is likely to put pressure on firms to cut labor cost through for example, less discrimination against women workers. At the macro-level, higher GDP per capita is an important proxy for several factors such as lower fertility rates, higher education among women relative to men, greater women’s political empowerment, and so forth, that could have a positive impact on women’s relative to men’s employment (see Gaddis and Klasen 2014). There is also a view that structural transformations 9 that accompany the early and later phases of economic development benefit women more than men, yielding a U-shaped relatonship between GDP per capita and female employment (see Tunali et al. 2021, Mammen and Paxson 2000, Goldin 1995). However, recent work indicates that the structural effects may be much weaker than previously thought (Gaddis and Klasen 2014). Higher growth rate of GDP per capita may signal tight labor markets, forcing many employers to hire women when they would not do so otherwise (see Klasen 2019). Increase in women’s education and the rapidly declining gender gap in education have also been found to impact women’s labor market outcomes, although this relationship is not always straightforward (see Klasen et al. 2021, Bussemakers et al. 2017, Klasen and Pieters 2015). Ceteris paribus, more females in total adult population implies a greater share of women workers in the private sector. The impact of lower fertility rate on female employment has been analyzed in several studies. Some studies show that lower fertility can free up time from child care and can increase female labor supply (Bloom et al. 2009). However, other studies find no effect of fertility on women’s employment (Heath 2017), and a shift from formal to informal sector by women (Schmieder 2021). Gender-disparity in the laws related to employment, working hours, pay, sexual harassment at work, and so forth can significantly impact the demand and supply of females relative to male workers (see Hyland et al. 2020, Islam et al. 2019). Cultural, social, and religious norms regarding women’s work outside home have also been found to be improtant drivers of female labor force participation (see Jayachandran 2021, Chen and Ge 2018, Dildar 2015, Fernandez et al. 2004). Urbanization can fuel changes in a society that makes women’s work outside home socially more acceptable (World Bank 2011). Women’s representation in the public sector and their political empowerment more generally may have important effects on female employment. As Aslim et al. (2021) note, the duration of maternity leave could be endogenous if 10 more women in politics lead to more female politicians lobbying for longer maternity leave. Further, women in high positions could serve as role models for other women leading to higher participation of women in the labor markets (Aslim et al. 2021, Macmillan et al., 2018, Chatopadhyay and Duflo 2004). Paternity leave can also affect labor market outcome for women, although evidence on the issue is limited (Druedahl et al. 2019, Amin et al. 2016). As discussed above, freedom to do business and laws that faciliate easy entry and exit of firms are likely to increase competition in the product markets and therefore lead to less discrimination against women workers. In this study we further explore heterogeneities by examining how the impact of maternity leave varies depending on whether it is fully funded by the government or not, the share of women in the 20-29 years age group, and the skill level of employees at the firm proxied by labor productivity of the firm and wage bill per worker. It is possible that these differential effects or heterogeneities may spuriously proxy for other heterogeneities or for non-linear effects of maternity leave and share of women in the 20-29 age group, and so forth. Theory and empirical evidence do not provide any guidance on the sorts of heterogeneities that may be at play. However, the possibility cannot be ruled out. For instance, the impact of maternity leave on female workers may be higher in countries where women have a higher fertility rate. We pay due attention to such possible heterogeneities in our empirical analysis. 2.3 Endogeneity tests 2.3.1 Industries that employ few women One source of reverse causality problem is that an increase in female labor market participation may create political presssure for maternity leave (see Ruhm 1998, Blau and Kahn 2013, Olivetti 11 and Petrongolo 2017, Del Rey et al. 2021). Since our results show a positive impact of maternity leave on female employment (Hypothesis 1), the positive feedback effect implies that our regression results may overstate the true impact of maternity leave on female employment (upward bias). Del Rey et al. (2021) argue that this source of upward bias can be significantly mitigated by the inclusion of women in political office and social norms as controls in the regressions. The rationale for this is that female employment affects maternity leave policies, at least partly, through the impact of female employment on social norms and women in political office. In his seminal study, Ruhm (1998) argues that if there is reverse causality from female employment to maternity leave, then lead values (t+1 values) of maternity leave should impact female employment currently (at time t). They test for this using panel macro-level data for 9 European countries and confirm no significant reverse causality. Micro-level studies that focus on a single country seek to compare outcomes between observationally similar women in the tretament and control groups (see Dahl et al. 2016, Baker and Milligan 2008; Rossin-Slater 2017 provides a review of the related literature). Our first defense against the revese causality problem is that we use lagged values of maternity leave, and as suggested by Del Rey et al. (2021), control for women in the public sector (women in parliament) and social norms proxied by the level of the country’s urbanization and its religious composition. Furthermore, we argue that while higher female employment may create demand and political pressure for maternity leave, this is unlikely to be so in individual industries that account for a small share of all formal sector female workers in the country. Simply put, some industries are too small to exert any meaningful effect on maternity leave laws that are determined at the macro-level. Thus, finding a large, postiive, and statistically significant relationship between maternity leave and female employment in an industry that employs a small share of female 12 workers in the formal sector will help raise our confidence against reverse causality. We sumarize the test in the following hypothesis. Hypothesis 2: Reverse causality from female employment to maternity leave is unlikely to be an issue in an industry that employs a small share of all formal sector female workers. Thus, a positive relationship between maternity leave and female employment in such industry helps raise our confidence agaisnt the reverse causality problem. The discussion above implies that if reverse causality is present, it should affect more those industries that employ a relatively large share of formal sector female workers. Thus, we compare the strength of the relationship between maternity leave and female employment in industries that account for a small share of formal sector female workers (e.g. manufacture of basic metals) with the same in industries that account for a relatively large share of female workers in the formal sector (e.g. retail). We note that this comparison should be treated with due caution because there are several other differences between industries that potentially may influence the impact of maternity leave. 2.3.2 Who pays for maternity leave? Our next defense against endogeneity concerns involves testing for some sensible predictions about the channels through which maternity leave affects female employment. Since the predictions we test are specific to the way maternity leave affects female employment, there is little reason for them to hold in the data if the maternity variable is a mere proxy for other correlated drivers of female employment, or if causality runs from female employment to maternity leave. 13 This way of guarding against the endogeneity problem has been used in the literature (see for example, Rajan and Zingales 1998, Nunn 2007, Duchin et al. 2010, Distinguin et al. 2016). For instance, Rajan and Zingales (1998) estimate the impact of financial development on industry level growth. They suggest that one way to make progress on causality is to focus on the details of theoretical mechanisms through which financial development affects economic growth and document their working. Building on this idea, they argue that financial markets and institutions help a firm overcome problems of moral hazard and adverse selection, thus reducing the firm's cost of raising money from outsiders. So financial development should disproportionately help firms (or industries) typically more dependent on external finance for their growth. They note that such a finding could be the “smoking gun” in the debate about causality. Based on the strategy of Rajan and Zingales (1998), we argue that maternity leave discourages female employment more if employers have to pay the salaries of those on maternity leave (Lee and Cho 2005, Karshenas et al. 2014, Olivetti and Petrongolo 2017). This effect is absent when maternity leave is fully paid by the government. Thus, we predict that the positive impact of maternity leave on female employment is much larger when maternity leave is fully funded by the government vs. when it is not. The prediction is based on the fact that the government funding reduces the cost of maternity leave to the employer and thereby increases demand for female workers. There is no apparent reason why the prediction should hold if maternity leave were a proxy for other correlated drivers of female employment such as women’s education, social norms, childcare availability, etc. We summarize the argument in the following hypothesis. 14 Hypothesis 3: The impact of maternity leave on female employment is larger or more positive when maternity leave is fully funded by the government compared to when it is partly or fully paid by the employer. 2.3.3 Proportion of women in the childbearing age group For the second endogeneity check based on the logic of Rajan and Zingales (1998), we argue that the impact of maternity leave should be larger (more postive) when there are more females in the country that are elegible for maternity benefits. These are women in the childbearing age group. The idea that family policies such as maternity leave are more effective for women in the childrearing age group than the rest has been analyzed in other studies such as Ruhm (1998), Besamusca et al. (2015), and Bennet et al. (2020). Thus, we predict that the positive impact of higher maternity leave on female workers is much larger (more positive) in countries where there are more females in the 20-29 age group as a proportion of the country’s total population. The share of females in the 20-29 age group depends on the overall demographic structure of the population and therefore is relatively exogenous to female employment. One caveat here is that the nexus between maternity leave and the share of females in the 20-29 age group may by a proxy for the nexus between fertility rate and the share of females in the 20-29 age group. We show below that our results are robust to this possibility. We summarize the endogeneity test as follows. Hypothesis 4: The impact of maternity leave on female employment is larger or more positive when there are more females in the age group of 20-29 years as a proportion of the country’s total population. 15 2.3.4 Falsification test using female ownership Our last endogeneity check is a falsification test. Maternity leave laws are applicable to women employees but not to women entrpreneurs (firm owners). Thus, maternity leave should not have any positive effect on female entrepreneurship. In fact, by encouraging female wage labor, it is possible that maternity leave could cause a shift from entreprenurship to wage work among women. This implies a negative impact of maternity leave on female entrepreneurship. Thus, we predict that if our maternity leave variable is only capturing maternity leave and not other factors then higher maternity leave should have either no effect or a negative impact on female entrepreneurship in the country. However, some of the factors that drive female employment such as social norms, education among women, childcare and so forth, are also likely to affect women’s participation as entrepreneurs. Thus, if our maternity leave variable is a proxy for the other determinants of female employment then we should find a significant positive link between maternity leave and the likelihood of having women entrepreneurs in the country. We summarize the falsfication test in the following hypothesis. Hypothesis 5: Absence of a significant positive relationship between maternity leave and the probability that a firm has (one or more) female owners will bolster our confidence against endogeneity concerns with our main results. 2.4 Hetergoneity by skill level The impact of maternity leave and other family policies on skilled vs. unskilled workers or low- wage vs. high-wage workers has been discussed in the literature, although not in the context of developing countries (see Del Boca et al. 2009, Cipollone et al. 2014, Olivetti and Petrongolo 16 2017). For instance, Del Boca et al. (2009) use education as a proxy for the expected wage rate of a worker. They find that across nine European countries, duration of parental leave significantly impacts a woman’s chances of participation in the labor market,2 but only for women with lower educational attainment. The authors argue that women who have made greater investment in education are more likely to work as they need to recoup the cost of investment in education, face better job opportunities in terms of wages and benefits, and in all likelihood, differ in their preferences for market work in comparison with women with less education. Thus, more educated women are less senstive to external environment factors such as maternity leave. In contrast, the opposite is found by Cipollone et al. (2014). The authors analyze 15 EU countries and find that labor supply of the relatively more educated women is much more responsive to a combined measure of family leave policies (parental leave, family subsidies, and elderly subsidies). They argue that the relatively weaker impact on less educated and low-skilled women is because in many EU countries, family policies are associated with a deterioration of job quality, especially for low educated women, increasing their chances of dropping out from the labor force. Both Del Boca (2009) and Cipollone et al. (2014) focus on differences in the supply of labor by low- vs. high- skilled/educated workers. Another possibility that we explore in this paper is that demand for female workers may adjust differently for low- vs. high-skilled female workers. High-skilled women workers embody firm-specific human capital which incentivizes firms to retain them post childbirth regardless of maternity leave laws. Thus, maternity leave plays only a limited role in job continutity for high-skilled women. In contrast, low-skilled women workers accumulate much less knowledge on the job. They can be easily replaced with new workers with some training. Absent maternity leave laws, employers are less likely to wait for low-skilled expecting mothers to return 2 The study finds that women’s probability of participating in the labor market is positively related to parental leave duration initially, then starts to decline. 17 to their job than high-skilled expecting mothers. Thus, by ensuring job continuity of low-skilled women workers who would otherwise lose their job due to childbirth, maternity leave is expected to impact low-skilled women workers more than high-skilled workers. Our emprical results below provide suggestive and preliminary evidence to support this view. 3. Data and methods 3.1 Data description In this section, we discuss the data and estimation method used in the empirical analysis. The main data sources are Enterprise Surveys (ES) and Women, Business and the Law (WBL) project. Both these data are collected by the World Bank. The ES is firm-level survey data on formal or registered private sector firms with five or more full-time permanent workers. The surveys cover all of manufacturing but exclude agriculture and certain services industries such as finance & insurance, health care, and education. Informal or unregistered firms are also excluded and so are fully state-owned enterprises. The surveys use a common sampling methodology, stratified random sampling, as well as a common questionnaire across all countries.3 Stratification is done by firm size, industry, and location within the country. Sampling weights are provided in the ES and used in all our regressions so that the sample is representative of the targeted private sector in the country.4 The WBL provides information on legal institutions that discriminate by gender across the world as well as on the length and cost of paid maternity leave. This information is collected through several rounds of interactions with practitioners with expertise in the different areas covered by the database. Inputs collected from practitioners are verified against codified 3 Details of the sampling methodology and other survey related information are available at www.enterprisesurveys.org. 4 The median sampling weights provided by ES are used in the regressions below. 18 sources of national law and compiled into indicators by the WBL team. We complement the ES and WBL with other data sources such as World Development Indicators (WDI), World Bank; Economic Freedom of the World, Fraser Institute, Maoz and Henderson (2013), and Inter Parliamentary Union (IPU). The sample we use includes all countries and all firms surveyed by ES for which data on our main variables are available. Some countries in our sample have multiple rounds of ES. The ES are not conducted every year in a country and their frequency varies across regions and countries.5 Our main regression results are based on a sample of 80,272 manufacturing and services firms in 111 countries. The number of ES rounds (country-year pair) in the sample is 191 spanning 2008-2019. In the Appendix, table A1 provides the list of countries in the sample along with the survey year, table A2 provides a formal definition of all the variables used in the regressions, and table A3 provides the summary statistics of the variables. 3.2 Estimation methodology Our baseline results are obtained from estimating the following equation: = + 1 + + + + −1 + … (1) where subscript i denotes the firm, j denotes the country where the firm operates, t denotes the year, and τ represents average values over t-1 to t-5. Y is the share of females in total workers at the firm; Maternity leave is (log of 1 plus) number of days of paid maternity leave mandated by 5 Typically, ES is repeated in a country every 5-6 years. 19 law in the country and averaged over the last 5 years; YFE is a set of dummy variables for the year the ES was conducted (Year fixed effects), and IFE is a set of dummy variables for the firm’s industry (Industry fixed effects). Firm controls include various controls for firm characteristics, Country controls are controls for country characteristics that vary over time, and u is the error term. Equation (1) is estimated using Ordinary Least Squares (OLS). All regressions use Huber- White robust standard errors and clustered at the country times year (or ES round) level. As discussed above, we go beyond and explore how the relationship between female workers and maternity leave depends on other country and firm characteristics. This heterogeneity is estimated using interaction terms between the relevant variables and the split-sample method. The estimation using interaction terms is based on the following equaion: = + 1 + 2 ∗ + 3 + + + + −1 + + … (2) Equation (2) differs from equation (1) in two ways. First, it includes the interaction terms between maternity leave and country and firm characteristics captured by Z. These interaction terms estimate how the relationship between maternity leave and female workers varies with country and firm characteristics (details below). The second change in equation (2) is that it includes several interaction terms with control variables. The additional interaction term controls ensure that our main interaction term in the equation is not spuriously picking up the differential impact of maternity leave in rich vs. poor countries, small vs. large firms, and so forth. Equation (2) is 20 estimated like equation (1). The split-sample method used for heterogeneity analysis involves estimating separately equation (1) for low vs. high values of the relevant variables. 3.3 Dependent variable The dependent variable is the share of female workers at the firm. The ES asked each firm the number of female full-time permanent workers employed at the firm at the end of the last fiscal year. A similar question was asked about total (male plus female) workers. We define our dependent variable as the number of female full-time permanent workers expressed as a percentage of all (male plus female) workers at the firm (Female workers). The mean value of female workers equals 33.7 percent, and the standard deviation is 27.8 percent. 3.4 Main explanatory variable Our main explanatory variable is the length of paid maternity leave in the country. It equals (log of 1 plus) number of days of paid maternity leave mandated by law and averaged over the last 5 years prior to the year the ES was conducted (Maternity leave). The mean value of Maternity leave equals 4.64 (log days) and the standard deviation is 0.62. In our sample, Portugal and Papua New Guinea have the least maternity leave (0 days) while Bulgaria (410 days) has the maximum. We use 5-year average values of maternity leave to capture any lagged effects and eliminate any unduly large effects of annual fluctuations. For most countries, the length of maternity leave does not change over the period considered. Nevertheless, we explore other lagged structures for robustness purpose. That is, we also use as our main explanatory variable (log of 1 plus) paid maternity leave in the last year prior to the ES, and (log of 1 plus) paid maternity leave averaged 21 over the last 10 years prior to the year the ES was conducted in the country. Last, we show that our results hold if we use the number of days of paid maternity leave instead of its log value. 3.5 Variables used for heterogeneity analysis Our first heterogeneity test relates to who pays for maternity leave in the country. To this end, we first define a dummy variable equal to one if the maternity leave is 100 percent paid by the government and zero otherwise. We use average value of this dummy over the last five years prior to the year the ES was conducted (Govt. Funded). Data source for the variable is WBL. The mean value of Govt. Funded equals 0.65 and the standard deviation is 0.48. The next heterogeneity test analyzes how the impact of maternity leave on female workers varies with the share of females in the country that are in the childbearing age group. We assume this group to include all women who are 20-29 years old. Thus, we compute the number of females that are in the 20-29 age group in the country expressed as a percentage of the country’s total population. We take average values of the variable over the last 5 years prior to the year the ES was conducted (Females: 20-29 yrs. old (%)). Data source for the variable is WDI, World Bank. The last heterogeneity we analyze is with respect to the skill level of women workers. The ES does not provide information on employees’ skill level or education. Thus, we use two proxy measures defined at the firm-level. The first is (log of) total real wage bill (in constant 2009 USD) divided by the number of employees at the firm (Wage rate). The second measure is (log) labor productivity of the firms as defined above. The underlying assumption is that firms that pay lower wages and have lower labor productivity have proportionately more low-skilled workers, including women workers, than firms that pay higher wages and have higher labor productivity. In separate regressions, we interact each of these proxy measures with maternity leave. These interaction terms 22 capture how the impact of maternity leave on female workers varies with the level wage rate and labor productivity of the firm. 3.6 Baseline control variables The maternity leave variable does not change over time for most countries in our sample. Thus, the identification of the female workers and maternity leave relationship (henceforth, main result) is based on cross-country variation. It is well-known that identification based on cross-country differences is more prone to endogeneity problems than changes over time for a given country. Thus, we provide several layers of defense against endogeneity concerns. To begin with, we control for several determinants of female workers that may also vary systematically with the duration of maternity leave. The motivation for the controls comes from the existing literature on the drivers of female employment and is discussed in detail in section 2 above. The controls used in the baseline model are as follows (see table A2 in the Appendix for a detailed definition): dummy variables indicating the year the ES was conducted (Year fixed effects); dummy variables indicating the industry (2-digit, ISIC Rev. 3.1) to which the firm belongs (Industry fixed effects)6; log of GDP per capita last year (one-year lagged) taken from WDI; annual growth rate of GDP per capita taken from WDI; percentage of females in total adult population last year taken from WDI; fertility rate last year obtained from WDI; gender parity in primary education proxied by the ratio of female primary education enrollment rate to male primary education enrollment rate averaged over the last five years and taken from WDI7; freedom from business regulations last year taken from Fraser Institute; index of gender parity in workplace laws 6 There are 46 industries at the 2-digit level in the baseline sample. 7 Data on primary education enrollment rates are not available for all countries for each year. Thus, we use average values of the variable over the last 5 years for which data are available to minimize missing values. 23 averaged over the last five years and taken from WBL; urbanization rate or the percentage of country’s population living in urban area last year taken from WDI; paternity leave defined analogously to maternity leave as (log of 1 plus) number of days of paid paternity leave mandated by law averaged over the last 5 years and taken from WBL; and the percerntage of seats in the lower house of parliaments that are held by women (average value over the last five years prior to the ES8) obtained from Inter Parliamentary Union (Women in Parliament). Firm-level controls all taken from the ES include firm-size measured by (log of) number of workers at the firm; (log of) age of the firm; labor productivity equal to (log of) real annual sales of the firm in the last year (in 2009 USD) divided by the number of workers; dummy variable equal to 1 if the firm has one or more female owners and 0 otherwise; dummy variable equal to 1 if the firm’s top manager is female and 0 otherwise; percentage of firm’s annual sales last year made abroad (Exports); percentage of firm’s ownership that is with foreign entities (Foreign ownership); dummy variable equal to 1 if the firm provides training to its workers and 0 otherwise; dummy variable equal to 1 if the firm experienced losses due to crime last year and 0 otherwise; dummy variable equal to 1 if the firm was registered when it started operations and 0 otherwise; dummy variable equal to 1 if the firm has its own website and 0 otherwise; severity level of inadequately education workers (on 0-4 scale) as an obstacle to firm’s current operations as reported by the firms; severity level of access to finance (on 0-4 scale) as an obstacle to firm’s current operations as reported by the firms; severity level of tax rates (on 0-4 scale) as an obstacle to firm’s current operations as reported by the firms. 3.7 Robustness controls 8 We use average values over the last 5 years for which data are available to avoid loss of sample due to missing data for some countries in a single year. 24 We use additional controls for robustness purposes. Some of these controls are not included in the baseline controls above because they involve a noticeable decline in sample size due to missing data (see table A3 in the Appendix). Firm-level controls all taken from ES include: dummy variable equal to 1 if the firm competes against unregistered or informal firms and 0 otherwise; firm’s growth proxied by the average annual growth rate of employment over the last 3 years prior the year ES was conducted; a dummy variable equal to 1 if the firm spent on R&D activity last year and 0 otherwise; dummy variable equal to 1 if the firm introduced a new product or product in the last three years and 0 otherwise; and severity level (on 0-4 scale) of labor regulations as an obstacle to firm’s current operations. Macro-level controls include the square of (log of) GDP per capita for any non-liner effects of higher incomes; ease of entry proxied by Fraser Insitutues sub-index of Freedom to Start a business; and social norms proxied by the percentage of country’s population that is Muslim, Catholic, Protestant (residual category is all other religions) taken from Maoz and Henderson (2013). 3.8 Interaction term controls As stated above, we focus on three heterogenous effects of maternity leave. These relate to how the impact of maternity leave depends on who pays for maternity leave (government or employer), share of females in 20-29 years age group in the country’s total population, and the skill level of workers proxied by the wage rate and labor productivity of the firm. To ensure that these heterogeneities do not proxy for other heterogenous effects, we control for several interaction terms as well as non-linear effects of maternity leave, etc. These controls are discussed in section 5 and 6. 25 4. Base regression results 4.1 Base regression results Baseline regression results are provided in Table 1. All the specifications control for industry- fixed effects and year fixed effects. The remaining baseline controls are added sequentially. Table 1 shows that the relationship between female workers and maternity leave is large, positive, and statistically significant at 5 percent level. For each log point increase in the maternity leave, the associated increase in female workers ranges between 2.08 to 2.84 percentage points across the different specifications. Without any other controls (except for year and industry-industry fixed effects), the estimated coefficient value of maternity leave equals 2.84 (column 1). The coefficient value declines to 1.77 (significant at 5 percent level, not shown) when we control for GDP per capita and increases to 1.96 (significant at the 5 percent level, not shown) when we also control for the growth rate of GDP per capita. It increases marginally to 2.08 when we add the remaining firm- and country-level baseline controls (column 5). Thus, for the final baseline specification with all the baseline controls included, female workers’ share increases by about 2.08 percentage points for each log point increase in maternity leave. The estimate implies that increasing maternity leave over its current range (from 0 to 410 days) increases female workers by 12.5 percentage points or by about 0.91 percentage points on average for each additional month of maternity leave days.9 This estimate is almost similar to what we find using the number of days of maternity leave instead of its log values (see section 4.4 below).10 9 This is calculated as follows. Maternity leave days in our sample vary between 0 and 13.7 months. In (1 plus) log terms, the range is between 0 and 6.02. So, based on the final baseline estimate, an increase in maternity leave from its minimum to maximum values (0 to 6.02 log points) increases female workers by 2.08*6.02 or 12.5 percentage points. Since this increase is over 13.7 months, the average monthly increase equals 12.5/13.7 or 0.91 percentage points. 10 Note that an increase in the number of days of maternity leave translates into much bigger increases in the log of maternity leave at low or initial values of maternity leave days than at higher values of it. Thus, our model automatically implies that the impact of an increase in maternity leave by a given number of days, say a month, will have much larger effect on female workers initially than later. Based on the final baseline specification above, figure 26 Several controls show a significant relationship with female workers in the expected direction. Unless stated otherwise, all relationships discussed in this paragraph are significant at 5 percent level or less. Regarding the macro-level controls, the share of female workers is significantly higher in countries that have higher level of GDP per capita, higher growth rate of GDP per capita, higher share of females in total adult population, and greater gender parity in the laws. For the firm-level controls, the share of female workers is significantly higher for firms that have a female top manager, have female owners, have greater foreign ownership, export more (significant at 10 percent and 5 percent level across specifications), and offer training to their workers. It is significantly lower for firms that report access to finance is a relatively bigger obstacle for their operations, and for firms that are larger in size (significant at 10 percent level in some specifications and insignificant in others). Since there is no prior evidence or theory for access to finance and female workers relationship, we confirm that the qualitative nature of the results discussed throughout the paper is preserved even if we do not control for the finance obstacle. For instance, controlling for all the baseline controls except for the finance obstacle, the estimated coefficient value of maternity leave equals 2.06, significant at the 5 percent level. The corresponding value with the finance obstacle control is almost the same equaling 2.08 (column 5). Another issue is the lack of significant relationship between female workers and fertility rate. The lack of significance is entirely due to the control for GDP per capita with which fertility rate is highly correlated. For instance, with all the baseline controls except for GDP per capita included 2 shows how the increase in female workers declines for each additional month of maternity leave over the current range of maternity leave in our sample (0 to 410 days). For instance, the first month of maternity leave leads to 7.13 percentage point increase in female workers. In contrast, the increase in maternity leave from 11 months to 12 months leads to an increase of female workers by only 0.18 percentage points. 27 in the specification, a unit increase in fertility rate is associated with 1.52 percentage points lower value of female workers and this relationship is statistically significant at the 5 percent level. 4.2 Robustness for other lagged values of maternity leave In our baseline, we used values of maternity leave averaged over the last 5 years prior to the ES. We check that our results are robust to alternative lagged values. Thus, we use (log of 1 plus) 1 year lagged values of maternity leave and (log of 1 plus) maternity leave averged over the last 10 years prior the ES. Regression results with these alternative lagged values are provided in table A4 in the Appendix. They show that our main result of a large, positive, and statistically significant relationship (at 5 percent level) between female workers and maternity leave is preserved. 4.3 Robustness for other controls Starting with the final specification in Table 1, we add robustness controls sequentially to the specification. The results are provided in Table 2. We first control for the square of (log of) GDP per capita and find that it makes no noticeable difference to the maternity leave and female workers relationship (column 1). That is, adding the control increases the estimated coefficient value of maternity leave marginally from 2.08 (column 5, table 1) to 2.12 (column 1, table 2) and it remains significant at the 5 percent level. In column 2, we add the controls for the percentage of population in the country that is Catholic, Muslim, and Protestant. The omitted category is all other religions. Adding these controls causes the estimated coefficient value of maternity leave to increase from 2.12 above (column 1) to 2.49 (column 2) and it remains significant at the 5 percent level. Adding the remaining firm- and country-level controls somewhat lowers the estimated coefficient value of maternity leave from 2.49 above (column 2) to 2.24 (column 4), significant at 5 percent level. 28 4.4 Maternity leave days (without logs) We repeat the baseline regressions replacing (log of 1 plus) maternity leave days with the actual number of days of maternity leave (without taking logs). Regression results are provided in table 3. These results are qualitatively similar to the baseline results above. That is, they show a large and positive relationship between maternity leave days and female workers. The relationship is significant at the 1 percent level in all the specifications. For the final baseline specification, an increase in maternity leave by 1 day is associated with an increase in female workers by 0.032 percentage points. This amounts to 0.96 percentage point increase in female workers for each additional month of maternity leave days. The estimate is almost similar to the corresponding estimate of 0.91 percentage point increase based on the final baseline specification (section 4.1). 4.5 Developing country sample Above, we argued that the present study is one of the few ones to focus on developing countries. While our baseline sample includes mostly developing countries, there are a few (non-OECD) high-income countries based on the income classification provided by the World Bank for the year the ES was conducted in the country. To ensure that the positive relationship between maternity leave and female workers is not driven by high-income countries, we repeat the baseline estimation excluding the high-income countries from the sample. The results are provided in table A5 in the Appendix. They show that the relationship between maternity leave and female workers remains postive, large, and statistically significant (at 5 percent level or close to it) for all the baseline specificaitons. In fact, quantitatively, the relationship is much larger for the case of the developing country sample than for the full sample. For instance, for the final baseline specification with all 29 the controls included, the estimated coefficient vlaue of the maternity leave variable equals 2.74 for the developing country sample (column 5, table A5), significant at close to the 5 percent level (p vlaue of .056). The correspnding value for the full sample is much lower at 2.08 (column 5, table 1). 5. Endoegeneity tests 5.1 Government funded maternity leave As discussed above, unlike employer funded maternity leave, a fully government funded maternity leave does not depress employer demand for female workers. Thus, the expectation is that the positive relationship between maternity leave and female workers is much stronger (more positive) when maternity leave is fully funded by the government vs. when it is not. Regression results provided in table 4 strongly confirm this prediction. That is, the interaction term between maternity leave and Govt. Funded is positive, large, and statistically significant at the 5 percent level or less. This holds for all the baseline specifications. It continues to hold when we include the square of maternity leave and several other interaction term controls (column 6). The interaction term controls ensure that our main interaction term between maternity leave and government funding of it does not pick up the effect of other possible heterogeneities between maternity leave and female workers, and between government funding of maternity leave and female workers. Theory offers little help in identifying these other possible heterogeneities. Thus, our choice of interaction term controls here is heuristic, although rich enough to account for several possible heterogeneities. That is, the interaction term controls include those between maternity leave and the following variables: GDP per capita (logs), firm-size or (log of) number of workers at the firm, dummy for the presence of female owner, fertility rate, gender parity in workplace laws, and 30 gender parity in primary education enrollment rate. The other set of interaction term controls include those between Govt. Funded and the following variables: GDP per capita (logs), firm-size or (log of) number of workers at the firm, dummy for the presence of female owner, gender parity in workplace laws, and gender parity in primary education enrollment rate. To get a sense of the magnitude of the heterogeneity, consider the specification with all the baseline controls included (column 5, table 4). For this specification, we find that 1 log point increase in maternity leave is associated with an increase in female workers by a large 3.1 percentage points (significant at 5 percent level) when maternity leave is fully funded by the government. In contrast, when maternity leave is fully or partly paid by the employer, the corresponding change in female workers is a decline of about 0.95 percentage point, although this is not statistically significant at the 10 percent level. Split-sample results are provided in table A6 in the Appendix. Columns 1-5 repeat the baseline regressions for countries where maternity leave is partly or fully funded by the employer while columns 6-10 do the same for countries with fully government funded maternity leave. Consistent with the interaction term results discussed in the previous paragraph, there is no significant relatonship between maternity leave and female workers when maternity leave is partly or fully employer funded. In contrast, there is a large, positive, and significant relationship between the two when maternity leave is fully government funded. 5.2 Maternity leave and females in 20-29 age group In this section, we provide results for how the impact of maternity leave depends on the share of women in the child-bearing age group of 20-29 years. The prediction is that the positive impact of higher maternity leave on female workers is much stronger (more positive) in countries with a 31 higher share of women in the 20-29 age group. Regression results are provided in table 5. As is evident from the table, the interaction term between maternity leave and the share of females in the 20-29 age group is positive, but small and statistically insignificant without any other controls (column 1). It becomes much larger equaling 2.42 and significant at the 5 percent level when we also control for the interaction term between maternity leave and government funding of maternity leave (column 2). It remains almost unchanged, equaling 2.50, when we also control for the share of all females in the adult population and its interaction term with maternity leave (column 3); it rises to 2.94 when we also include the baseline controls in the specification (column 7). Last, we add to the previous set of controls the square of maternity leave, square of the share of females in the 20-29 age group in the country, and several interaction term controls (column 8). The interaction term controls are similar to the ones used in the previous section. They include interaction terms between maternity leave and the following variables: GDP per capita (logs), firm- size or (log of) number of workers at the firm, dummy for the presence of female owner, fertility rate, gender parity in workplace laws, and gender parity in primary education enrollment rate. The other set of interaction term controls include those between the share of females in the 20-29 age group and the following: GDP per capita (logs), firm-size or (log of) number of workers at the firm, dummy for the presence of female owner, gender parity in workplace laws, and gender parity in primary education enrollment rate. Adding these controls causes the estimated coefficient value of the interaction term between maternity leave and the share of females in the 20-29 age group to increase from 2.94 above (column 7) to 3.42 (column 8). The latter estimate is significant at the 1 percent level. We note that the specification discussed in the previous paragraph uses controls for the interaction term between maternity leave and the share of females in the total (adult) population 32 of the country (columns 3-8). This interaction term is small and statsitically insignificant. Two implications follow. First, the larger positive impact of maternity leave in countries with a higher share of females in the 20-29 age group is not because these countries happen to have more females in total population. Second, the positive impact of maternity leave is magnified because of more women in the 20-29 years age group and not because of more women in other age groups. This is an unlikely result if our maternity leave variable were a proxy for other correlated drivers of female workers. Split-sample results based on below median vs. above median values of the share of females in the 20-29 age group are provided in table A7 in the Appendix. The specifications used are the same as in the baseline model discussed above (table 1). The split-sample results are consistent with the ones in table 4. That is, the maternity leave and female workers relationship is positive, large, and significant for countries with above median share of females in the 20-29 age group. For countries with below median values of the share of females in the 20-29 age group, the relationship between maternity leave and female workers is also positive but much smaller and statistically insignificant. Note that unlike the interaction term results discussed in the previous paragrph, the split-sample reuslts do not require any conditioning on government funding of maternity leave. 5.3 Falsification test Our next endogeneity check involves a falsification test. The dependent variable here is a dummy variable equal to 1 if the firm has at least one female owner and 0 otherwise. The prediction is that there is either no relationship or a negative relationship between maternity leave and the presence of a female owner. One issue here is that the ES does not provide information on the extent of 33 female ownership of the firm or on how many female owners are present. To somewhat account for this, we also provide results for a sub-sample of firms that have a single owner (sole- proprietorship firms). For these firms, female ownership is either 0 or 100 percent. The estimation method used is logit and the specifications (controls) are the same as in the baseline model except that we do not include the dummy for the presence of female owners as a control. Regression results are provide in table 6. The coefficient estimates shown are log odds ratios in Panel A and marginal effects in Panel B. For brevity, the results are shown for only some of the baseline specification. Columns 1-3 show the results for the full sample while columns 4-6 show the results for the sample of sole proprietorship firms. In both these samples, we find no evidence of any signifiant positive relatonship between maternity leave and female ownership. In fact, in the full sample, the relationship betwee the two is negative and significant in some of the specifications (column 2) suggesting that by making wage employment more attractive, higher maternity leave may discourage female entrepreneurship. However, this negative relationship is not robust and it becomes insignificant when we add some of the macro-level controls in column 3. For the sole proprietorship sample, the relatonship between maternity leave and the female ownership dummy is positive but statitically insignificant (at 10 percent level) in all the specifications. Thus, we conclude that unlike female emplpoyment, female ownerhsip does not increase with higher duration of maternity leave. The falsification test is passed. 5.4 Results for individual industries Our last endogeneity check involves repeating the baseline regressions for industries that account for a relatively small share of all formal sector female workes in the country. To this end, we repeat the baseline regressions for the manufacture of basic metals industry (industry code 27, ISIC Rev. 34 3.1) that is known to be a male dominted industry (see Cumming et al. 2015). For instance, using the ILO data for the 31 mostly developing countries for which information is available over 2010- 2021, the manufacture of basic metals industry accounts for 0.07 percent (Uruguay, 2020) to 1.3 percent (Bangladesh, 2010) of all formal sector female employment in the country. For the median country, the industry employs 0.17 percent of all formal sector female workers. For robustness we also look at the manufacture of food and beverages industry and the retail sector. For the median country in the ILO data mentioned above, 3.5 percent of all formal sector female workers are employed in the manufacture of food and beverages industry and 11.3 percent in the retail industry.11 As discussed above, presence of reverse causality implies a larger (more positive) relationship between female workers and maternity as we move from the industry with the smallest to the largest share of female workers (manufacture of basic metals industry to food and beverages industry to the retail sector). Regression results for the three industries are provided in table 7. For brevity, only some of the speicfications are shown. Columns 1-2 show the results for for the manufacture of metals, columns 3-4 contain results for food and beverages industry, and columns 5-6 provide results for the retail sector. We find a large, positive, and statistically significant relationship between maternity leave and female workers in all the three industries. Further, there is no evidence that this relatonship becomes stronger (more psotiive) as we move from metals manufacutring to retail sector. In fact, quite the oppsitie holds. For the final baseline specification, the estimated coefficient value of maternity leave equals 3.5 in metals manufacutirng, 3.7 in food and beverages manufacutirng, and a much lower 2.3 in the retail sector. Thus, the endogeneity test is passed. 11 The ES data can be used to estimate the proportion of all formal sector female workers that are employed in an industry. For robustness, we used the ES to identify industries that account for very small shares of all formal female workers in the country’s private sector covered by the ES. Regression results for these industries are like the ones reported below for the manufacture of basic metals. These results are available on request from the authors. 35 6. Extensions 6.1 Maternity leave and skill level In this section, we analyze how the relationship between female workers and maternity leave varies as skill intensity of workers at the firm rises. We also consider how this nexus between maternity leave and skill intensity depends on skills shortage confronting firms. As stated above, we use two proxy measures of skill intensity of workers at the firm. These are labor producitivty and the wage rate (log values) defined at the firm-level. 6.1.1 Wage rate as a proxy for skilled workers Regression results for the interaction term between wage rate and maternity leave are provided in table 8. The specification includes all the baseline specifications with the difference that we now control for the wage rate (log values) instead of labor productivity. As explained above, we also include several interaction term controls which include those between maternity leave and the following variables: GDP per capita (logs), firm-size or (log of) number of workers at the firm, dummy for the presence of female owners, fertility rate, gender parity in workplace laws, and gender parity in primary education enrollment rate. We also control for the interaction term between wage rate and gender parity in workplace laws. Regression results in table 8 reveal that the interaction term between maternity leave and wage rate is large, negative, and significant at the 5 percent level or less in all the specifications. This holds even when we control for the square of maternity leave and other interaction terms (column 6). Thus, maternity leave is more effective in improving female employment for the relatively low-skilled workers. To get a sense of the magnitude, consider the specificaiton in 36 column 5 that includes all the baseline controls. An increase in maternity leave by 1 log point is associated with an increase in female workers by 4.80 percentage points for firms at the 25th percentile value of wage rate, significant at the 1 percent level. The corresponding increase at the 75th percentile value of wage rate is a much smaller equaling 1.47 percentage points, significant at the 10 percent level. Split-sample results based on below vs. above median values of wage rate are provided in table A8 in the Appendix. These results are qualitatively similar to the ones based on interactions. That is, the relationship between maternity leave and female workers is positive, large, and significant (at 5 percent level) for below median values of wage rate. It is also positive, large, and significant for above median values of wage rate but much smaller in magnitude. Next, we provide some suggestive evidence of the role of demand side factors in affecting the maternity leave and skill-level nexus. We argued above that employers would like to retain their high skilled women workers regardless of maternity leave provisions. For the low skilled women workers, employers would retain them post childbirth only when they are required to do so due to maternity leavel laws. This partly explains our finding above that maternity leave has a larger positive impact on employment of females in relatively low-wage firms. Arguably, employers are more keen to retain their skilled workers when there is a greater shortage of skilled workers. Thus, our suggested explanation applies more strongly when employers face a greater shortage of skilled workers. To test for this, we re-estimate the interaction term between materntiy leave and wage rate (as above) separately for the sample of firms that report the shortage of adequately educated workers as no, minor or moderate obstacle for their current operations and for firms that report it as a major or very severe obstacle. The prediction is that the interaction term between maternity leave and wage rate is more negative for firms that report the obstacle as major or very severe obstacle than the rest. The regression results provided in table 9 confirm this 37 prediction. For instance, for the final specification with all the baseline controls and the itneraction term controls included, the estimated coefficient value of the interaction term between wage rate and maternity leave equals -1.2 (significant at 10 percent level, column 3) for firms reporting skills shortage as no, minor or moderate obstacle. The corresponding figure for firms that report skills shortage to be major or very severe obstacle is much larger equaling -1.7 (significant at 10 percent level, column 6). 6.1.2 Labor productivity as a proxy for skilled workers Regression results for the interaction term between labor productivty and maternity leave are qualitatively similar as the ones discussed in the previous section for wage rate. These results are available on request from the authors. 6.2 Adding all interaction terms simultaneously Last, we check if the various interaction term results discussed above are separate and independent of each other or capture the same phenomenon. We do so by including all the interaction terms simultaneously in a single regression. The results are provided in table 10. As evident from the table, all the interaction terms maintain their signs and statistical significance. Thus, they are independent and separate effects. 7. Conclusion Women workers can make substantial contributions to developing economies. However, they face several hurdles in accessing the formal labor markets. Childbearing and childrearing activities put a significant demand on women’s time, forcing many of them to drop out of the formal labor 38 market altogether or suffer from job discountinuity. Countries around the world have implemented maternity leave policies to provide a better balance between work and family life. However, the efficacy of these policies in improving women’s employment remains to be properly established, especially for the case of developing countries. Theoretically, maternity leave can either increase or decrease women’s employment. On the one hand, higher maternity leave increases the cost of hiring female workers and therefore discourages female employment. On the other hand, higher maternity leave encourages women to enter the labor market, leading to greater female employment. The net effect of these competing effects determines the overall relationship between maternity leave and female employment. The present paper takes one step in estimating this relationship using firm-level data on 111 developing countries. The results show that a large positive impact of greater maternity leave duration on the share of workers in the private formal sector firms that are female. Several important heterogeneities were uncovered in this relationship that allow us to better understand who benefits more from maternity leave. The findings are important for policy makers in a variety of ways. First, we find that maternity leave increases women’s employment in the formal sector in the developing and emerging countries. However, this is so only when maternity leave is fully funded by the government and not otherwise. Thus, an effective maternity leave policy has budgetary implications that policy makers need to grapple with. Second, maternity leave policies seem to benefit more women with relatively low wages and skills. This is important for the design and targeting of maternity leave. For instance, policy makers may wish to condition the duration and the extent of government funding of maternity leave on the education or income level of the beneficiaries. Such conditioning is more egalitarian and also more effective in improving the overall employment of women in the formal sector. Third, by providing greater job continuity, 39 maternity leave is most likely to improve women’s long-term career progression and help them break the “glass ceiling”. This is an important issue for policy makers because despite improvements in labor market activity, women tend to be concentrated in low-paying and vulnerable jobs. Our preliminary analysis shows that maternity leave is indeed associated with more female top managers in the country. In terms of future research, several issues remain to be explored. We provide a few examples to motivate future research in the area. First, by providing better job continuity, maternity leave helps preserve women’s firm- and industry-specific human capital accumulation. This is likely to help women break the glass ceiling and acquire managerial positions. Thus, it will be interesting to see if maternity leave increases women’s chances of becoming top managers of firms. Second, our analysis focuses squarely on women in the formal or registered private sector. However, opporutnities available in the informal sector are likely to affect how women in the formal sector react to maternity leave. In countries where the informal sector offers better work opporunities, women are likely to be more sensitive to (absence of) maternity leave in the formal sector. The interplay between formal and informal employment in determining the efficacy of maternity and other family leave policies is an important area for future research. Third, we found some indication above that by encouraging wage employment, longer maternity leave may discourage women’s entrepreneurship. This remains to be properly explored especially in the context of the developing and emerging countries. Fourth, we argued above that there are contrasting effects of maternity leave on demand and supply of female workers. Our analysis only captures the net effect of the contrasting forces. It will be useful to isolate and estimate both the positive and negative effects of maternity leave on female workers. Such an exercise can help in targeting appropriate policies for employers vs. women workers. Last, we found that maternity 40 leave benefits more women employed in low labor productivity and low-wage paying firms. This suggests distributional implications of maternity leave that favor the less educated and low-income women. However, data limitations do not allow us to explore the issue is greater detail. This is a fruitful area for future research perhaps using employer-employee matched data that has information on the wage rate, skills, and education of the employees in the formal sector. 41 References Albagli, Pinjas and Rau, Tomás, (2019), The Effects of a Maternity Leave Reform on Children's Abilities and Maternal Outcomes in Chile, Economic Journal, 129, issue 619, p. 1015-1047. Acevedo, P., Cruces, G., Gertler, P., and Martinez, S. (2017). ‘Living up to expectations: How job training made women better off and men worse off’. NBER Working Paper 23264. Cambridge, MA: National Bureau of Economic Research. https://doi.org/10.3386/w23264 Amin, Mohammad, Islam, Asif and Sakhonchik, Alena, (2016), Does paternity leave matter for female employment in developing economies? Evidence from firm-level data, Applied Economics Letters, 23, issue 16, p. 1145-1148. Aslim, Erkmen, Panovska, Irina and Taş, M. Anıl, (2021), Macroeconomic effects of maternity leave legislation in emerging economies, Economic Modelling, 100 (C): 1-21. Baker, M. and K. Milligan (2008a). How Does Job-Protected Maternity Leave Affect Mothers’ Employment? Journal of Labor Economics 26 (4), 655–691. Baum, Charles L. and Ruhm, Christopher, (2016), The Effects of Paid Family Leave in California on Labor Market Outcomes, Journal of Policy Analysis and Management, 35, issue 2, p. 333-356 Bennett, Benjamin, Erel, Isil, Stern, Léa H. and Wang, Zexi, (2020), Paid Leave Pays Off: The Effects of Paid Family Leave on Firm Performance, No 27788, NBER Working Papers, National Bureau of Economic Research. Berger, Lawrence M. and Waldfogel, Jane, (2004), Maternity leave and the employment of new mothers in the United States, Journal of Population Economics, 17, issue 2, p. 331-349. Bertrand, Marianne, Sandra E Black, Sissel Jensen, Adriana Lleras-Muney (2019), Breaking the Glass Ceiling? The Effect of Board Quotas on Female Labour Market Outcomes in Norway, The Review of Economic Studies, Volume 86 (1): 191–239. https://doi.org/10.1093/restud/rdy032 Besamusca, Janna, Tijdens, Kea, Keune, Maarten and Steinmetz, Stephanie, (2015), Working Women Worldwide. Age Effects in Female Labor Force Participation in 117 Countries, World Development, 74, issue C, p. 123-141. Bloom, D.E., Canning, D., Fink, G. et al. Fertility, female labor force participation, and the demographic dividend. J Econ Growth 14, 79–101 (2009). https://doi.org/10.1007/s10887-009- 9039-9 Borrowman, Mary & Stephan Klasen (2020) Drivers of Gendered Sectoral and Occupational Segregation in Developing Countries, Feminist Economics, 26:2, 62-94, DOI: 10.1080/13545701.2019.1649708 Bossler, M., Mosthaf, A. and Schank, T. (2020). Are female managers more likely to hire more female managers? Evidence from Germany. Industrial and Labor Relations Review, 73, 190–222. 42 Bratti, M., Del Bono, E. and Vuri, D. (2005), "New mothers' labour force participation in Italy: The role of job characteristics", Labour: Review of Labour Economics and Industrial Relations, Vol. 19 No. S1, pp. 79-121. Bussemakers, Carlijn, van Oosterhout, Kars, Kraaykamp, Gerbert and Spierings, Niels, (2017), Women’s Worldwide Education–employment Connection: A Multilevel Analysis of the Moderating Impact of Economic, Political, and Cultural Contexts, World Development, 99, issue C, p. 28-41. Chang, M.L. (2004) Growing Pains: Cross-National Variation in Sex Segregation in Sixteen Developing Countries. American Sociological Review, 69(1): 114-137. doi:10.1177/000312240406900107 Chatopadhyay, R., & Duflo, E. (2004). Women as policy makers: Evidence from a randomized policy experiment in India. Econometrica, 72(5), 1409–1443. https://doi.org/10.1111/j.1468-0262.2004.00539x Chen, Xi and Ge, Suqin, (2018), Social norms and female labor force participation in urban China, Journal of Comparative Economics, 46, issue 4, p. 966-987. Cooke, Dudley, Fernandes, Ana P. and Ferreira, Priscila, (2019), Product market competition and gender discrimination, Journal of Economic Behavior & Organization, 157, issue C, p. 496-522 Cumming, Douglas, T. Y. Leung, and Oliver Rui, (2015), Gender Diversity and Securities Fraud, Academy of Management 58: 1572-1593. https://doi.org/10.5465/amj.2013.0750 Da Mata, Daniel, Oliveira, Rodrigo C. and Silva, Diana, (2021), Who benefits from job training programmes?: Evidence from a high-dosage programme in Brazil, No wp-2021-121, WIDER Working Paper Series, World Institute for Development Economic Research (UNU-WIDER) Dahl, Gordon, Løken, Katrine, Mogstad, Magne and Salvanes, Kari Vea, (2016), What Is the Case for Paid Maternity Leave?, The Review of Economics and Statistics, 98, issue 4, p. 655-670. David Card, Jochen Kluve, Andrea Weber, What Works? A Meta Analysis of Recent Active Labor Market Program Evaluations, Journal of the European Economic Association, Volume 16, Issue 3, June 2018, Pages 894–931, https://doi.org/10.1093/jeea/jvx028 David Card, Ana Rute Cardoso, Patrick Kline, Bargaining, Sorting, and the Gender Wage Gap: Quantifying the Impact of Firms on the Relative Pay of Women , The Quarterly Journal of Economics, Volume 131, Issue 2, May 2016, Pages 633–686, https://doi.org/10.1093/qje/qjv038 Del Boca, Daniela, Silvia Pasqua, Chiara Pronzato (2009), Motherhood and market work decisions in institutional context: a European perspective, Oxford Economic Papers, 61(suppl_1): i147– i171. https://doi.org/10.1093/oep/gpn046 43 Del Rey, Elena, Andreas Kyriacou, and José I. Silva. 2021. “Maternity Leave and Female Labor Force Participation: Evidence from 159 Countries.” Journal of Population Economics 34 (3): 803– 24. doi.org /10.1007/s00148-020-00806-1. Dildar, Yasemin (2015), Patriarchal Norms, Religion, and Female Labor Supply: Evidence from Turkey, World Development, 76, issue C, p. 40-61. Druedahl, Jeppe, Ejrnæs, Mette and Jørgensen, Thomas, (2019), Earmarked paternity leave and the relative income within couples, Economics Letters, 180, issue C, p. 85-88. Duchin, R., O. Ozbas, B. A. Sensoy (2010). “Costly External Finance, Corporate Investment, and the Subprime Mortgage Credit Crisis,” Journal of Financial Economics 97(3): 418–435. Fallon, Kathleen M., Mazar, Alissa and Swiss, Liam, (2017), The Development Benefits of Maternity Leave, World Development, 96, issue C, p. 102-118 Fernández, Raquel, Fogli, Alessandra and Olivetti, Claudia, (2004), Mothers and Sons: Preference Formation and Female Labor Force Dynamics, The Quarterly Journal of Economics, 119, issue 4, p. 1249-1299. Gaddis, I., Klasen, S. (2014), Economic development, structural change, and women’s labor force participation:. J Popul Econ 27, 639–681. https://doi.org/10.1007/s00148-013-0488-2 Goldin C (1995) The U-shaped female labor force function in economic development and economic history. In: Schultz TP (ed) Investment in women’s human capital. University of Chicago Press, Chicago, pp 61–90. Goldin, Claudia. 2006. "The Quiet Revolution That Transformed Women's Employment, Education, and Family." American Economic Review, 96 (2): 1-21.DOI: 10.1257/000282806777212350 Heath, Rachel. 2017. “Fertility at Work: Children and Women's Labor Market Outcomes in Urban Ghana.” Journal of Development Economics. 126: 190-214. Hyland, Marie, Djankov, Simeon and Goldberg, Pinelopi, (2020), Gendered Laws and Women in the Workforce, American Economic Review: Insights, 2, issue 4, p. 475-90 İlkkaracan, İpek, (2012), Why so Few Women in the Labor Market in Turkey?, Feminist Economics, 18, issue 1, p. 1-37. Islam, Asif, Silvia Muzi & Mohammad Amin (2019) Unequal Laws and the Disempowerment of Women in the Labour Market: Evidence from Firm-Level Data, The Journal of Development Studies, 55:5, 822-844, DOI: 10.1080/00220388.2018.1487055 44 Jain, Ritika (2021). "Information and Communication Technology Adoption and the Demand for Female Labor: The Case of Indian Industry" The B.E. Journal of Economic Analysis & Policy, 21(2): 695-722. https://doi.org/10.1515/bejeap-2020-0295 Jayachandran, Seema, (2021), Social Norms as a Barrier to Women’s Employment in Developing Countries, IMF Economic Review, 69, issue 3, p. 576-595. Juhn, Chinhui, Gergely Ujhelyi, Carolina Villegas-Sanchez (2014), “Men, Women, and Machines: How Trade Impacts Gender Inequality,” Journal of Development Economics, 106: 179-193. Karshenas, M., Moghadam, V.M., & Alami, R. (2014). Social policy after the Arab Spring: States and social rights in the MENA Region. World Development, 64, 726–739. Klasen, Stephan (2019), What Explains Uneven Female Labor Force Participation Levels and Trends in Developing Countries?, The World Bank Research Observer, 34(2): 161–197. https://doi.org/10.1093/wbro/lkz005 Klasen, Stephan, TU THI NGOC Le, Janneke Pieters, and Manuel Santos Silva (2021) What Drives Female Labour Force Participation? Comparable Micro-level Evidence from Eight Developing and Emerging Economies, The Journal of Development Studies, 57(3): 417- 442, DOI: 10.1080/00220388.2020.1790533 Kodama, N, Javorcik, BS, Abe, Y. (2018), Transplanting corporate culture across international borders: Foreign direct investment and female employment in Japan. World Economy 41: 1148– 1165. https://doi.org/10.1111/twec.12612 Lee, Kye Woo and Cho, Kisuk, (2005), Female labour force participation during economic crises in Argentina and the Republic of Korea, International Labour Review, 144, issue 4, p. 423-450. Mammen K, Paxson C (2000) Women’s work and economic development. J Econ Perspect 14(4):141–164. Macmillan, R., Shofia, N., & Sigle, W. (2018). Gender and the politics of death: Female representation, political and developmental context, and population health in a cross-national panel. Demography, 55(5), 1905–1934. Maoz, Z., & Henderson, E. A. (2013). The world religion dataset, 1945-2010: Logic, estimates, and trends. International Interactions, 39(3), 265–291. Mishra, Ankita, Mishra, Vinod and Parasnis, Jaai, (2021), The asymmetric role of crime in women's and men's labour force participation: Evidence from India, Journal of Economic Behavior & Organization, 188, issue C, p. 933-961. Nelson, Robert L. and William P. Bridges (1999), Legalizing Gender Inequality: Courts, Markets, and Unequal Pay for Women in America, Cambridge University Press. 45 Nunn, Nathan (2007). “Relationship-Specificity, Incomplete Contracts, and the Pattern of Trade,” The Quarterly Journal of Economics 122(2): 569–600. Olivetti, Claudia and Petrongolo, Barbara, (2014), Gender gaps across countries and skills: Demand, supply and the industry structure, Review of Economic Dynamics, 17, issue 4, p. 842- 859. https://doi.org/10.1016/j.red.2014.03.001 Olivetti, Claudia, and Barbara Petrongolo. 2017. "The Economic Consequences of Family Policies: Lessons from a Century of Legislation in High-Income Countries." Journal of Economic Perspectives, 31 (1): 205-30.DOI: 10.1257/jep.31.1.205 Ouedraogo, Rasmane and Marlet, Elodie, (2018), Foreign Direct Investment and Women Empowerment: New Evidence on Developing Countries, No 2018/025, IMF Working Papers, International Monetary Fund. Rajan, R. G., and L. Zingales (1998). “Financial Dependence and Growth,” American Economic Review 88(3): 559-586. Rossin-Slater, Maya (2017), Maternity and Family Leave Policy, No 23069, NBER Working Papers, National Bureau of Economic Research. Ruhm, Christopher, (1998), The Economic Consequences of Parental Leave Mandates: Lessons from Europe, The Quarterly Journal of Economics, 113, issue 1, p. 285-317 Schmieder, Julia (2021), Fertility as a driver of maternal employment, Labour Economics, 72(C): 1-16. Tunalı, İnsan, Murat Kırdar, and Metlem Dayioglu (2021), Down and up the “U” – A synthetic cohort (panel) analysis of female labor force participation in Turkey, 1988–2013, World Development, 146(C): 1-17. Uribe, Ana Maria Tribin, Vargas, Carmina and Bustamante, Natalia Ramírez, (2019), Unintended consequences of maternity leave legislation: The case of Colombia, World Development, 122, issue C, p. 218-232. Viollaz, M., and H. Winkler. 2020. Does the Internet Reduce Gender Gaps?: The Case of Jordan. Washington, D.C., USA: The World Bank Working Paper.10.1596/1813-9450-9183 Vo, Thang T. and Ha, Truong Thiet, (2021), Decomposition of gender bias in enterprise employment: Insights from Vietnam, Economic Analysis and Policy, 70, issue C, p. 182-194. https://doi.org/10.1016/j.eap.2021.02.007 World Bank (2011). World Development Report 2012: Gender Equality and Development. Washington, D.C.: World Bank. 46 World Bank (2020), Women and Trade: The Role of Trade in Promoting Gender Equality , World Bank, Washington DC. World Bank. 2022. Women, Business and the Law 2022. Washington, DC: World Bank. doi:10.1596/978-1-4648-1817-2. License: Creative Commons Attribution CC BY 3.0 IGO Zhang, Jian, Jin, Songqing, Li, Tao and Wang, Haigang, (2021), Gender discrimination in China: Experimental evidence from the job market for college graduates, Journal of Comparative Economics, 49, issue 3, p. 819-835. Zveglich, Joseph and Rodgers, Yana, (2003), The Impact of Protective Measures for Female Workers, Journal of Labor Economics, 21, issue 3, p. 533-556. 47 Figure 1: Female workers and maternity leave relationship Source: Author’s own calculations based on Enterprise Surveys data. 48 Figure 2: Impact of maternity leave on Female workers over the current range of maternity leave 8.00 % point increase in Female workers 7.13 7.00 6.00 5.00 4.00 3.00 2.00 1.41 0.83 0.59 1.00 0.46 0.38 0.32 0.28 0.24 0.22 0.20 0.18 0.17 0.15 0.00 Increase in maternity leave days Source: Author’s own calcualtions using ES data. Note: The values shows are based on the estimated coefficient value of Maternity leave (logs) variable in the final baseline specification (column 5, table 1). 49 Table 1: Base regression results (OLS) Dependent variable: Female (1) (2) (3) (4) (5) workers Maternity leave (logs) 2.845** 2.068** 2.129** 2.193** 2.076** (1.320) (0.922) (0.878) (0.869) (0.876) GDP per capita (logs) 4.606*** 4.766*** 4.555*** 3.360** (0.555) (0.551) (0.547) (1.373) Growth rate of GDP per 0.513*** 0.515*** 0.519*** 0.445*** capita (%, annual) (0.181) (0.175) (0.174) (0.164) Females in adult 1.674*** 1.722*** 1.718*** 1.567*** population (%) (0.475) (0.472) (0.469) (0.458) Number of workers (logs) -0.025 -0.385 -0.503* -0.466* (0.299) (0.282) (0.290) (0.279) Female owner present Y:1 4.984*** 4.949*** 4.899*** 4.481*** N:0 (0.500) (0.493) (0.488) (0.490) Top manager is female 13.300*** 13.172*** 13.166*** 13.046*** Y:1 N:0 (0.801) (0.795) (0.793) (0.774) Age of firm (logs) -0.237 -0.293 -0.380 (0.382) (0.391) (0.387) Labor productivity (logs) -0.220 -0.267 -0.236 (0.194) (0.200) (0.187) Exports (% of sales) 0.029** 0.029** 0.023* (0.014) (0.014) (0.013) Foreign ownership (%) 0.023** 0.022** 0.023** (0.009) (0.009) (0.009) Firm offers formal training 2.282*** 2.168*** 1.938*** to workers Y:1 N:0 (0.572) (0.567) (0.560) Inadequately educated -0.232 -0.179 -0.125 workers: Obstacle severity (0-4 scale) (0.189) (0.203) (0.206) Firm registered when -0.179 -0.343 started operations Y:1 N:0 (1.024) (0.999) Firm experienced losses 0.672 0.701 due to crime Y:1 N:0 (0.611) (0.614) Access to finance: Obstacle -0.589*** -0.505** severity (0-4 scale) (0.225) (0.217) Tax rates: Obstacle 0.294 0.303 severity (0-4 scale) (0.232) (0.223) Firm has own website Y:1 0.586 0.469 N:0 (0.550) (0.538) Fertility rate -0.189 (0.966) Gender parity in 0.064*** 50 workplace laws (0.022) Gender parity in primary 13.774 education enrollment (12.732) Freedom from business 0.276 regulations (0.527) Urbanization rate (%) -0.003 (0.040) Women in Parliament (%) 0.089 (0.063) Paternity leave (logs) -0.065 (0.467) Industry fixed effects Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Constant 27.379*** -109.518*** -112.471*** -109.748*** -111.316*** (8.096) (27.471) (27.182) (27.073) (33.556) Number of observations 80,272 80,272 80,272 80,272 80,272 R-squared 0.197 0.279 0.282 0.283 0.290 OLS method used. Huber-White robust standard errors clustered on country-year pair (ES round) in brackets. *** (1%), ** (5%), * (10%). 51 Table 2: Robustness for additional controls Dependent variable: Female (1) (2) (3) (4) workers Maternity leave (logs) 2.124** 2.491** 2.285** 2.238** (0.896) (1.066) (1.019) (1.015) GDP per capita (logs) 26.987** 35.352*** 23.952* 23.436* (10.962) (12.750) (14.208) (14.004) Growth rate of GDP per 0.410** 0.350** 0.603*** 0.633*** capita (%, annual) (0.162) (0.150) (0.181) (0.180) Females in adult population 1.581*** 0.951** 1.192** 1.224** (%) (0.459) (0.470) (0.566) (0.572) Number of workers (logs) -0.495* -0.621** -0.449 -0.486 (0.278) (0.291) (0.318) (0.320) Female owner present Y:1 4.417*** 3.997*** 4.282*** 4.258*** N:0 (0.481) (0.514) (0.550) (0.553) Top manager is female Y:1 13.059*** 12.195*** 13.213*** 13.301*** N:0 (0.768) (0.805) (0.842) (0.848) Age of firm (logs) -0.332 -0.464 -0.673 -0.695 (0.384) (0.412) (0.480) (0.483) Labor productivity (logs) -0.194 -0.271 -0.069 -0.071 (0.186) (0.190) (0.195) (0.196) Exports (% of sales) 0.024* 0.018 0.015 0.015 (0.013) (0.014) (0.014) (0.014) Foreign ownership (%) 0.021** 0.014 0.011 0.011 (0.009) (0.009) (0.010) (0.010) Firm offers formal training 1.861*** 1.365** 0.771 0.823 to workers Y:1 N:0 (0.548) (0.555) (0.583) (0.584) Inadequately educated -0.109 -0.201 -0.289 -0.407* workers: Obstacle severity (0-4 scale) (0.203) (0.215) (0.225) (0.244) Firm registered when started -0.300 -0.991 0.426 0.364 operations Y:1 N:0 (0.998) (1.012) (0.815) (0.801) Firm experienced losses due 0.767 0.713 1.812** 1.804** to crime Y:1 N:0 (0.603) (0.624) (0.765) (0.767) 52 Access to finance: Obstacle -0.539** -0.511** -0.517** -0.554** severity (0-4 scale) (0.216) (0.221) (0.232) (0.232) Tax rates: Obstacle severity 0.326 0.345 0.542** 0.442* (0-4 scale) (0.223) (0.239) (0.248) (0.248) Firm has own website Y:1 0.526 0.475 0.118 0.099 N:0 (0.536) (0.554) (0.512) (0.516) Fertility rate 0.305 1.911** 1.448 1.437 (0.985) (0.868) (0.898) (0.904) Gender parity in workplace 0.068*** 0.056*** 0.073*** 0.072*** laws (0.021) (0.019) (0.026) (0.026) Gender parity in primary 9.401 7.837 4.083 2.690 education enrollment (13.435) (15.454) (21.176) (21.453) Freedom from business 0.432 0.665 0.502 0.254 regulations (0.520) (0.574) (0.708) (0.777) Urbanization rate (%) -0.006 0.020 0.048 0.048 (0.040) (0.041) (0.049) (0.051) Women in Parliament (%) 0.102 0.079 0.044 0.040 (0.063) (0.069) (0.079) (0.079) Paternity leave (logs) 0.066 -0.113 -0.189 -0.185 (0.440) (0.472) (0.558) (0.548) Square of GDP per capita -1.311** -1.734** -1.205 -1.182 (logs) (0.579) (0.689) (0.771) (0.760) % of population that is -2.252 -1.553 -1.464 Catholic (1.895) (2.349) (2.310) % of population that is 0.984 -1.373 -1.514 Protestant (4.398) (5.737) (5.717) % of population that is -9.338*** -8.193*** -8.307*** Muslim (2.339) (2.746) (2.737) Annual employment growth -0.077*** -0.077*** (%) (0.014) (0.014) Informal competition Y:1 -1.433** -1.445** N:0 53 (0.558) (0.554) Firm spent on R&D Y:1 N:0 0.007 0.006 (0.008) (0.008) Introduced new product Y:1 0.004 0.004 N:0 (0.006) (0.006) How Much Of An Obstacle: 0.446* Labor Regulations? (0.264) Ease of Starting a Business 0.499 (EFW, Fraser Institute)) (0.477) Industry fixed effects Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Constant -217.710*** -235.298*** -181.004*** -183.793*** (56.994) (60.993) (66.981) (66.279) Number of observations 80,272 69,136 53,440 53,230 R-squared 0.291 0.293 0.290 0.290 OLS method used. Huber-White robust standard errors clustered on country-year pair (ES round) in brackets. *** (1%), ** (5%), * (10%). 54 Table 3: Number of days of paid maternity leave Dependent variable: Female (1) (2) (3) (4) (5) workers Number of days of paid 0.041*** 0.034*** 0.033*** 0.033*** 0.032*** maternity leave (0.010) (0.008) (0.008) (0.008) (0.008) GDP per capita (logs) 4.392*** 4.557*** 4.353*** 3.394** (0.548) (0.546) (0.542) (1.352) Growth rate of GDP per 0.515*** 0.516*** 0.518*** 0.456*** capita (%, annual) (0.180) (0.174) (0.174) (0.166) Females in adult population 1.833*** 1.872*** 1.867*** 1.712*** (%) (0.470) (0.469) (0.467) (0.454) Number of workers (logs) -0.022 -0.373 -0.485* -0.452 (0.298) (0.283) (0.290) (0.279) Female owner present Y:1 5.051*** 5.003*** 4.954*** 4.550*** N:0 (0.498) (0.492) (0.487) (0.488) Top manager is female Y:1 13.299*** 13.186*** 13.185*** 13.069*** N:0 (0.800) (0.793) (0.790) (0.774) Age of firm (logs) -0.197 -0.255 -0.338 (0.379) (0.387) (0.385) Labor productivity (logs) -0.227 -0.272 -0.247 (0.193) (0.199) (0.186) Exports (% of sales) 0.029** 0.029** 0.023* (0.014) (0.014) (0.013) Foreign ownership (%) 0.020** 0.019** 0.020** (0.009) (0.009) (0.009) Firm offers formal training to 2.205*** 2.096*** 1.873*** workers Y:1 N:0 (0.573) (0.569) (0.562) Inadequately educated -0.184 -0.130 -0.079 workers: Obstacle severity (0-4 scale) (0.190) (0.205) (0.206) Firm registered when started -0.245 -0.380 operations Y:1 N:0 (1.030) (0.999) Firm experienced losses due 0.564 0.600 to crime Y:1 N:0 (0.605) (0.609) Access to finance: Obstacle -0.572** -0.498** severity (0-4 scale) (0.223) (0.215) Tax rates: Obstacle severity 0.287 0.298 (0-4 scale) (0.229) (0.220) Firm has own website Y:1 0.587 0.486 N:0 (0.549) (0.536) Fertility rate 0.027 (0.972) Gender parity in workplace 0.062*** laws 55 (0.021) Gender parity in primary 16.650 education enrollment (12.842) Freedom from business 0.286 regulations (0.528) Urbanization rate (%) -0.005 (0.039) Women in Parliament (%) 0.086 (0.061) Paternity leave (logs) -0.027 (0.471) Industry fixed effects Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Constant 36.421*** -109.229*** -111.443*** -108.422*** -115.495*** (5.572) (26.111) (26.049) (25.972) (32.925) Number of observations 80,272 80,272 80,272 80,272 80,272 R-squared 0.200 0.282 0.284 0.285 0.291 OLS estimation method used. Huber-White robust standard errors clustered on country-year pair (ES round) in brackets. *** (1%), ** (5%), * (10%). 56 Table 4: Government funded maternity leave Dependent variable: Female (1) (2) (3) (4) (5) (6) workers Maternity leave (logs)*Govt. 4.240** 3.874** 3.571** 3.499** 4.043** 13.867*** Funded Y:1 N:0 (2.107) (1.676) (1.659) (1.641) (1.628) (4.512) Maternity leave (logs) -0.923 -0.669 -0.426 -0.316 -0.952 73.381* (1.181) (0.979) (0.996) (0.989) (0.978) (40.344) Govt. Funded Y:1 N:0 -12.516 -15.552** -14.198* -13.892* -15.327** -85.403*** (9.579) (7.486) (7.415) (7.327) (7.284) (26.734) GDP per capita (logs) 4.116*** 4.301*** 4.117*** 2.852** 45.641*** (0.586) (0.578) (0.573) (1.396) (11.361) Growth rate of GDP 0.422** 0.427** 0.430** 0.325* 0.213 per capita (%, annual) (0.181) (0.176) (0.176) (0.166) (0.174) Females in adult 1.651*** 1.691*** 1.688*** 1.484*** 1.408*** population (%) (0.462) (0.462) (0.459) (0.434) (0.399) Number of workers -0.073 -0.400 -0.493* -0.459* -3.901*** (logs) (0.297) (0.281) (0.289) (0.275) (1.413) Female owner present Y:1 N:0 4.989*** 4.950*** 4.907*** 4.470*** 2.604 (0.499) (0.493) (0.489) (0.491) (3.101) Top manager is female Y:1 N:0 13.330*** 13.206*** 13.198*** 13.080*** 13.123*** (0.794) (0.788) (0.786) (0.762) (0.749) Age of firm (logs) -0.208 -0.269 -0.359 -0.113 (0.378) (0.388) (0.383) (0.372) Labor productivity (logs) -0.260 -0.299 -0.283 -0.268 (0.198) (0.204) (0.195) (0.196) Exports (% of sales) 0.030** 0.030** 0.024* 0.021 (0.014) (0.014) (0.013) (0.013) Foreign ownership (%) 0.020** 0.019** 0.019** 0.023*** (0.009) (0.009) (0.009) (0.009) Firm offers formal training to 2.165*** 2.081*** 1.822*** 1.816*** workers Y:1 N:0 (0.565) (0.562) (0.559) (0.549) Inadequately educated workers: -0.235 -0.179 -0.120 -0.042 Obstacle severity (0-4 scale) (0.188) (0.203) (0.205) (0.205) Firm registered when started -0.405 -0.639 -0.400 operations Y:1 N:0 (1.020) (0.988) (0.958) Firm experienced losses due to 0.488 0.503 0.657 crime Y:1 N:0 (0.595) (0.599) (0.597) Access to finance: Obstacle -0.579** -0.503** -0.516** severity (0-4 scale) (0.224) (0.214) (0.204) 57 Tax rates: Obstacle severity (0- 0.300 0.297 0.231 4 scale) (0.229) (0.219) (0.213) Firm has own website Y:1 N:0 0.457 0.306 0.251 (0.548) (0.536) (0.528) Fertility rate 0.097 23.356** (0.969) (11.772) Gender parity in workplace 0.072*** 0.433** laws (0.022) (0.200) Gender parity in primary 19.854 -113.067 education enrollment (12.639) (171.920) Freedom from business 0.204 0.323 regulations (0.517) (0.535) Urbanization rate (%) -0.002 -0.007 (0.036) (0.037) Women in Parliament (%) 0.086 0.078 (0.061) (0.062) Paternity leave (logs) -0.070 0.283 (0.452) (0.448) Square of Maternity leave -0.488 (logs) (0.524) Maternity leave (logs)*GDP -9.429*** per capita (logs) (2.440) Maternity leave (logs)*Number 0.559* of workers (logs) (0.331) Maternity leave (logs)*Female 0.377 owner present Y:1 N:0 (0.691) Maternity leave (logs)*Fertility -5.024* rate (2.574) Maternity leave (logs)*Gender -0.071 parity in workplace laws (0.046) Maternity leave (logs)*Gender 27.781 parity in primary education enrollment (37.391) Govt. Funded Y:1 N:0*GDP 0.526 per capita (logs) (1.643) 58 Govt. Funded Y:1 N:0*Number 1.138* of workers (logs) (0.610) Govt. Funded Y:1 N:0*Female 0.128 owner present Y:1 N:0 (1.196) Govt. Funded Y:1 N:0*Gender 22.335 parity in primary education enrollment (23.477) Govt. Funded Y:1 N:0*Gender -0.062 parity in workplace laws (0.038) Industry fixed effects Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Constant 40.544*** -91.266*** -94.694*** -92.410*** -94.847*** -414.580** (7.647) (25.223) (25.324) (25.254) (30.899) (201.372) Number of observations 80,272 80,272 80,272 80,272 80,272 80,272 R-squared 0.208 0.281 0.284 0.284 0.292 0.299 OLS estimation method used. Huber-White robust standard errors clustered on country-year pair (ES round) in brackets. *** (1%), ** (5%), * (10%). 59 Table 5: Females in 20-29 age group Dependent variable: Female (1) (2) (3) (4) (5) (6) (7) (8) workers Maternity leave 0.538 2.421** 2.502*** 2.759*** 2.733*** 2.737*** 2.939*** 3.423*** (logs)*Females: 20-29 yrs. (0.717) (1.001) (0.828) (0.585) (0.580) (0.578) (0.580) (1.230) old (%) Maternity leave (logs) -1.111 -21.763*** -12.128 -23.202 -24.061 -23.394 -35.886 -23.794 (4.644) (8.304) (32.570) (21.629) (21.804) (21.663) (29.789) (52.162) Females: 20-29 yrs. old (%) -3.420 -11.347** -11.972*** -12.011*** -11.891*** -11.859*** -12.668*** -27.535 (3.358) (4.764) (3.972) (2.743) (2.711) (2.714) (2.693) (24.728) Govt. Funded Y:1 N:0 -39.319*** -43.424*** -43.583*** -41.888*** -41.739*** -45.524*** -69.219*** (12.800) (12.210) (10.152) (10.058) (9.988) (9.737) (21.140) Maternity leave (logs)*Govt. 9.983*** 10.777*** 9.889*** 9.513*** 9.481*** 10.522*** 15.682*** Funded Y:1 N:0 (2.801) (2.661) (2.233) (2.213) (2.196) (2.175) (4.724) Maternity leave (logs)*% of -0.206 -0.023 0.004 -0.008 0.192 0.871 females in total adult population (5 yr. avg.) (0.635) (0.445) (0.446) (0.443) (0.581) (0.640) % of females in total adult 2.436 10.264*** 10.084*** 9.798*** 9.571*** 2.546 population (5 yr. avg.) (3.001) (3.109) (3.119) (3.088) (3.595) (4.678) GDP per capita (logs) 4.195*** 4.357*** 4.191*** 2.946** 28.428 (0.595) (0.583) (0.576) (1.421) (20.747) Growth rate of GDP per 0.399** 0.406** 0.408** 0.305* 0.245 capita (%, annual) (0.176) (0.171) (0.171) (0.160) (0.156) Females in adult population -8.539** -8.433** -8.101** -8.982** -5.117 (%) (3.826) (3.839) (3.799) (3.593) (4.019) Number of workers (logs) -0.066 -0.421 -0.519* -0.483* -4.856 (0.294) (0.278) (0.287) (0.274) (3.189) Female owner present Y:1 5.158*** 5.100*** 5.055*** 4.599*** 6.700 N:0 (0.496) (0.491) (0.487) (0.484) (4.766) Top manager is female Y:1 13.277*** 13.166*** 13.163*** 13.063*** 13.089*** N:0 (0.795) (0.790) (0.788) (0.760) (0.760) Age of firm (logs) -0.052 -0.114 -0.187 -0.059 (0.373) (0.383) (0.375) (0.373) Labor productivity (logs) -0.255 -0.291 -0.282 -0.242 (0.200) (0.206) (0.195) (0.188) Exports (% of sales) 0.031** 0.031** 0.025* 0.024* (0.013) (0.013) (0.013) (0.013) Foreign ownership (%) 0.019** 0.018** 0.018** 0.018** (0.009) (0.009) (0.009) (0.008) Firm offers formal training to 2.103*** 2.013*** 1.727*** 1.838*** workers Y:1 N:0 (0.568) (0.561) (0.550) (0.524) Inadequately educated -0.158 -0.098 -0.041 -0.011 workers: Obstacle severity (0- 4 scale) (0.191) (0.206) (0.207) (0.204) Firm registered when started -0.432 -0.639 -0.437 operations Y:1 N:0 (1.028) (0.976) (0.974) Firm experienced losses due 0.519 0.536 0.643 to crime Y:1 N:0 (0.597) (0.601) (0.587) Access to finance: Obstacle -0.557*** -0.483** -0.548*** severity (0-4 scale) (0.212) (0.201) (0.201) 61 Tax rates: Obstacle severity 0.252 0.240 0.224 (0-4 scale) (0.224) (0.217) (0.222) Firm has own website Y:1 0.495 0.357 0.416 N:0 (0.549) (0.538) (0.541) Fertility rate 0.290 21.056 (0.990) (12.991) Gender parity in workplace 0.069*** -0.280 laws (0.019) (0.363) Gender parity in primary 24.517* 67.958 education enrollment (13.185) (306.924) Freedom from business 0.240 0.362 regulations (0.532) (0.556) Urbanization rate (%) 0.004 0.010 (0.032) (0.033) Women in Parliament (%) 0.088 0.107 (0.061) (0.065) Paternity leave (logs) 0.251 0.669 (0.459) (0.465) Square of Maternity leave -0.956 (logs) (0.718) Square of Females: 20-29 yrs. 1.334** old (%) (0.548) Maternity leave (logs)*GDP -7.049** per capita (logs) 62 (3.248) Maternity leave 0.934** (logs)*Number of workers (logs) (0.404) Maternity leave 0.289 (logs)*Female owner present Y:1 N:0 (0.639) Maternity leave -4.428 (logs)*Fertility rate (2.786) Maternity leave -0.003 (logs)*Gender parity in workplace laws (0.048) Maternity leave 28.696 (logs)*Gender parity in primary education enrollment (39.037) Females: 20-29 yrs. old 0.798 (%)*GDP per capita (logs) (0.908) Females: 20-29 yrs. old -0.001 (%)*Number of workers (logs) (0.236) Females: 20-29 yrs. old -0.436 (%)*Female owner present Y:1 N:0 (0.433) 63 Females: 20-29 yrs. old -20.469 (%)*Gender parity in primary education enrollment (23.692) Females: 20-29 yrs. old 0.042* (%)*Gender parity in workplace laws (0.024) Industry fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Constant 53.903** 138.421*** 20.159 6.794 8.006 7.465 52.227 26.070 (23.227) (40.945) (153.277) (98.700) (99.623) (99.046) (143.164) (317.728) Number of observations 80,272 80,272 80,272 80,272 80,272 80,272 80,272 80,272 R-squared 0.198 0.210 0.214 0.286 0.288 0.289 0.297 0.301 OLS estimation method used. Huber-White robust standard errors clustered on country-year pair (ES round) in brackets. *** (1%), ** (5%), * (10%). 64 Table 6: Falsification test using female ownership Dependent variable: Female (1) (2) (3) (4) (5) (6) owner present Y:1 N:0 All firms Sole proprietorship firms Panel A: Log odds ratios Maternity leave (logs) -0.017 -0.114** -0.099 0.127 0.094 0.107 (0.044) (0.049) (0.068) (0.082) (0.074) (0.098) GDP per capita (logs) 0.114** -0.123 0.117 0.074 (0.057) (0.129) (0.083) (0.207) Growth rate of GDP per 0.005 -0.005 0.021 0.027 capita (%, annual) (0.017) (0.018) (0.037) (0.037) Females in adult population -0.007 -0.014 0.105 0.127* (%) (0.035) (0.037) (0.064) (0.069) Number of workers (logs) 0.062** 0.074*** 0.087 0.078 (0.027) (0.026) (0.067) (0.063) Top manager is female Y:1 2.647*** 2.639*** 3.544*** 3.563*** N:0 (0.091) (0.092) (0.145) (0.145) Age of firm (logs) 0.230*** 0.224*** -0.095 -0.105 (0.040) (0.038) (0.081) (0.080) Labor productivity (logs) 0.013 0.017 -0.047 -0.054 (0.018) (0.017) (0.037) (0.036) Exports (% of sales) 0.001 0.000 0.002 0.002 (0.001) (0.001) (0.003) (0.003) Foreign ownership (%) -0.005*** -0.005*** -0.007* -0.007* (0.001) (0.001) (0.004) (0.004) Firm offers formal training 0.249*** 0.226*** 0.205 0.197 to workers Y:1 N:0 (0.058) (0.057) (0.141) (0.131) Inadequately educated 0.024 0.027 0.085** 0.078* workers: Obstacle severity (0-4 scale) (0.020) (0.020) (0.043) (0.041) Firm registered when started 0.053 0.024 0.070 0.080 operations Y:1 N:0 (0.086) (0.081) (0.116) (0.111) Firm experienced losses due 0.062 0.064 -0.009 -0.030 to crime Y:1 N:0 (0.070) (0.071) (0.184) (0.174) Access to finance: Obstacle -0.033* -0.019 0.008 0.023 severity (0-4 scale) (0.019) (0.018) (0.046) (0.045) Tax rates: Obstacle severity -0.002 -0.007 0.024 0.009 (0-4 scale) (0.024) (0.023) (0.042) (0.043) Firm has own website Y:1 0.173*** 0.154*** 0.178 0.162 N:0 (0.055) (0.054) (0.119) (0.115) Fertility rate -0.150* -0.018 (0.088) (0.129) Gender parity in workplace 0.004* 0.002 laws (0.002) (0.003) Gender parity in primary -0.330 -0.767 education enrollment (1.361) (1.861) Freedom from business 0.041 -0.075 regulations (0.060) (0.089) Urbanization rate (%) 0.000 0.004 (0.004) (0.006) Women in Parliament (%) 0.008 0.018** (0.006) (0.009) Paternity leave (logs) 0.104** 0.179** (0.044) (0.090) Industry fixed effects Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Constant -2.341*** -4.667* -2.135 -3.211** -10.475** -10.445* (0.906) (2.395) (3.229) (1.541) (4.089) (5.660) Number of observations 80,261 80,261 80,261 25,244 25,244 25,244 Panel B: Marginal effects Maternity leave (logs) -0.004 -0.026** -0.022 0.025 0.017 0.020 (0.010) (0.011) (0.015) (0.016) (0.013) (0.018) Logit estimation used. Huber-White robust standard errors clustered on country-year pair (ES round) in brackets. *** (1%), ** (5%), * (10%). 66 Table 7: Results for individual industries Dependent variable: (1) (2) (3) (4) (5) (6) Female workers Manufacture of basic Manufacture of food and Retail sector (ISIC Rev. metals (ISIC Rev. 3.1 beverages (ISIC Rev. 3.1 3.1 Industry: 52) industry: 27) industry: 15) Maternity leave (logs) 3.118* 3.541* 5.940*** 3.709** 2.878* 2.341** (1.798) (2.001) (2.095) (1.752) (1.673) (1.138) GDP per capita (logs) 1.687 2.120 5.701*** (4.169) (2.661) (1.909) Growth rate of GDP 0.382 0.144 0.646** per capita (%, annual) (0.526) (0.338) (0.267) Females in adult 0.570 1.449* 2.018*** population (%) (1.003) (0.847) (0.735) Number of workers -0.944 -1.109* 1.108* (logs) (0.960) (0.647) (0.603) Female owner present 3.643 3.819*** 5.865*** Y:1 N:0 (2.704) (1.443) (1.099) Top manager is female 2.399 13.778*** 20.748*** Y:1 N:0 (3.861) (1.871) (1.174) Age of firm (logs) -0.374 -1.688** 0.151 (1.913) (0.762) (0.801) Labor productivity (logs) 0.752 -2.157*** -0.289 (0.718) (0.494) (0.348) Exports (% of sales) -0.061 0.013 -0.123*** (0.050) (0.038) (0.042) Foreign ownership (%) 0.098** -0.001 0.045** (0.047) (0.023) (0.019) Firm offers formal training -1.963 2.752* 0.388 to workers Y:1 N:0 (2.600) (1.402) (1.131) Inadequately educated -0.869 1.280** -0.014 workers: Obstacle severity (0-4 scale) (0.954) (0.611) (0.408) Firm registered when -9.454 -0.012 1.198 started operations Y:1 N:0 (6.503) (1.821) (1.841) Firm experienced losses 0.085 -1.397 7.513*** due to crime Y:1 N:0 (3.124) (1.454) (1.307) Access to finance: Obstacle -1.477* -0.469 -1.715*** severity (0-4 scale) 67 (0.880) (0.514) (0.477) Tax rates: Obstacle -1.122 0.269 0.176 severity (0-4 scale) (0.852) (0.556) (0.436) Firm has own website Y:1 -3.885 2.307 -5.063*** N:0 (2.733) (1.708) (1.094) Fertility rate 0.708 -4.066** 0.593 (2.237) (1.709) (1.252) Gender parity in workplace 0.078 0.073** 0.089** laws (0.057) (0.034) (0.039) Gender parity in primary -38.464 -11.086 1.648 education enrollment (34.963) (22.530) (20.163) Freedom from business 1.856 1.237 1.003 regulations (1.809) (0.997) (0.959) Urbanization rate (%) -0.023 0.007 -0.039 (0.126) (0.075) (0.063) Women in Parliament (%) 0.074 -0.089 -0.019 (0.163) (0.106) (0.087) Paternity leave (logs) 0.553 0.599 0.127 (1.244) (0.806) (0.766) Year fixed effects Yes Yes Yes Yes Yes Yes Constant 4.646 -7.079 38.727*** -13.303 24.063*** -148.764*** (9.407) (73.102) (10.151) (63.785) (8.103) (48.809) Number of observations 1,398 1,398 9,289 9,289 11,360 11,360 R-squared 0.106 0.210 0.112 0.273 0.062 0.280 OLS method used. Huber-White robust standard errors clustered on country-year pair (ES round) in brackets. *** (1%), ** (5%), * (10%). 68 Table 8: Maternity leave and wage rate interaction term results Dependent variable: (1) (2) (3) (4) (5) (6) Female workers Maternity leave (logs)* -2.001** -1.723** -1.747** -1.746*** -1.785** -1.292** Wage rate (real, logs) (0.805) (0.680) (0.677) (0.664) (0.686) (0.599) Maternity leave (logs) 20.015** 17.123*** 17.406*** 17.437*** 17.515*** 67.795 (7.755) (6.508) (6.442) (6.330) (6.435) (44.700) Wage rate (real, logs) 10.205*** 7.776** 7.717** 7.648** 7.672** 5.815** (3.719) (3.139) (3.127) (3.072) (3.161) (2.659) GDP per capita (logs) 4.484*** 4.592*** 4.428*** 3.758*** 35.549*** (0.627) (0.623) (0.612) (1.388) (12.154) Growth rate of GDP 0.564*** 0.564*** 0.564*** 0.480*** 0.437** per capita (%, annual) (0.189) (0.180) (0.180) (0.168) (0.178) Females in adult 1.691*** 1.746*** 1.735*** 1.630*** 1.753*** population (%) (0.476) (0.472) (0.468) (0.458) (0.468) Number of workers 0.331 -0.160 -0.277 -0.226 -3.348** (logs) (0.262) (0.256) (0.261) (0.255) (1.354) Female owner present 4.869*** 4.801*** 4.760*** 4.359*** 3.573 Y:1 N:0 (0.527) (0.521) (0.517) (0.500) (2.790) Top manager is 13.684*** 13.566*** 13.556*** 13.438*** 13.616*** female Y:1 N:0 (0.801) (0.795) (0.795) (0.780) (0.763) Age of firm (logs) -0.147 -0.114 -0.183 -0.127 (0.388) (0.394) (0.393) (0.392) Exports (% of sales) 0.037** 0.036** 0.030** 0.027** (0.014) (0.014) (0.014) (0.014) Foreign ownership 0.031*** 0.029*** 0.029*** 0.023** (%) (0.010) (0.010) (0.009) (0.009) Firm offers formal 2.493*** 2.385*** 2.148*** 2.091*** training to workers Y:1 (0.607) (0.610) (0.588) (0.585) N:0 Inadequately -0.221 -0.164 -0.112 -0.058 educated workers: (0.207) (0.222) (0.218) (0.220) Obstacle severity (0-4 scale) Firm registered when 1.277 1.118 1.026 started operations Y:1 N:0 (0.811) (0.760) (0.739) Firm experienced 0.701 0.662 0.437 losses due to crime Y:1 (0.633) (0.635) (0.611) N:0 Access to finance: -0.460** -0.390* -0.399* Obstacle severity (0-4 (0.228) (0.215) (0.208) scale) Tax rates: Obstacle 0.185 0.177 0.170 severity (0-4 scale) (0.237) (0.226) (0.222) 69 Firm has own website 0.324 0.201 0.025 Y:1 N:0 (0.564) (0.551) (0.543) Fertility rate 0.024 33.785*** (0.960) (12.352) Gender parity in 0.067*** 0.195 workplace laws (0.022) (0.190) Gender parity in 10.586 -115.999 primary education (15.893) (191.349) enrollment Freedom from 0.424 0.494 business regulations (0.543) (0.548) Urbanization rate (%) -0.006 -0.011 (0.040) (0.040) Women in Parliament (%) 0.103 0.092 (0.064) (0.063) Paternity leave (logs) -0.098 0.278 (0.472) (0.469) Square of Maternity leave -0.069 (logs) (0.588) Maternity leave -6.928*** (logs)*GDP per capita (logs) (2.606) Maternity leave 0.661** (logs)*Number of workers (logs) (0.288) Maternity leave 0.156 (logs)*Female owner present Y:1 N:0 (0.589) Maternity leave -7.322*** (logs)*Fertility rate (2.659) Maternity leave -0.021 (logs)*Gender parity in workplace laws (0.043) Maternity leave 29.146 (logs)*Gender parity in primary education enrollment (41.262) 70 Wage rate (real, -0.006 logs)*Gender parity in (0.008) workplace laws Industry fixed effects Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Constant -55.228 -177.871*** -182.542*** -181.32*** -189.283*** -435.017* (35.268) (37.674) (37.072) (36.774) (44.889) (225.839) Number of observations 72,342 72,342 72,342 72,342 72,342 72,342 R-squared 0.209 0.288 0.291 0.292 0.298 0.303 OLS estimation used. Huber-White robust standard errors clustered on country-year pair (ES round) in brackets. *** (1%), ** (5%), * (10%). 71 Table 9: Split sample results for the interaction term between maternity leave and wage rate Low skills obstacle: Inadequately High skills obstacle: Inadequately educated workers is no obstacle, educated workers is a major or very minor, or moderate obstacle severe obstacle Dependent variable: (1) (2) (3) (4) (5) (6) Female workers Maternity leave -1.874** -1.649** -1.227* -2.699*** -2.174** -1.656* (logs)*Wage rate (0.803) (0.676) (0.642) (1.012) (0.933) (0.861) (real, logs) Maternity leave (logs) 19.119** 16.478*** 64.567 25.050*** 20.111** 60.415 (7.815) (6.305) (44.662) (9.214) (8.714) (60.519) Wage rate (real, logs) 9.718*** 7.092** 5.377* 13.005*** 9.557** 7.995** (3.710) (3.114) (2.833) (4.674) (4.300) (3.949) GDP per capita (logs) 3.878*** 31.511** 1.389 43.907** (1.440) (12.299) (1.730) (17.151) Growth rate of GDP 0.594*** 0.555*** 0.153 0.095 per capita (%, annual) (0.164) (0.174) (0.260) (0.268) Females in adult 1.683*** 1.837*** 1.171** 1.184** population (%) (0.469) (0.481) (0.526) (0.541) Number of workers -0.065 -3.014** -0.647 -7.398** (logs) (0.277) (1.366) (0.439) (3.241) Female owner present 4.400*** 5.992** 3.795*** -14.839** Y:1 N:0 (0.570) (3.037) (0.989) (6.540) Top manager is 13.665*** 13.831*** 12.592*** 12.710*** female Y:1 N:0 (0.846) (0.829) (1.433) (1.416) Age of firm (logs) -0.283 -0.215 0.371 0.372 (0.432) (0.430) (0.609) (0.613) Exports (% of sales) 0.033** 0.030** 0.019 0.014 (0.015) (0.015) (0.022) (0.022) Foreign ownership 0.029*** 0.023** 0.021 0.017 (%) (0.011) (0.010) (0.015) (0.016) Firm offers formal 2.337*** 2.256*** 1.581 1.565 training to workers Y:1 N:0 (0.618) (0.614) (1.061) (1.061) Inadequately educated -0.744** -0.675* 1.215 1.116 workers: Obstacle severity (0-4 scale) (0.341) (0.344) (0.985) (0.981) Firm registered when 0.659 0.577 2.474 2.474* started operations Y:1 N:0 72 (0.821) (0.809) (1.523) (1.477) Firm experienced 1.131 0.872 -0.148 -0.094 losses due to crime Y:1 N:0 (0.765) (0.738) (1.040) (1.054) Access to finance: -0.371 -0.360 -0.245 -0.303 Obstacle severity (0-4 scale) (0.235) (0.227) (0.317) (0.313) Tax rates: Obstacle 0.272 0.248 -0.177 -0.141 severity (0-4 scale) (0.246) (0.245) (0.362) (0.358) Firm has own website 0.732 0.559 -1.721* -1.821* Y:1 N:0 (0.571) (0.563) (1.010) (0.992) Fertility rate 0.129 31.492** -0.836 34.653** (1.016) (12.415) (1.155) (16.622) Gender parity in 0.079*** 0.191 0.014 0.033 workplace laws (0.024) (0.194) (0.026) (0.223) Gender parity in 2.625 -87.208 34.682* -239.445 primary education enrollment (16.712) (196.672) (19.460) (211.525) Freedom from 0.208 0.296 1.369* 1.392* business regulations (0.572) (0.572) (0.732) (0.762) Urbanization rate (%) 0.016 0.011 -0.050 -0.059 (0.043) (0.042) (0.054) (0.056) Women in Parliament 0.091 0.080 0.177** 0.163* (%) (0.062) (0.062) (0.087) (0.090) Paternity leave (logs) -0.223 0.150 0.563 0.782 (0.475) (0.475) (0.636) (0.658) Square of Maternity 0.096 -0.742 leave (logs) (0.604) (0.690) Maternity leave -6.036** -9.172** (logs)*GDP per capita (logs) (2.640) (3.610) Maternity leave 0.627** 1.428** (logs)*Number of workers (logs) (0.290) (0.688) 73 Maternity leave -0.359 4.008*** (logs)*Female owner present Y:1 N:0 (0.641) (1.369) Maternity leave -6.803** -7.677** (logs)*Fertility rate (2.670) (3.538) Maternity leave -0.022 0.015 (logs)*Gender parity in workplace laws (0.044) (0.053) Maternity leave 21.310 60.408 (logs)*Gender parity in primary education enrollment (42.210) (45.956) Wage rate (real, -0.003 -0.011 logs)*Gender parity in workplace laws (0.009) (0.012) Industry fixed effects Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Constant -50.668 -179.257*** -418.128* -68.407 -187.211*** -370.531 (35.469) (45.362) (227.814) (42.387) (59.374) (299.058) Number of 57,410 57,410 57,410 14,932 14,932 14,932 observations R-squared 0.200 0.295 0.299 0.255 0.333 0.338 OLS estimation method. Huber-White robust standard errors clustered on country-year pair (ES round) in brackets. *** (1%), ** (5%), * (10%). 74 Table 10: All interaction terms included simultaneously Dependent variable: (1) (2) (3) (4) (5) (6) Female workers Maternity leave 9.693*** 9.707*** 9.189*** 9.137*** 10.512*** 17.079*** (logs)*Govt. Funded Y:1 N:0 (2.882) (2.159) (2.168) (2.150) (2.240) (4.792) Maternity leave 2.065* 2.065*** 2.008*** 2.042*** 2.040*** 3.584*** (logs)*Females: 20- 29 yrs. old (%) (1.067) (0.746) (0.749) (0.749) (0.765) (1.183) Maternity leave -1.337* -1.396** -1.400** -1.377** -1.380** -0.994* (logs)*Wage rate (real, logs) (0.725) (0.635) (0.643) (0.634) (0.608) (0.543) Maternity leave -7.334 -6.799 -5.971 -6.333 -7.601 21.829 (logs) (12.216) (9.625) (9.674) (9.617) (9.221) (52.878) Govt. Funded Y:1 -39.031*** -42.504*** -40.150*** -39.980*** -45.085*** -110.227*** N:0 (13.097) (9.796) (9.846) (9.760) (10.108) (30.309) Females: 20-29 yrs. -9.637* -9.123** -8.859** -8.972** -8.497** 4.200 old (%) (5.070) (3.525) (3.536) (3.542) (3.552) (23.910) Wage rate (real, 6.676** 6.221** 6.084** 5.923** 5.747** 3.836 logs) (3.379) (2.938) (2.973) (2.935) (2.806) (2.399) GDP per capita 4.110*** 4.200*** 4.072*** 3.411** 21.000 (logs) (0.655) (0.650) (0.639) (1.386) (18.214) Growth rate of GDP 0.410** 0.417** 0.419** 0.295* 0.235 per capita (%, annual) (0.185) (0.179) (0.178) (0.166) (0.169) Females in adult 1.659*** 1.707*** 1.696*** 1.532*** -3.092 population (%) (0.466) (0.465) (0.462) (0.427) (2.491) Number of workers 0.269 -0.198 -0.298 -0.237 -5.648** (logs) (0.260) (0.256) (0.263) (0.254) (2.846) Female owner 5.034*** 4.948*** 4.911*** 4.464*** 6.028 present Y:1 N:0 (0.524) (0.518) (0.514) (0.495) (4.782) 75 Top manager is 13.683*** 13.582*** 13.572*** 13.473*** 13.371*** female Y:1 N:0 (0.789) (0.783) (0.784) (0.756) (0.750) Age of firm (logs) -0.062 -0.032 -0.105 0.042 (0.378) (0.385) (0.380) (0.385) Exports (% of sales) 0.040*** 0.039*** 0.034** 0.030** (0.014) (0.014) (0.014) (0.014) Foreign ownership 0.025*** 0.024** 0.021** 0.024*** (%) (0.010) (0.010) (0.009) (0.009) Firm offers formal 2.284*** 2.199*** 1.828*** 1.992*** training to workers Y:1 N:0 (0.597) (0.600) (0.561) (0.549) Inadequately -0.176 -0.109 -0.073 -0.001 educated workers: Obstacle severity (0- 4 scale) (0.209) (0.224) (0.214) (0.209) Firm registered 1.106 0.853 0.897 when started operations Y:1 N:0 (0.811) (0.749) (0.737) Firm experienced 0.503 0.381 0.462 losses due to crime Y:1 N:0 (0.624) (0.624) (0.602) Access to finance: -0.472** -0.418** -0.400** Obstacle severity (0- 4 scale) (0.224) (0.210) (0.201) Tax rates: Obstacle 0.177 0.162 0.086 severity (0-4 scale) (0.239) (0.223) (0.222) Firm has own 0.244 0.069 -0.063 website Y:1 N:0 (0.562) (0.551) (0.536) Fertility rate 0.499 21.890* (0.954) (11.359) Gender parity in 0.075*** -0.137 workplace laws (0.022) (0.382) Gender parity in 20.147 611.998** primary education enrollment (15.616) (298.909) 76 Freedom from 0.432 0.489 business regulations (0.527) (0.562) Urbanization rate 0.003 0.022 (%) (0.034) (0.032) Women in 0.098 0.090 Parliament (%) (0.064) (0.063) Paternity leave 0.391 0.796 (logs) (0.475) (0.485) Square of Maternity -0.462 leave (logs) (0.711) Maternity leave -6.117** (logs)*GDP per capita (logs) (2.680) Maternity leave 0.783** (logs)*Number of workers (logs) (0.365) Maternity leave 0.179 (logs)*Female owner present Y:1 N:0 (0.578) Maternity leave -4.497* (logs)*Fertility rate (2.440) Maternity leave 0.022 (logs)*Gender parity in workplace laws (0.050) Maternity leave -32.639 (logs)*Gender parity in primary education enrollment (36.518) Govt. Funded Y:1 1.200 N:0*GDP per capita (logs) (1.554) 77 Govt. Funded Y:1 0.490 N:0*Number of workers (logs) (0.586) Govt. Funded Y:1 -0.052 N:0*Female owner present Y:1 N:0 (1.247) Govt. Funded Y:1 29.313 N:0*Gender parity in primary education enrollment (26.603) Govt. Funded Y:1 -0.078** N:0*Gender parity in workplace laws (0.038) Wage rate (real, 0.002 logs)*Gender parity in workplace laws (0.008) Square of Females: 1.217** 20-29 yrs. old (%) (0.554) Females: 20-29 yrs. 1.103 old (%)*GDP per capita (logs) (1.073) Females: 20-29 yrs. 0.173 old (%)*Number of workers (logs) (0.229) Females: 20-29 yrs. -0.303 old (%)*Female owner present Y:1 N:0 (0.474) Females: 20-29 yrs. -52.807** old (%)*Gender parity in primary education enrollment (24.971) 78 Females: 20-29 yrs. 0.017 old (%)*Gender parity in workplace laws (0.026) Maternity leave 0.981* (logs)*% of females in total adult population (5 yr. avg.) (0.558) Industry fixed Yes Yes Yes Yes Yes Yes effects Year fixed effects Yes Yes Yes Yes Yes Yes Constant 69.919 -66.084 -72.896 -70.503 -84.873* -447.567 (58.008) (46.169) (46.768) (46.687) (51.017) (319.343) Number of 72,342 72,342 72,342 72,342 72,342 72,342 observations R-squared 0.219 0.292 0.295 0.295 0.304 0.312 OLS estimation method. Huber-White robust standard errors clustered on country-year pair (ES round) in brackets. *** (1%), ** (5%), * (10%). 79 Appendix Table A1: Sample description Country Survey year Country Survey year Country Survey year Albania 2013 2019 Ethiopia 2011 2015 Myanmar 2014 2016 Angola 2010 Gambia, The 2018 Namibia 2014 Argentina 2010 2017 Georgia 2008 2013 2019 Nepal 2009 2013 Armenia 2009 2013 Ghana 2013 Nicaragua 2010 2016 Azerbaijan 2009 2013 2019 Greece 2018 Niger 2009 2017 Bahamas, 2010 Guatemala 2010 2017 Nigeria 2014 The Bangladesh 2013 Guinea 2016 North Macedonia 2009 2013 2019 Barbados 2010 Guyana 2010 Pakistan 2013 Belarus 2018 Honduras 2010 2016 Panama 2010 Belize 2010 Hungary 2009 2013 2019 Papua New Guinea 2015 Benin 2009 2016 India 2014 Paraguay 2010 2017 Bhutan 2015 Indonesia 2009 2015 Peru 2010 2017 Bolivia 2010 2017 Israel 2013 Philippines 2009 2015 Botswana 2010 Italy 2019 Poland 2009 2013 2019 Brazil 2009 Jamaica 2010 Portugal 2019 Bulgaria 2009 2013 2019 Jordan 2013 2019 Romania 2009 2013 2019 Burkina Faso 2009 Kazakhstan 2009 2013 2019 Russian Federation 2009 2012 2019 Burundi 2014 Kenya 2013 2018 Rwanda 2011 2019 Cambodia 2016 Kyrgyz Republic 2009 2013 2019 Senegal 2014 Cameroon 2009 2016 Lao PDR 2016 2018 Serbia 2009 2013 2019 Central 2011 Latvia 2009 2013 2019 Sierra Leone 2017 African Republic Chad 2009 2018 Lesotho 2016 Slovak Republic 2009 2013 2019 Chile 2010 Liberia 2017 Slovenia 2009 2013 2019 China 2012 Lithuania 2009 2013 2019 Sri Lanka 2011 Colombia 2010 2017 Madagascar 2009 Tajikistan 2013 2019 Congo, Dem. 2010 2013 Malawi 2009 2014 Tanzania 2013 Rep. Costa Rica 2010 Malaysia 2015 Thailand 2016 Côte d'Ivoire 2009 2016 Mali 2010 2016 Timor-Leste 2015 Croatia 2013 2019 Malta 2019 Togo 2009 2016 Cyprus 2019 Mauritania 2014 Trinidad and Tobago 2010 Czech 2009 2013 2019 Mauritius 2009 Turkey 2008 2013 2019 Republic 80 Dominican 2010 2016 Mexico 2010 Uganda 2013 Republic Ecuador 2010 2017 Moldova 2019 Ukraine 2008 2013 2019 Egypt, Arab 2013 2016 Mongolia 2009 2013 2019 Uruguay 2010 2017 Rep. El Salvador 2010 2016 Montenegro 2009 2013 2019 Vietnam 2009 2015 Estonia 2009 2013 2019 Morocco 2013 2019 Zambia 2013 2019 Eswatini 2016 Mozambique 2018 Zimbabwe 2016 Source: Author’s own calculations based on Enterprise Surveys. 81 Table A2: Description of variables Variable Description of variable Female workers Percentage of all full-time workers employed at the firm at the end of the last fiscal year that are females. Source: Enterprise Surveys. www.enterprisesurveys.org Maternity leave (logs) Log of 1 plus the number of days of paid maternity leave mandated by law and averaged over the last 5 years prior to the year the ES was conducted. Source: Women, Business and Law, World Bank. https://wbl.worldbank.org/en/wbl Number of days of paid maternity leave Number of days of paid maternity leave mandated by law and averaged over the last 5 years prior to the year the ES was conducted. Source: Women, Business and Law, World Bank. https://wbl.worldbank.org/en/wbl GDP per capita (logs) Log of GDP per capita (PPP adjusted and at constant 2017 Int’l. Dollars) in the year prior to the year the ES was conducted. Source: World Development Indicators (WDI), World Bank. Growth rate of GDP per capita (%, annual) Annual percentage growth rate of GDP per capita during the year the Enterprise Survey was conducted in the country. The growth rate is computed from GDP per capita in constant local currency units. Source: World Development Indicators (WDI), World Bank. Females in adult population (%) Percentage of females in the total adult population in the country in the year prior to the year Enterprise Surveys was conducted. Source: World Development Indicators, World Bank. Number of workers (logs) Log of total number of workers employed at the firm at the end of the last fiscal year. Total workers include permanent and temporary with the latter adjusted for the number of months employed on average in a firm. Source: Enterprise Surveys. www.enterprisesurveys.org Female owner present Y:1 N:0 Dummy variable equal to 1 if the firm has at least one female owners and 0 otherwise. Source: Enterprise Surveys. www.enterprisesurveys.org Top manager is female Y:1 N:0 Dummy variable equal to 1 if the top manager of the firm is female and 0 otherwise. Source: Enterprise Surveys. www.enterprisesurveys.org 82 Age of firm (logs) Log of age of the firm. Source: Enterprise Surveys. www.enterprisesurveys.org Labor productivity (logs) Log of real annual sales of the firm (in 2009 USD) uring the last fiscaly year divided by the total number of workers at the firm at the end of the last fiscal year. Total workers include permanent and temporary with the latter adjusted for the number of months employed on average in a firm. Data on sales and number of workers are from Enterprise Surveys. Exchange rate used to convert sales from local currencies to USD and GDP deflator used to convert them to 2009 prices is taken from World Development Indicators, World Bank. Source: Enteprrise Surveys and World Development Indictors, World Bank. www.enterprisesurveys.org Exports (% of sales) Percentage of firm’s sales made abroad directly by the firm in the last fiscal year. Source: Enterprise Surveys. www.enterprisesurveys.org Foreign ownership (%) Percentage of the firm that is owned by private foreign individuals, companies or organizations. Source: Enterprise Surveys. www.enterprisesurveys.org Firm offers formal training to workers Y:1 N:0 Dummy variable equal to 1 if the firm has formal training programs for its permanent, full-time employees Source: Enterprise Surveys. www.enterprisesurveys.org Inadequately educated workers: Obstacle severity Severity level of inadequately educated workers as (0-4 scale) an obstacle to firm’s current operations as reported by the firm. The obstacle severity level is one of the following: no obstacle (0), minor obstacle (1), moderate obstacle (2), major obstacle (3), and a very severe obstacle (4). Source: Enterprise Surveys. www.enterprisesurveys.org Firm registered when started operations Y:1 N:0 Dummy variable equal 1 if the firm was registered when it started operations and 0 otherwise. Source: Enterprise Surveys. www.enterprisesurveys.org Firm experienced losses due to crime Y:1 N:0 Dummy variable equal to 1 if the firm reported losses from cirme, theft, and vandalism during the last year and 0 otherwise. Source: Enterprise Surveys. www.enterprisesurveys.org Access to finance: Obstacle severity (0-4 scale) Severity level of (lack of) access to finance as an obstacle to firm’s current operations as reported by the firm. The obstacle severity level is one of the 83 following: no obstacle (0), minor obstacle (1), moderate obstacle (2), major obstacle (3), and a very severe obstacle (4). Source: Enterprise Surveys. www.enterprisesurveys.org Tax rates: Obstacle severity (0-4 scale) Severity level of tax rates as an obstacle to firm’s current operations as reported by the firm. The obstacle severity level is one of the following: no obstacle (0), minor obstacle (1), moderate obstacle (2), major obstacle (3), and a very severe obstacle (4). Source: Enterprise Surveys. www.enterprisesurveys.org Firm has own website Y:1 N:0 Dummy variable equal to 1 if the firm has own website for business and 0 otherwise. Source: Enterprise Surveys. www.enterprisesurveys.org Fertility rate Total number of birth per woman in the last year prior to the year the Enterprise Survey was conducted in the country. Source: World Development Indicators (WDI), World Bank. Gender parity in workplace laws Sub-index of gender inequality in laws related to workplace. The sub-index is compiled by the Women, Business and Law, World Bank team using answers to the following questions: Can a woman get a job in the same way as a man? Does the law prohibit discrimination in employment based on gender? Is there legislation on sexual harassment in employment? Are there criminal penalties or civil remedies for sexual harassment in employment? Higher values of the variable imply greater gender parity or more favorable laws for women. Average value of the sub-index over the last 5 years prior to the year the Enterprise Survey was conducted in the country are used. Source: Women, Business and Law, World Bank. https://wbl.worldbank.org/en/wbl Gender parity in primary education enrollment Ratio of gross primary enrollment rate of females to males. To minimize sample loss due to missing data, average value of the variable over the last 5 years (for which data are available) prior to the year the ES was conducted are used. Source: World Development Indicators (WDI), World Bank. Freedom from business regulations A sub-index of Economic Freedom of the World. It is summary measure over various regulatory 84 aspects that restrict entry into markets and interfere with the freedom to engage in voluntary exchange in product markets. The regulations and regulatory aspects covered by the sub-index include: Administrative requirements, bureaucracy costs, starting a business, impartial public administration, licensing restrictions, and cost of tax compliance. Higher values of the variable imply less restrictive regulations for the private firms. Values of the variable for the year prior to the year Enterprise Surveys was conducted in the country are used. Source: Economic Freedom of the World, Fraser Institute. https://www.fraserinstitute.org/studies/economic- freedom Urbanization rate (%) Percentage of country’s total population that is living in urban areas in the year prior to the year the ES was conducted in the country. Source: World Development Indicators (WDI), World Bank. Women in Parliament (%) Percentage of parliamentary seats in a single or lower chamber held by women. To minimize sample loss due to missing data, average value of the variable over the last 5 years (for which data are available) prior to the year the ES was conducted are used. Source: Inter-Parliamentary Union (IPU). Data retrieved from World Development Indicators (WDI), World Bank. Paternity leave (logs) Log of 1 plus the number of days of paid paternity leave mandated by law and averaged over the last 5 years prior to the year the ES was conducted. Source: Women, Business and Law, World Bank. https://wbl.worldbank.org/en/wbl Year fixed effects A set of dummy variables, one for each year, indicating the year the Enterprise Survey was conducted in the country. Source: Enterprise Surveys. www.enterprisesurveys.org Industry fixed effects A set of dummy variables, one for each industry, indicating the industry to which the firm belongs. Industry if defined at ISIC Rev. 3.1 at 2 digit level. In the baseline sample, there are 46 industries. Source: Enterprise Surveys. www.enterprisesurveys.org Govt. Funded Y:1 N:0 Average value over the last five years prior to the year the Enterprise Survey was conducted in the country of the dummy variable equal to 1 if maternity leave is administered and fully paid by the government and 0 otherwise. 85 Source: Women, Business and Law, World Bank. https://wbl.worldbank.org/en/wbl Wage rate (real, logs) Log of real total labor cost of the firm (in 2009 USD) during the last fiscaly year divided by the total number of workers (permanent plus temporary with temporary adjusted for the number of months worked by the average worker in the firm). Data on labor cost and number of workers are from Enterprise Surveys. Exchange rate used to convert labor cost from local currencies to USD and GDP deflator used to convert it to 2009 prices is taken from World Development Indicators, World Bank. Source: Enteprrise Surveys and World Development Indictors (WDI), World Bank. www.enterprisesurveys.org Females: 20-29 yrs. old (%) Females in the age group of 20-29 years age group as a percentage of the country’s total population. Average values of the variable over the last 5 years prior to the year the Enterprise Survey was conducted in the country are used. Source: World Development Indicators (WDI), World Bank. % of females in total adult population (5 yr. avg.) Percentage of females in the total adult population in the country averaged over the last five years prior the the year Enterprise Surveys was conducted in the country. Source: World Development Indicators, World Bank. % of population that is Catholic Percentage of country’s population that folows Catholicism. Values are available for select years. We use values for the year 2010 which is the latest available. Source: Maoz and Henderson (2013) % of population that is Protestant Percentage of country’s population that folows Protestantism. Values are available for select years. We use values for the year 2010 which is the latest available. Source: Maoz and Henderson (2013) % of population that is Muslim Percentage of country’s population that folows Islam. Values are available for select years. We use values for the year 2010 which is the latest available. Source: Maoz and Henderson (2013) Annual employment growth (%) Total number of permanent full-time workers in the final year (last fiscal year) minus the same in the initial year (beginning of the growth period which is 2 fiscal years ago for most countries and 3 fiscal years ago for other countries) divided by the average (mean value) 86 number of permanent workers in the final and initial year. The ratio is divided by the number of years between the final and initial year and multiplied by 100 to convert to percentage terms. Source: Enterprise Surveys. www.enterprisesurveys.org Informal competition Y:1 N:0 Dummy variable equal to 1 if the firm reports competing against informal or unregistered firms and 0 otherwise. Source: Enterprise Surveys. www.enterprisesurveys.org Firm spent on R&D Y:1 N:0 Dummy variable equal to 1 if the firm spent on R&D acvtivity during the last year and 0 otherwise. Source: Enterprise Surveys. www.enterprisesurveys.org Introduced new product Y:1 N:0 Dummy variable equal to 1 if the firm introduced a new product in the last three years and 0 otherwise. Source: Enterprise Surveys. www.enterprisesurveys.org How Much Of An Obstacle: Labor Regulations? Severity level of labor regulations as an obstacle to firm’s current operations as reported by the firm. The obstacle severity level is one of the following: no obstacle (0), minor obstacle (1), moderate obstacle (2), major obstacle (3), and a very severe obstacle (4). Source: Enterprise Surveys. www.enterprisesurveys.org Ease of Starting a Business (EFW, Fraser A sub-component of the Economic Freedom of the Institute) World index. It is based on data collected by the World Bank’s Doing Business project on three aspects related to starting a business: time (days) necessary to comply with regulations when starting a limited liability company; (2) money costs of the fees paid to regulatory authorities (measured as a share of per-capita income); and (3) minimum capital requirements, that is, funds that must be deposited into a company bank account (measured as a share of per-capita income). These three ratings were then averaged to arrive at the final rating for this sub-component. Higher values of the variable imply less time or cost of starting a business. Values of the variable for the year prior to the year Enterprise Surveys was conducted in the country are used. Source: Economic Freedom of the World, Fraser Institute. https://www.fraserinstitute.org/studies/economic- freedom 87 Table A3: Summary Statistics Variable Mean Std. Minimum Maximum Observations deviation Female workers 33.67 27.77 0 100 80,272 Maternity leave (logs) 4.64 0.62 0 6.02 80,272 Number of days of paid maternity leave 116.65 61.66 0 410 80,272 GDP per capita (logs) 9.06 1.02 6.73 10.66 80,272 Growth rate of GDP per capita (%, annual) 2.43 3.58 -9.44 13.17 80,272 Females in adult population (%) 50.57 1.14 46.70 54.27 80,272 Number of workers (logs) 2.85 1.11 0 8.52 80,272 Female owner present Y:1 N:0 0.36 0.48 0 1 80,272 Top manager is female Y:1 N:0 0.19 0.39 0 1 80,272 Age of firm (logs) 2.66 0.68 0 5.28 80,272 Labor productivity (logs) 10.15 1.79 -4.32 24.76 80,272 Exports (% of sales) 6.72 20.65 0 100 80,272 Foreign ownership (%) 7.69 24.72 0 100 80,272 Firm offers formal training to workers Y:1 N:0 0.35 0.48 0 1 80,272 Inadequately educated workers: Obstacle severity 1.38 1.31 0 4 80,272 (0-4 scale) Firm registered when started operations Y:1 N:0 0.90 0.30 0 1 80,272 Firm experienced losses due to crime Y:1 N:0 0.18 0.39 0 1 80,272 Access to finance: Obstacle severity (0-4 scale) 1.51 1.34 0 4 80,272 Tax rates: Obstacle severity (0-4 scale) 1.81 1.33 0 4 80,272 Firm has own website Y:1 N:0 0.47 0.50 0 1 80,272 Fertility rate 2.86 1.52 1.23 7.57 80,272 Gender parity in workplace laws 72.02 29.50 0 100 80,272 Gender parity in primary education enrollment 0.97 0.06 0.68 1.07 80,272 Freedom from business regulations 6.13 1.06 3.34 8.97 80,272 Urbanization rate (%) 52.84 19.51 11.19 95.05 80,272 Women in Parliament (%) 18.20 9.41 1.97 62.75 80,272 Paternity leave (logs) 0.84 1.02 0 3.46 80,272 Govt. Funded Y:1 N:0 0.65 0.48 0 1 80,272 Wage rate (real, logs) 7.97 1.50 -9.10 14.13 72,342 Females: 20-29 yrs. old (%) 8.21 1.06 5.07 10.38 80,272 % of females in total adult population (5 yr. avg.) 50.61 1.14 46.73 54.25 80,272 % of population that is Catholic 0.29 0.32 0 0.96 69,136 % of population that is Protestant 0.10 0.13 0 0.61 69,136 % of population that is Muslim 0.21 0.33 0 0.99 69,136 Annual employment growth (%) 4.58 16.31 -97.04 100 75,578 Informal competition Y:1 N:0 0.53 0.50 0 1 77,163 Firm spent on R&D Y:1 N:0 13.22 33.87 0 100 68,710 Introduced new product Y:1 N:0 33.99 47.37 0 100 69,708 88 How Much Of An Obstacle: Labor Regulations? 1.01 1.13 0 4 80,024 Ease of Starting a Business (EFW, Fraser 8.76 1.24 2.51 9.92 80,170 Institute)) Source: Author's own calculations based on Enterprise Surveys data. Sample size varies due to missing data. 89 Table A4: Lagged values of maternity lave Dependent (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) variable: Female workers Maternity 4.075** 2.612** 2.722** 2.798** 2.726** leave (10-yr. avg., logs) (1.610) (1.313) (1.230) (1.208) (1.309) Maternity 2.811** 2.080** 2.145** 2.210** 2.049** leave (1 yr. lagged, logs) (1.290) (0.916) (0.872) (0.864) (0.862) GDP per 4.490*** 4.649*** 4.436*** 3.310** 4.608*** 4.767*** 4.557*** 3.398** capita (logs) (0.561) (0.554) (0.550) (1.376) (0.553) (0.550) (0.546) (1.375) Growth rate 0.520*** 0.523*** 0.527*** 0.451*** 0.509*** 0.512*** 0.515*** 0.442*** of GDP per capita (%, annual) (0.183) (0.177) (0.177) (0.165) (0.180) (0.174) (0.174) (0.164) Females in 1.651*** 1.699*** 1.695*** 1.543*** 1.681*** 1.730*** 1.726*** 1.575*** adult population (%) (0.475) (0.472) (0.469) (0.458) (0.476) (0.472) (0.469) (0.459) Number of -0.012 -0.378 -0.500* -0.467* -0.022 -0.385 -0.502* -0.465* workers (logs) (0.299) (0.282) (0.290) (0.280) (0.299) (0.282) (0.290) (0.279) Female 4.990*** 4.956*** 4.905*** 4.494*** 4.990*** 4.954*** 4.904*** 4.488*** owner present Y:1 N:0 (0.498) (0.491) (0.487) (0.489) (0.500) (0.493) (0.489) (0.491) 90 Top 13.298*** 13.166*** 13.160*** 13.036*** 13.299*** 13.171*** 13.165*** 13.048*** manager is female Y:1 N:0 (0.801) (0.795) (0.792) (0.774) (0.802) (0.796) (0.793) (0.774) Age of firm -0.240 -0.298 -0.383 -0.230 -0.286 -0.375 (logs) (0.379) (0.389) (0.386) (0.382) (0.391) (0.387) Labor -0.228 -0.275 -0.244 -0.221 -0.268 -0.237 productivity (logs) (0.195) (0.200) (0.187) (0.194) (0.200) (0.187) Exports (% 0.029** 0.028** 0.022* 0.029** 0.029** 0.023* of sales) (0.014) (0.014) (0.013) (0.014) (0.014) (0.013) Foreign 0.024*** 0.023** 0.023*** 0.024** 0.022** 0.023** ownership (%) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) Firm offers 2.317*** 2.199*** 1.978*** 2.287*** 2.174*** 1.944*** formal training to workers Y:1 N:0 (0.573) (0.568) (0.560) (0.572) (0.567) (0.559) Inadequately -0.212 -0.154 -0.096 -0.230 -0.179 -0.125 educated workers: Obstacle severity (0-4 scale) (0.189) (0.204) (0.206) (0.188) (0.203) (0.205) Firm -0.182 -0.346 -0.189 -0.349 registered when started 91 operations Y:1 N:0 (1.030) (1.000) (1.024) (0.999) Firm 0.749 0.780 0.674 0.699 experienced losses due to crime Y:1 N:0 (0.612) (0.613) (0.611) (0.614) Access to -0.590*** -0.510** -0.589*** -0.506** finance: Obstacle severity (0-4 scale) (0.224) (0.216) (0.225) (0.217) Tax rates: 0.267 0.282 0.302 0.308 Obstacle severity (0-4 scale) (0.230) (0.221) (0.233) (0.223) Firm has 0.600 0.492 0.581 0.466 own website Y:1 N:0 (0.548) (0.536) (0.550) (0.538) Fertility rate -0.112 -0.170 (0.984) (0.965) Gender 0.064*** 0.064*** parity in workplace laws (0.022) (0.022) Gender 15.961 13.887 parity in primary 92 education enrollment (12.976) (12.717) Freedom 0.334 0.267 from business regulations (0.530) (0.528) Urbanization -0.007 -0.003 rate (%) (0.040) (0.040) Women in 0.080 0.088 Parliament (%) (0.062) (0.063) Paternity -0.052 -0.056 leave (logs) (0.468) (0.468) Industry Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes fixed effects Year fixed Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes effects Constant 21.676** -109.688*** -112.853*** -110.112*** -114.839*** 27.536*** -109.954*** -112.968*** -110.259*** -111.983*** (9.033) (27.771) (27.414) (27.320) (34.525) (8.004) (27.550) (27.249) (27.129) (33.668) Number of 80,272 80,272 80,272 80,272 80,272 80,272 80,272 80,272 80,272 80,272 observations R-squared 0.199 0.280 0.282 0.283 0.290 0.197 0.279 0.282 0.283 0.290 OLS estimation method used. All coefficient values are log odds ratios obtained from logit estimation. Huber-White robust standard errors clustered on country in brackets. *** (1%), ** (5%), * (10%). 93 Table A5: Excluding high-income countries Dependent variable: Female (1) (2) (3) (4) (5) workers Maternity leave (logs) 4.328** 2.838* 2.986** 3.080** 2.742* (1.965) (1.537) (1.427) (1.409) (1.423) GDP per capita (logs) 4.739*** 4.884*** 4.630*** 3.667** (0.760) (0.734) (0.726) (1.690) Growth rate of GDP per capita 0.742*** 0.737*** 0.729*** 0.620*** (%, annual) (0.215) (0.209) (0.208) (0.193) Females in adult population (%) 1.507*** 1.570*** 1.571*** 1.405*** (0.514) (0.507) (0.503) (0.496) Number of workers (logs) -0.125 -0.605* -0.826** -0.846** (0.360) (0.342) (0.350) (0.335) Female owner present Y:1 N:0 5.300*** 5.182*** 5.085*** 4.702*** (0.589) (0.571) (0.563) (0.558) Top manager is female Y:1 N:0 12.540*** 12.440*** 12.493*** 12.317*** (0.913) (0.901) (0.897) (0.875) Age of firm (logs) -0.249 -0.318 -0.441 (0.425) (0.437) (0.426) Labor productivity (logs) -0.180 -0.241 -0.196 (0.211) (0.216) (0.200) Exports (% of sales) 0.038** 0.036** 0.031* (0.017) (0.017) (0.017) Foreign ownership (%) 0.026** 0.024** 0.023** (0.010) (0.010) (0.010) Firm offers formal training to 2.877*** 2.666*** 2.436*** workers Y:1 N:0 (0.661) (0.652) (0.634) Inadequately educated workers: -0.016 0.113 0.207 Obstacle severity (0-4 scale) (0.220) (0.232) (0.232) Firm registered when started -0.293 -0.472 operations Y:1 N:0 (1.098) (1.060) Firm experienced losses due to 0.812 0.903 crime Y:1 N:0 (0.704) (0.698) Access to finance: Obstacle -0.627** -0.572** severity (0-4 scale) (0.256) (0.247) Tax rates: Obstacle severity (0- 0.005 -0.025 4 scale) (0.261) (0.249) Firm has own website Y:1 N:0 1.385** 1.316** (0.637) (0.624) Fertility rate 0.318 (1.021) Gender parity in workplace 0.078*** laws (0.023) Gender parity in primary 16.384 education enrollment (13.429) Freedom from business 0.428 94 regulations (0.573) Urbanization rate (%) 0.017 (0.050) Women in Parliament (%) 0.053 (0.067) Paternity leave (logs) -0.636 (0.629) Industry fixed effects Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Constant 8.582 -110.602*** -114.467*** -111.079*** -118.225*** (12.527) (29.465) (28.925) (28.852) (35.885) Number of observations 67,645 67,645 67,645 67,645 67,645 R-squared 0.178 0.260 0.264 0.266 0.274 Huber-White robust standard errors clustered on country-year pair (ES round) in brackets. *** (1%), ** (5%), * (10%). The sample excludes countries that are classified as high-income countries at the time of the survey. Income classification follows the World Bank. 95 Table A6: Split sample results for government vs. employer funded maternity leave Partly or fully employer funded maternity leave Fully government funded maternity leave Dependent variable: (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Female workers Maternity leave 0.687 -0.554 -0.158 0.034 -1.296 2.095* 2.294** 2.286** 2.341** 2.081** (logs) (1.434) (1.391) (1.381) (1.396) (1.507) (1.170) (0.994) (0.957) (0.954) (0.863) GDP per capita 2.675** 2.622** 2.146* 2.340 4.085*** 4.553*** 4.483*** 1.624 (logs) (1.073) (1.083) (1.081) (2.599) (0.823) (0.811) (0.799) (1.261) Growth rate of -0.211 -0.199 -0.205 -0.386 0.565*** 0.554*** 0.545*** 0.363* GDP per capita (%, (0.307) (0.301) (0.303) (0.298) (0.216) (0.207) (0.208) (0.205) annual) Females in adult 1.221 1.165 1.147 0.310 1.445*** 1.480*** 1.457*** 1.181** population (%) (0.809) (0.799) (0.787) (0.657) (0.433) (0.438) (0.442) (0.460) Number of workers -0.893 -1.433** -1.803*** -1.817*** 0.493* 0.128 0.186 0.171 (logs) (0.633) (0.603) (0.596) (0.561) (0.272) (0.253) (0.263) (0.253) Female owner 6.115*** 5.897*** 5.749*** 4.912*** 3.815*** 3.695*** 3.700*** 3.583*** present Y:1 N:0 (0.936) (0.914) (0.900) (0.876) (0.548) (0.544) (0.546) (0.555) Top manager is 12.689*** 12.666*** 12.648*** 12.257*** 13.640*** 13.521*** 13.421*** 13.456*** female Y:1 N:0 (1.373) (1.342) (1.323) (1.270) (0.940) (0.931) (0.926) (0.879) Age of firm (logs) -0.813 -0.996 -1.151** 0.680 0.740* 0.818* (0.615) (0.615) (0.528) (0.428) (0.428) (0.417) Labor productivity 0.285 0.215 0.152 -0.785*** -0.770*** -0.629** (logs) (0.230) (0.234) (0.202) (0.268) (0.268) (0.260) Exports (% of 0.050** 0.047* 0.046* 0.025 0.026 0.018 sales) (0.025) (0.025) (0.024) (0.016) (0.016) (0.015) Foreign ownership 0.015 0.014 0.012 0.021* 0.020 0.023** (%) (0.013) (0.013) (0.013) (0.012) (0.012) (0.012) 96 Firm offers formal 3.182*** 2.719*** 2.119** 2.077*** 2.138*** 1.915*** training to workers Y:1 N:0 (0.952) (0.893) (0.836) (0.646) (0.653) (0.649) Inadequately -0.058 0.090 0.127 -0.127 -0.109 -0.035 educated workers: Obstacle severity (0-4 scale) (0.329) (0.358) (0.370) (0.218) (0.229) (0.222) Firm registered -2.271 -2.387* 1.241 1.301 when started operations Y:1 N:0 (1.476) (1.394) (1.127) (1.144) Firm experienced 0.247 0.304 0.887 0.823 losses due to crime Y:1 N:0 (0.956) (1.000) (0.687) (0.678) Access to finance: -0.577* -0.566* -0.477* -0.365 Obstacle severity (0-4 scale) (0.343) (0.330) (0.252) (0.244) Tax rates: Obstacle -0.113 -0.166 0.337 0.427* severity (0-4 scale) (0.396) (0.402) (0.244) (0.222) Firm has own 3.823*** 3.522*** -1.263** -1.370** website Y:1 N:0 (0.903) (0.875) (0.621) (0.592) Fertility rate 1.735 -0.594 (1.593) (0.837) Gender parity in 0.119*** 0.027 workplace laws (0.035) (0.023) Gender parity in 35.264* 56.733*** primary education enrollment 97 (17.684) (15.909) Freedom from -0.267 0.438 business regulations (0.980) (0.691) Urbanization rate -0.018 -0.043 (%) (0.079) (0.044) Women in 0.114 0.104 Parliament (%) (0.078) (0.074) Paternity leave -1.045 0.334 (logs) (0.947) (0.514) Industry fixed Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes effects Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Constant 14.632* -64.678 -62.542 -55.528 -48.915 4.757 -112.289*** -113.074*** -112.948*** -139.284*** (8.691) (44.252) (43.183) (42.976) (41.256) (8.342) (27.393) (27.634) (27.780) (35.483) Number of 31,530 31,530 31,530 31,530 31,530 48,742 48,742 48,742 48,742 48,742 observations R-squared 0.156 0.235 0.241 0.246 0.263 0.255 0.318 0.321 0.322 0.329 OLS estimation method used. All coefficient values are log odds ratios obtained from logit estimation. Huber-White robust standard errors clustered on country in brackets. *** (1%), ** (5%), * (10%). 98 Table A7: Split sample results for share of females in 20-29 age group in total population Below median values of Females: 20-29 yrs. old (%) Above median values of Females: 20-29 yrs. old (%) Dependent variable: (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Female workers Maternity leave 1.694** 0.937 0.955 0.958 1.109 8.360* 6.309* 6.360* 6.592* 6.680* (logs) (0.845) (0.715) (0.711) (0.720) (0.842) (4.982) (3.581) (3.518) (3.500) (3.664) GDP per capita 4.888*** 4.994*** 4.845*** 4.257** 3.055*** 3.150*** 2.999*** 2.798 (logs) (0.557) (0.618) (0.615) (2.097) (1.069) (1.034) (1.033) (1.760) Growth rate of GDP 0.365* 0.382* 0.370* 0.377* 0.701** 0.717** 0.724** 0.558** per capita (%, annual) (0.219) (0.218) (0.215) (0.216) (0.315) (0.294) (0.295) (0.276) Females in adult 1.877** 1.971** 1.942** 1.828** 1.896*** 1.870*** 1.858*** 1.770*** population (%) (0.744) (0.764) (0.775) (0.820) (0.649) (0.642) (0.636) (0.617) Number of workers 0.112 -0.091 -0.102 -0.124 -0.222 -0.778* -0.978** -1.008** (logs) (0.350) (0.321) (0.343) (0.319) (0.483) (0.441) (0.456) (0.432) Female owner 4.600*** 4.580*** 4.581*** 4.447*** 4.941*** 4.880*** 4.776*** 4.182*** present Y:1 N:0 (0.624) (0.630) (0.623) (0.644) (0.712) (0.688) (0.693) (0.679) Top manager is 13.030*** 12.981*** 12.934*** 12.978*** 13.175*** 12.99*** 13.016*** 12.991*** female Y:1 N:0 (1.051) (1.051) (1.045) (1.027) (1.154) (1.114) (1.107) (1.067) Age of firm (logs) 0.250 0.284 0.133 -0.556 -0.662 -0.605 (0.587) (0.609) (0.581) (0.439) (0.451) (0.469) Labor productivity -0.210 -0.221 -0.218 -0.107 -0.174 -0.088 (logs) (0.298) (0.301) (0.281) (0.242) (0.246) (0.233) Exports (% of sales) 0.027 0.026 0.020 0.020 0.020 0.022 (0.017) (0.017) (0.016) (0.022) (0.022) (0.021) 99 Foreign ownership 0.005 0.003 0.011 0.037*** 0.035*** 0.029** (%) (0.012) (0.013) (0.012) (0.013) (0.013) (0.012) Firm offers formal 1.384** 1.395** 1.374** 2.995*** 2.800*** 2.512*** training to workers Y:1 N:0 (0.691) (0.684) (0.689) (0.846) (0.827) (0.757) Inadequately -0.404 -0.290 -0.267 0.309 0.330 0.396 educated workers: Obstacle severity (0- 4 scale) (0.263) (0.290) (0.289) (0.271) (0.296) (0.281) Firm registered 0.854 0.633 -0.576 -0.785 when started operations Y:1 N:0 (1.713) (1.624) (1.100) (1.145) Firm experienced 0.015 0.157 1.162 1.310* losses due to crime Y:1 N:0 (0.853) (0.866) (0.802) (0.783) Access to finance: -0.544* -0.525* -0.503* -0.500* Obstacle severity (0- 4 scale) (0.303) (0.300) (0.294) (0.277) Tax rates: Obstacle 0.133 0.149 0.266 0.200 severity (0-4 scale) (0.315) (0.303) (0.306) (0.302) Firm has own -0.376 -0.421 1.227 1.129 website Y:1 N:0 (0.665) (0.654) (0.845) (0.810) Fertility rate -0.004 0.775 (1.438) (1.075) Gender parity in 0.023 0.083*** workplace laws 100 (0.028) (0.024) Gender parity in 22.531 19.214 primary education enrollment (17.655) (25.140) Freedom from -0.275 1.399* business regulations (0.697) (0.710) Urbanization rate -0.019 -0.032 (%) (0.052) (0.053) Women in 0.159 -0.017 Parliament (%) (0.099) (0.076) Paternity leave 0.058 -0.125 (logs) (0.538) (0.844) Industry fixed Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes effects Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Constant 38.488*** -113.389*** -118.80*** -116.936*** -129.68** -13.357 -136.446*** -134.51*** -131.234*** -162.718*** (6.581) (40.940) (42.047) (42.618) (49.732) (24.518) (35.065) (34.333) (34.084) (42.376) Number of 43,923 43,923 43,923 43,923 43,923 36,349 36,349 36,349 36,349 36,349 observations R-squared 0.243 0.321 0.323 0.323 0.327 0.187 0.261 0.266 0.267 0.278 OLS estimation method used. All coefficient values are log odds ratios obtained from logit estimation. Huber-White robust standard errors clustered on country in brackets. *** (1%), ** (5%), * (10%). 101 Table A8: Split sample results for low vs. high wage rate Below median values of Wage rate (real, logs) Above median values of Wage rate (real, logs) Dependent (1) (2) (3) (4) (5) (6) (7) (8) (19) (10) variable: Female workers Maternity leave 5.975** 4.473** 4.498** 4.539** 3.890** 1.333* 1.317** 1.338** 1.422** 1.323* (logs) (2.507) (1.922) (1.831) (1.816) (1.756) (0.734) (0.658) (0.651) (0.651) (0.673) GDP per capita 5.623*** 5.474*** 5.203*** 5.263*** 2.815*** 3.274*** 3.190*** 1.171 (logs) (0.770) (0.754) (0.742) (1.678) (0.847) (0.843) (0.831) (1.374) Growth rate of 0.473** 0.468** 0.460** 0.458** 0.557** 0.533** 0.526** 0.446** GDP per capita (%, annual) (0.232) (0.223) (0.224) (0.202) (0.232) (0.229) (0.228) (0.190) Females in adult 1.673*** 1.726*** 1.701*** 1.586*** 1.589*** 1.574*** 1.532*** 1.483*** population (%) (0.618) (0.613) (0.603) (0.574) (0.481) (0.487) (0.485) (0.424) Number of 1.014*** 0.321 -0.002 -0.104 -0.259 -0.632** -0.572** -0.326 workers (logs) (0.385) (0.401) (0.403) (0.393) (0.290) (0.283) (0.288) (0.280) Female owner 5.333*** 5.094*** 5.015*** 4.748*** 4.268*** 4.173*** 4.181*** 3.767*** present Y:1 N:0 (0.759) (0.746) (0.744) (0.731) (0.601) (0.600) (0.592) (0.563) Top manager is 14.088*** 13.963*** 14.048*** 13.931*** 13.255*** 13.145*** 13.033*** 12.761*** female Y:1 N:0 (1.087) (1.093) (1.099) (1.076) (0.965) (0.958) (0.956) (0.918) Age of firm (logs) -0.403 -0.355 -0.367 0.373 0.371 0.297 (0.543) (0.548) (0.557) (0.515) (0.508) (0.494) Wage rate (real, 0.423 0.307 0.194 -1.391** -1.353** -1.245*** logs) (0.378) (0.379) (0.345) (0.550) (0.545) (0.472) Exports (% of 0.070*** 0.067*** 0.067*** 0.021 0.022 0.010 sales) (0.018) (0.018) (0.018) (0.016) (0.016) (0.016) Foreign 0.035** 0.033** 0.028** 0.026** 0.025** 0.025** ownership (%) (0.014) (0.015) (0.014) (0.010) (0.010) (0.010) Firm offers 3.233*** 2.993*** 2.635*** 2.093*** 2.197*** 1.752*** formal training to workers Y:1 N:0 102 (0.920) (0.901) (0.840) (0.656) (0.673) (0.641) Inadequately -0.048 0.047 0.090 -0.390* -0.379* -0.258 educated workers: Obstacle severity (0-4 scale) (0.343) (0.353) (0.338) (0.208) (0.229) (0.228) Firm registered 0.957 0.984 0.247 0.430 when started operations Y:1 N:0 (0.961) (0.873) (1.248) (1.271) Firm experienced 0.556 0.619 0.863 0.716 losses due to crime Y:1 N:0 (0.946) (0.928) (0.690) (0.702) Access to -0.434 -0.527* -0.531** -0.417* finance: Obstacle severity (0-4 scale) (0.335) (0.317) (0.243) (0.226) Tax rates: -0.063 -0.138 0.437 0.389 Obstacle severity (0-4 scale) (0.334) (0.313) (0.267) (0.243) Firm has own 1.895** 1.836** -1.190* -1.371** website Y:1 N:0 (0.807) (0.799) (0.655) (0.621) Fertility rate 1.742* -1.402 (1.053) (1.059) Gender parity in 0.073*** 0.045** workplace laws (0.025) (0.021) 103 Gender parity in 17.109 15.401 primary education enrollment (17.833) (16.716) Freedom from 0.397 0.648 business regulations (0.661) (0.597) Urbanization rate 0.081 -0.062* (%) (0.053) (0.036) Women in 0.088 0.168** Parliament (%) (0.072) (0.068) Paternity leave -1.083* 0.830 (logs) (0.594) (0.526) Industry fixed Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes effects Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Constant 14.500 -134.075*** -135.994*** -131.856*** -157.792*** 28.68*** -81.362*** -76.415** -75.428** -73.428** (11.605) (34.196) (33.718) (33.327) (43.187) (4.872) (27.961) (29.434) (29.335) (30.338) Number of 36,131 36,131 36,131 36,131 36,131 36,211 36,211 36,211 36,211 36,211 observations R-squared 0.187 0.281 0.286 0.288 0.295 0.250 0.313 0.316 0.317 0.329 OLS estimation method used. All coefficient values are log odds ratios obtained from logit estimation. Huber-White robust standard errors clustered on country in brackets. *** (1%), ** (5%), * (10%). 104