WPS8715 Policy Research Working Paper 8715 Paid Maternity Leave and Female Employment Evidence Using Firm-Level Survey Data for Developing Countries Mohammad Amin Asif Islam Development Economics Global Indicators Group January 2019 Policy Research Working Paper 8715 Abstract The relationship between the length of paid maternity percentage points increase in the share of workers in a typi- leave and the proportion of female workers in the private cal firm that are female. As expected, the stated relationship sector is explored using firm-level survey data for 66 mostly is much larger when the government pays for maternity developing countries. The paper finds a large, positive, leave versus the employer. The results are robust to sev- and statistically significant relationship between the two. eral controls for firm and country characteristics and other According to the most conservative estimate, an increase possible heterogeneities in the maternity leave and female of one week of paid maternity leave is associated with a 2.6 workers relationship. This paper is a product of the Global Indicators Group, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/research. 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 Paid Maternity Leave and Female Employment: Evidence Using Firm-Level Survey Data for Developing Countries By: Mohammad Amin* and Asif Islam** Enterprise Analysis Unit, DECEA, World Bank JEL Codes: J13, J16, J18, J22 Keywords: Gender, Female employment, Maternity leave, Developing countries, Firms We would like to thank Jorge Luis Rodriguez Meza for helpful comments. All remaining errors are our own. * Senior Economist, Enterprise Analysis Unit, DECEA, World Bank Malaysia Country Office and Washington, DC. Email: mamin@worldbank.org ** Economist, Enterprise Analysis Unit, DECEA, Washington DC. Email: aislam@worldbank.org 1. Introduction Career interruptions due to motherhood are known to adversely impact labor market outcomes for females (Goldin 2006, Bertrand et al. 2010). To mitigate these adverse effects, maternity leave policies have been implemented by most countries in the world. According to the World Bank (2018), except for 7 countries, all other countries in the world have maternity leave policies. There are significant differences across countries in the length of maternity paid leave, who pays for it and the extent of job protection provided. Existing empirical studies, mostly confined to developed countries, and theoretical models (reviewed below) show mixed results for the impact of maternity leave policies on female job market prospects. We contribute to this literature by analyzing the relationship between the length of paid maternity leave (henceforth, maternity leave) and share of workers in a typical private firm that are female (henceforth, female workers). Focusing on 66 mostly developing countries, we find a large, positive and statistically significant relationship between female workers and maternity leave. Our conservative estimate suggests that each additional week of maternity leave is associated with an increase of 2.6 percentage points in female workers. As expected, the positive relationship between maternity leave and female workers is found to be much larger (more positive) when maternity leave is funded by the government vs. the employer and in countries with more females in the total population. Theoretically, the direction of the relationship between maternity leave and female employment is not straightforward. Klerman and Leibowitz (1997, 1999) develop a model that predicts that mothers’ return to work after a birth can happen either earlier or later when a maternity leave scheme is in place compared to when no such scheme is in place. The authors argue that maternity leave can delay return to work for mothers who would have returned to the pre-birth employer with and without a maternity leave scheme. For mothers who would have left the labor market (and their pre-birth employer) after giving birth if they had no access no maternity leave, 2    maternity leave leads to retention with the pre-birth employer and a potentially earlier return to work. While much of the literature is focused on the period immediately following child-birth, the impact of maternity leave may be felt many years before or after the maternity leave period. That is, there may be long-run effects. There are several reasons for this.1 First, according to the human capital model (Mincer and Polachek 1974), a temporary exit from the labor market around the time of child-birth by a mother implies less work-related human capital accumulated by her and even skill depreciation. Second, labor productivity will rise if maternity leave increases firm- specific human capital by allowing women to return to their old jobs. This will shift the demand curve for female workers to the right, further increasing women’s employment and wages (Ruhm 1998). Third, opportunities in the labor market are network dependent (Calvó-Armengol and Jackson 2004, Drange and Rege 2013). If temporary exit leads to loss of network, returning to work may be less rewarding to a mother post-birth. Fourth, mothers’ reentry into the labor force post-birth could discourage post-birth return if staying home with the children increases the accumulation of human capital related to home production (Becker 1991). Fifth, as argued by Ruhm (1998), women who would not otherwise participate in the labor market may choose to be active prior to childbirth to subsequently qualify for maternity leave benefits. Sixth, anticipating maternity leave related costs, employers may be discouraged from hiring female workers. This factor is likely to be more important when the employer pays the salary during maternity leave with no or little contribution from the government (see for example, Ruhm 1998 for a discussion on this point).                                                              1  Drange and Rege (2013) nicely summarize the arguments here.  3    Existing empirical evidence on the effects of maternity leave on female employment in the period immediately after giving birth and in the long-run is mixed. That is, for the period just after childbirth, while some studies show that maternity leave is associated with a faster return or higher probability of return to work post childbirth (Waldfogel et al. 1999, Berger and Waldfogel 2004), others show just the opposite (Lalive and Zweimüller 2009, Hanratty and Trzcinski 2009, Ludsteck and Schonberg 2007), or no effect of maternity leave (Klerman and Leibowitz 1997, 1999, Dahl et al. 2016). Empirical studies on the long-run impact of maternity leave on female employment also show a mixed picture, although such studies are relatively scarce. Using census data for the U.S. for 1980 and 1990, Klerman and Leibowitz (1997) find no significant long-run impact on female employment of maternity leave sanctioned under the federal Family and Leave Act. Similarly, Hanratty and Trzcinski (2009) find that the 2000 reform in maternity leave legislation in Canada that increased maternity leave from 25 to 50 weeks had no impact on return to work by mothers beyond the duration of the maternity leave. Lalive and Zweimuller (2009) also find no significant effects of increased maternity leave in Austria on female labor supply in the long-run (over 10 years after birth). In contrast, Kluev and Schmitz (2018) estimate policy impacts of a generous parental benefit in Germany. Their estimates for the medium or long-run impact (i.e. 3-5 years after childbirth) show an increase in the probability of mothers’ employment by up to 10 percent, higher job continuity with the pre-birth employer, and better quality of jobs. On the other hand, Schonberg and Ludsteck (2014) find that maternity leave expansion led to a fall in women’s employment 6 years after childbirth, but this decline was small. Ondrich et al. (2003) also arrive at a similar conclusion for the case of Germany. 4    The present paper analyzes the relationship between the number of days of paid maternity leave and the proportion of workers in a typical registered private firm that are females. As indicated above, we also explore the idea of lower demand by employers for female workers when the cost of maternity leave falls on the employers as opposed to the government (discussed in detail below). The analysis is based on firm-level survey data for a large cross-section of mostly developing countries. Thus, we capture the full impact (short-run plus long-run) of maternity leave. In this sense, our approach is like that of Fipps et al. (2001), who argue that “ever having” rather than “currently having” a child is generally the more appropriate measure for studying the family gap. We contribute to the existing literature in several ways. First, estimates of the long-run effects of maternity leave are largely restricted to a handful of rich countries. Whether the results apply to other countries is not discussed. We fill this gap in the literature by focusing on a large cross-section of countries. Second, there is substantial variation across countries in who bears the cost of maternity leave. Demand for female workers may contract when such costs fall on the employers as opposed to the government. However, there is no formal evidence on the issue. We attempt to fill this gap in the literature by analyzing how the maternity leave and female workers relationship differs depending on who pays for maternity leave (employer, government, or both). Our results confirm that the maternity leave and female workers relationship is much larger (more positive) when maternity leave is funded by the government vs. the employer. Third, the use of individual-level data that is pervasive in the existing literature suffers from selection bias. Women who are eligible for maternity leave and those who decide to have children are different in several characteristics from the rest. Failure to control for these selection effects can significantly bias the estimates of the impact of maternity leave (see for example, Hashimoto et al. 2004). Our results 5    based on cross-country differences in maternity leave and firm-level survey data are relatively immune from such selection bias. This is especially so since the survey data we use target only registered private firms with 5 or more full-time workers. Thus, the issue of self-selection of firms and workers into the micro (with less than 5 employees) and unregistered sectors where maternity leave laws do not apply or are difficult to enforce than in the rest of the private sector does not arise in our sample. Fourth, there is lingering concern that women are still concentrated in low- paying jobs such as those in the informal sector. Thus, our focus on registered private firms that provide the bulk of high-paying jobs in the country is important. We would like to caution that our results are based on pure cross-section data and so they may suffer from omitted variable bias. Apart from controlling for several firm and country characteristics, we offer two validity checks against the omitted variable bias problem. First, as discussed above, we explore the difference between government funded vs. employer funded maternity leave. The understanding is that the maternity leave and female employment relationship should be stronger (more positive or less negative) for the case of government funded maternity leave. However, there is no reason to expect this to hold if maternity leave is merely proxying for other correlated factors such as the overall level of development, cultural factors, etc. Second, since we consider the relationship between maternity leave and the proportion of females in total (male plus female) workers at the firm, we can expect the relationship be magnified (in absolute terms) when there are more adult females relative to adult males in the country. 2. Data and Main Variables Our main data source is firm-level surveys conducted by the World Bank’s Enterprise Surveys (ES) and the World Bank’s Women, Business and the Law (WBL) project that documents country- level gender-specific disparities in the law. These data sources are complemented with other data 6    sources including the World Development Indicators (WDI, World Bank) and La Porta et al. (1999). The ES consist of a stratified random sample of firms that is representative of the country’s formal (registered) private sector. Sampling weights are provided in the survey to allow for inferences to the targeted universe of firms. The ES do not cover firms in the informal sector, firms with fewer than five employees, the agriculture and mining sectors, as well as some services sectors such as banking and insurance, health care, and education. A common survey methodology and questionnaire is used across surveys. We use the latest round of the ES available in each country. The sample we use consists of over 48,647 private firms located in 66 different countries that are mostly developing countries. 2 The survey year for the ES varies across countries ranging from 2007 to 2016. Summary statistics of the variables used are provided in table 1; correlations between our main explanatory variables are provided in table 2.3 Dependent variable Our dependent variable is the percentage of all permanent full-time workers at the end of the last fiscal year at the firm that are females (Female workers). The variable ranges between 0 and 100, with a mean value of 32 (percent), median value of 26 and the standard deviation equals 27.3. The percentage of female workers in a typical firm is highest in Belarus (51%) and lowest in Pakistan (5%). Note that our dependent variable is women’s employment relative to total employment at the firm. Thus, factors such as overall economic development, job availability and labor market conditions that affect men’s and women’s employment equally do not pose any omitted variable                                                              2 Of the total 66 countries in our sample, there are 5 high-income countries per the definition of the World Bank. The high-income countries include Chile, Estonia, Poland, Latvia and Uruguay. 3  For brevity, in table 2, only the correlations between our main explanatory variable (Maternity leave) and GDP per capita with all the other explanatory variables are shown.  7    bias problem for us. As Ruhm (1998) notes, since women use virtually all parental leave in most countries, men constitute a reasonable comparison group, and the “natural” experiment examines how changes in leave entitlements affect the gap between female and male outcomes. Main explanatory variable Our main explanatory variable is the (log of) the number of weeks of paid maternity leave as mandated by law in the country (Maternity Leave). The data source for the variable is the World Bank’s Women, Business and the Law project (WBL). The mean value of the variable equals 2.6 and the standard deviation equals 0.31. Across countries, Maternity Leave has the highest value of 3.9 (or over 52 weeks) in Albania and lowest value of 1.9 (7 weeks) in Lebanon. As discussed above, we explore how the relationship between maternity leave and female workers varies with who pays for maternity leave. To this end, we define a dummy variable equal to 1 if the employer pays for the entire maternity leave and 0 otherwise (Employer Pays); a dummy variable equal to 1 if the government and the employer jointly pay for maternity leave and 0 otherwise (Gov.-Employer Pays); and a dummy variable equal to 1 if the government alone pays for maternity leave and 0 otherwise (Government Pays). The data source for these dummy variables is WBL. In our sample, the employer alone pays for maternity leave in 18 countries, government and employer jointly pay in 13 countries while the government alone pays in 34 countries. Other explanatory variables Note that our dependent variable varies at the firm-level while the main explanatory variable, Maternity Leave, varies at the country level. Thus, reverse causality is unlikely to be a problem for 8    our results, although the possibility cannot be ruled out completely. There are several examples in the literature to support our claim. For instance, Pounov (2016) uses ES data to analyze the effect of corruption on firm’s decision to obtain quality certificates and patents. For corruption, it uses an aggregate (country-level) measure – the proportion of firms in the country that were asked to pay to bribe to get a business license. The study notes that using the aggregate measure of corruption helps to avoid the endogeneity concerns that would arise with a firm-level measure of corruption; it also helps to control for potential measurement error if some firms choose not to respond. The omitted variable bias problem is a more serious concern for us. To address this potential problem, we include several firm-level and country-level variables as controls in our regressions. The various controls can be divided into macro-level and micro-level controls. Consider the controls at the micro or firm-level. The data source for all these controls is ES. First, we consider the tendency for men and women to be segregated by industry (Juhn et al. 2014, Amin and Islam 2014, Do et al. 2011). Some of the reasons suggested for this in the literature include the brawn vs. brains content of jobs, the ease with which work can be combined with family responsibility, and culture. Such segregation may cause spurious correlation for our main results if the share of segregated industries in total employment varies systematically with the length of maternity leave across countries. Thus, we control for dummy variables indicating the industry to which the firm belongs (Industry fixed effects). There are 36 industries in our sample defined at the 2-digit ISIC (Rev. 3.1) level.4                                                              4  Some services industries with very few observations in the data are grouped together into one single category.  9    Out next controls include (log of) age and size of the firm. The latter equals (log of) number of full-time permanent workers at the firm at the end of the last fiscal year (Employment). Firm- size and age are important proxy measures of several firm attributes (see for example, Acs and Audretsch 1988, Cohen and Klepper 1996, Soderbom and Teal 2004), which could in turn affect firm outcomes such as workforce composition. For example, being more visible to the public, large firms may discriminate less against women workers than the small firms. On the other hand, small firms may be more flexible, offering a better work-family life balance critical for women’s participation in the labor market. The argument can be extended to young vs. old firms since older firms are likely to be more visible and younger firms more flexible in hiring practices. If firm-size and/or firm’s age also happen to vary systematically with the level of maternity leave across countries, failure to control for them could affect our main results spuriously. Next, female workers may also be more likely to be employed in firms managed by females, as female managers are less likely to discriminate against female workers than male managers. However, existing evidence on the issue is somewhat mixed (Nelson and Bridges 1999, Penner and Toro-Tulla 2010). In addition, since women tend to lag men in education and technical skills in many parts of the developing world, provision of training to employees by firms may be particularly attractive to women especially when informal networks in the firm are male- dominated (see for example, Rowley 2013, Ragins and Sundstrom 1989). These factors could cause an omitted variable bias problem with our results if they happen to vary systematically across countries with high vs. low levels of maternity leave. Thus, we control for a dummy variable equal to 1 if the top manager of the firm is a female and 0 otherwise (Female top manager); and a dummy variable equal to 1 if the firm provides formal training to its employees and 0 otherwise (Training). 10    We worry about aspects of globalization that may affect women’s employment and be correlated with maternity leave. For instance, there is substantial work on the link between exporting activity and employment of women (see for example, Ozler 2000, Juhn et al. 2014). While the issue is still debated, the literature tends to favor the view that in the case of developing countries exporting activity encourages female relative to male employment. Further, foreign ownership of the firms in developing countries may come with foreign values and culture that may be more favorable towards women employees. The potential for omitted variable bias here is evident. We guard against such possibility by controlling for a dummy variable equal to 1 if the firm exported any of its output during the last fiscal year and 0 otherwise (Exporter); and the percentage of the firm that is owned by private foreign individuals and companies (Foreign ownership). Our last set of firm-level controls relates to crime, severity of labor laws and lack of skilled workers as obstacles to firms’ operations. Women may be particularly vulnerable to crime and lack of security, as criminals tend to target women more than men (Glaeser and Sacerdote 1999; Islam 2013). Employers with a taste for discrimination may find it costlier to discriminate against women when it is more difficult to find adequately educated or skilled workers. Similarly, stringency or severity of labor laws could potentially affect a firm’s decision to hire women vs. men. For instance, more stringent labor laws may make women workers more attractive if they are more docile and more easily willing to work at lower wages than required by law compared with men (see for example Seguino 2000, Elson 1996). If crime, skills shortages and labor laws vary systematically with maternity leave, our main results could be spuriously affected. Thus, we control for a dummy variable equal to 1 if the firm experienced losses due to crime during the last fiscal year and 0 otherwise (Crime); severity level of the availability of adequately educated 11    workers (0-4 scale) as an obstacle for firms’ operations (Skills obstacle); and severity level of the labor laws (0-4 scale) as an obstacle for firms’ operations (Labor laws obstacle). For macro-level controls, we begin with GDP per capita. Overall economic development as reflected in the level of GDP per capita may lead to improved outcomes for women’s labor market participation through several channels such as better education and health for women; better child care facilities; etc. If maternity leave also varies between rich vs. poor countries, our main results could suffer from an omitted variable bias problem. We account for this potential problem by controlling for the (log of) GDP per capita (PPP adjusted and at constant 2011 international $). Value of GDP per capita used is the average value over the five-year period prior to the year the ES was conducted. The data source is WDI, World Bank. Employment of men and women may also be affected by the current level of labor demand. If the current level of demand varies systematically with the maternity leave, our main results could be spuriously affected despite controlling for GDP per capita and related variables. Thus, following Wamboye and Seguino (2015), we control for the current state of labor markets by capturing aggregate demand through the growth rate of GDP per capita (Growth rate). The data source for the variable is WDI, World Bank. Average values of the variable over the five years prior to the year the ES was conducted are used. A larger proportion of females in the total (male plus female) adult population may positively impact the proportion of workers in the country that are females. Thus, demographic factors that influence female and male population in the country may cause our main results to suffer from an omitted variable bias problem if they vary systematically with the level of maternity leave. We guard against such possibility by controlling for the share of females in the total adult population in the country (Female population). We use average values of the variable taken over 12    the last five years prior to the year the ES was conducted. The data source for the variable is WDI, World Bank. We account for the education level of women relative to men. There are many ways in which greater education for women can increase their participation in the labor market. For example, improved access to education for women adds to their skills and creates awareness among them about available opportunities in the labor markets; it raises the cost to women of staying home; more educated women in society helps to change social attitudes towards women’s work, making women’s labor market activity socially more acceptable (World Bank 2011, Contrareras and Plaza 2010). This could spuriously affect our main results if maternity leave and women’s access to education happen to be correlated. We guard against this possibility by controlling for the level of education among females relative to males in the country (Education). The variable is constructed in two steps using data from WDI, World Bank. In the first step, average values over the last five years prior to the year the ES was conducted are obtained separately for gross female to male enrollment rate in primary education and in secondary education. In the second step, a simple average is taken over the primary and secondary enrollment rates obtained in the first step. The overall structure of the economy – share of manufacturing, agriculture and services sectors – could have direct or indirect effects on female employment. For example, a larger agricultural or services sector (relative to manufacturing) may affect female relative to male employment as women tend to be concentrated (more than men) in the services and agriculture sectors. If the share of the manufacturing sector also varies systematically with the level of maternity leave, our main results could be spuriously affected. We follow the literature5 in                                                              5  See for example, Wamboye and Seguino 2014.  13    controlling such effects using the value added in manufacturing as a proportion of GDP (Manufacturing VA), and value added in agriculture as a proportion of GDP (Agriculture VA). The data source for these variables is WDI, World Bank. As above, we use average values of the variable taken over the last five years prior to the year the ES was conducted in the country. Cultural and institutional factors are important factors that determine women’s engagement in the labor market. Studies show that social and cultural attitudes towards women and their market activity are correlated with women’s participation in the labor market, although the evidence on the issue is rather mixed (see for example, World Bank 2011, Contreras and Plaza 2010, Hajj and Panizza 2009). For culture or social attitudes, we follow the literature and use two sets of proxy measures. First, the proportion of population that is Catholic, Protestant and Muslim (the residual category of other religions is the omitted category). Second, a dummy variable equal to 1 if the legal origin of the country is Socialist and 0 otherwise. The data source is La Porta et al. (1999). Laws on maternity leave are not the only laws that impact women’s employment. Other gender-specific laws could also impact female relative to male employment (see for example, Amin et al. 2016, Gonzales et al. 2015, Branisa et al. 2013, Islam et al. 2018). Thus, our main results concerning maternity leave could suffer from omitted variable bias if the other laws happen to vary systematically with the length of maternity leave across countries. To guard against such possibility, we control for various gender-specific laws all taken from the WBL data and for the year 2016. Specifically, we control for an overall measure of gender disparity in the laws (WBL disparity).6 Additionally, following Amin et al. (2016), we control for a dummy variable equal to                                                              6  The  WBL disparity variable is an overall measure of Legal Gender Disparities based on WBL data (for 2016) and is  obtained from Iqbal et al. (2016). It is constructed as follows. Fifty‐one gender disparities in laws are considered. If  a law treats men and women differently (a legal gender disparity) then a score is assigned. If the disparity applies to  only married women, a score of 1 is assigned. If it applies to both married and unmarried women, a score of 2 is  assigned. If the law treats both men and women the same, then a score of 0 is assigned. The summation over all the  scores provides the WBL disparity variable.   14    1 if the country has laws mandating paid or unpaid paternity leave and 0 otherwise (Paternity Leave). Last, the ES data we use were collected in different years across countries between 2007 and 2016. This raises the possibility that global factors that vary over time could be affect women’s employment. If these factors also vary systematically with maternity leave across countries, our main results could be spuriously affected. Thus, we control for the year the ES was conducted in the country (Trend).7 One variable that we did not mention above is the fertility rate. Higher fertility rates are known to adversely affect women’s labor force participation rates; at the same time, maternity leave policies may be systematically correlated with the fertility rate. In our sample, fertility rate is very highly correlated with GDP per capita (correlation of -0.78). Thus, we expect that the effect of fertility rate on female workers is already accounted for via the GDP per capita control. Nevertheless, we experimented with controlling for fertility rate in all the specifications discussed below.8 However, controlling for fertility rate and even its interaction with maternity leave made no qualitative difference to any of the results discussed below. 3. Main empirical results All the regression results discussed below are based on the Ordinary Least Squares regression method. Sampling weights are used throughout to ensure that the results can be extrapolated to the targeted universe of the private economy.9                                                              7  Our main results do not change much if we control for year dummies (year fixed effects) instead of the year trend.  8  Data source for fertility rate is WDI, World Bank. As for GDP per capita and other macro‐level controls, we use  average  values  of  fertility  rate  taken  over  5  years  prior  the  year  the  ES  was  conducted  in  the  country.  Results  controlling for fertility rate are available on request from the authors.  9  Sampling weights are normalized so they sum to unity across all firms within a given country. Thus, all countries  have equal weightage.  15    Our main regression results for the relationship between female workers and maternity leave are provided in tables 3 and 4. Table 3 contains results for the base specification while table 4 provides robustness checks. Results in table 3 show that there is a positive relationship between female workers and maternity leave that is economically large and statistically significant at the 1 percent level. Without any other controls, the estimated coefficient value of maternity leave equals 16.6. This implies that an increase of 1 log unit in Maternity Leave (or an increase of about 19 days) is associated with an increase of 16.6 percentage points in the share of female workers (in a typical firm in the country). Alternatively, moving from a country with the smallest length of maternity leave (Lebanon) to the country with the highest length (Albania) is associated with an increase of about 33 percentage points in the share of female workers. While this is a large increase, later results show a more moderate increase. Next, we add sequentially the various baseline controls to the specification. That is, we add controls for industry and year trend (column 2), GDP per capita (column 3), size and age of the firm (column 4), exporting status and foreign ownership (column 5) and female population (column 6). None of these controls alters the results for maternity leave qualitatively from above. There is some decline in the estimated coefficient value of maternity leave, but it is still large and significant (at 1 percent level). For example, with all the controls included in the specification, the estimated coefficient value of maternity leave equals 13.3 (column 6), only somewhat lower than the value of 16.6 found above with no controls in place (column 1). As is evident from table 3, the decline in the coefficient value is largely due to the control for female population. For the macro-level controls and as we might expect, GDP per capita and the share of females in total population show a strong positive relationship with female workers, statistically significant at the 1 percent level. As suggested above, exporting firms have more female workers 16    than non-exporting firms and the difference is significant at the 5 percent level; higher foreign ownership is also positively correlated with female workers, significant at the 1 percent level. Regarding firm-size, smaller firms have more female workers and the difference is significant at the 1 percent or 5 percent level. Robustness checks for our main results are provided in table 4. These checks involve adding to the final specification above the remaining firm-level and country-level controls discussed above. As is evident from table 4, the estimated relationship between female workers and maternity leave remains positive, large and statistically significant at the 1 percent level even with the additional controls included in the specification. The magnitude of the estimated coefficient value of maternity leave is also not affected much except by the socialist legal origin dummy (column 5) and to some extent the WBL disparity variable. Controlling for the socialist legal origin dummy causes the estimated coefficient value of maternity leave to decline sharply from 12.3 (column 4) to 6.1, but it remains significant at the 1 percent level (column 5); further controlling for WBL disparity causes the coefficient value to decline further to 5.1 (not shown), significant at the 1 percent level. For the various controls, results are along expected lines. First, results for the controls in the base specification (table 3) discussed above survive the robustness checks except for GDP per capita which is still positively correlated with female workers but not significantly so due to some of the macro-level controls. Thus, GDP per capita could be a proxy for some of the macro-level controls. For the new controls and as expected, having a female top manager and provision of training by the firms show a strong positive correlation with female workers, significant at the 1 percent level. A larger share of Muslims in the population relative to Catholics, Protestants and the omitted category of all other religions is associated with a significantly lower (at 1 percent 17    level) share of female workers. As expected, greater gender disparity in the laws as measured by WBL disparity is associated with a significantly lower (at 1 percent level) share of female workers. Last, a firm in a Socialist legal origin country vs. elsewhere has a higher share of female workers by about 8 percentage points and this difference is significant at the 1 percent level (column 6). Summarizing, there is a large, positive and statistically significant relationship between female workers and maternity leave that is robust to several firm-level and country-level controls. According to the most conservative estimate so far (column 6, table 4), an increase in maternity leave days by 1 week (increase of 0.69 in the Maternity Leave variable) is associated with a 3.5 percentage point increase in the share of female workers in a typical firm in our sample. 4. Extensions 4.1 Employer vs. government funded maternity leave We argued above that the positive relationship between maternity leave and female workers is likely to be stronger as the cost of maternity leave shifts from the employer to the government. This hypothesis also serves as a robustness check against endogeneity concerns. We report regression results for employer vs. government funded maternity leave using the split sample and the interaction term method. Regression results using the split sample method are provided in table 5. Regression results for the sample of countries where maternity leave is entirely funded by the employer are provided in columns 1-6 while columns 7-12 contain results for the sample of countries where maternity leave is entirely funded by the government.10 For brevity, results are shown for only some of the specifications discussed above. As expected, the results show that the relationship between                                                              10  Countries where maternity leave is in part funded by the government and in part by the employer are excluded  in table 5.  18    maternity leave and female workers is much larger (more positive) in countries where maternity leave is funded by the government than in countries where it is funded by the employer. For example, for our baseline specification, the estimated coefficient value of maternity leave equals 10.3 when maternity leave is funded by the employer and a larger 13.3 percent when it is funded by the government (column 3 and 9); both the coefficient values are significant at the 1 percent level. The qualitative nature of the results discussed here remains unchanged when we add the remaining controls to the specification. For instance, with all the controls discussed above included in the specification, the estimated coefficient value of maternity leave equals 8.9 (column 12) when maternity leave is funded by the government and a much lower negative coefficient value of -9.2 in countries with employer funded maternity leave; both the coefficient values are significant at the 1 percent and 5 percent level, respectively. Some caveats are in order here. First, the negative and significant coefficient value of maternity leave here for countries with employer funded maternity leave is largely due to a single country, Mauritius. Excluding Mauritius from the sample, the estimated coefficient value of maternity leave drops sharply (in absolute value) from -9.3 above to -5.0 and is no longer significant (at the 10 percent level or less). Second, for the employer funded maternity leave sample, the estimated relationship between female workers and maternity leave is quite sensitive to the controls for Socialist legal origin dummy and paternity leave. While the results discussed above in this paragraph hold qualitatively even when controls for Socialist legal origin and paternity leave are excluded from the specification, more work is required to ascertain the robustness of the findings. Regression results using the interaction term between maternity leave and government vs. employer funded maternity leave are provided in table 6. In these regressions, we control for the interaction term between maternity leave and the dummy for employer funded maternity leave 19    (Maternity Leave*Employer pays) and the interaction term between maternity leave and the dummy for maternity leave jointly funded by the government and the employer (Maternity Leave* Gov-Employer pays). Thus, the coefficient value of Maternity Leave*Employer pays gives the difference in the strength of the relationship between maternity leave and female workers when maternity leave is fully funded by the employer versus when it is fully funded by the government. As discussed above, the expected sign of the coefficient is negative. Regression results provided in table 6 confirm this expectation and are consistent with the split sample results discussed above. That is, the estimated coefficient value of the interaction term between maternity leave and the dummy for employer funded maternity leave is negative, economically large and statistically significant at the 5 percent level in one specification and at the 1 percent level in the remaining specifications. One concern with the interaction term results discussed in the previous paragraph is that the interaction term could be spuriously picking up heterogeneity in the female workers and maternity leave relationship with respect to other variables. For instance, if richer countries are more likely to have government funded maternity leave, then our results for the interaction term between maternity leave and who pays for it could be spuriously picking up the differential impact of maternity leave on rich vs. poor countries. To guard against this possibility, we interact the maternity leave variable with all the control variables discussed above (as shown in table 4).11 Adding all the resulting interaction terms to the final specification above (i.e., the specification in column 6 of table 6) did not change the qualitative nature of the results discussed in the previous paragraph.                                                              11  For industry, we use the interaction term Maternity Leave and a dummy variable equal to 1 if the firm is in the  manufacturing sector and 0 otherwise.  20    4.2 Female population Table 7 provides the split sample results for high vs. low values of female population. Columns 1- 5 contain results for the sample of countries with below median female population levels while the same for above median female population countries are provided in columns 6-10. As suggested above, the regression results in table 7 clearly show that the relationship between maternity leave and female workers is much stronger (more positive) in countries with proportionately more females in total population. Results using the interaction term between maternity leave and female population are provided in table 8. We make one modification here in that we also control for the interaction term between GDP per capita and maternity leave (GDP per capita*Maternity Leave) variable to address any lingering concern that our results for the interaction term between maternity leave and female population are spuriously picking up the broader overall development effects. Note that the control for GDP per capita*Maternity Leave is not critical for the qualitative nature of the results discussed here. Results in table 8 confirm the findings using the split sample approach. That is, the interaction term between maternity leave and female population is positive, large and statistically significant at the 1 percent level in all the specifications. One concern with the interaction term results discussed in the previous paragraph could be that the interaction term could be spuriously picking up heterogeneity in the female workers and maternity leave relationship with respect to other variables. For instance, if richer countries have more females in the total population, our results for the interaction term between maternity leave and share of females in total population discussed in the previous paragraph could be spuriously picking up the differential impact of maternity leave on rich vs. poor countries. To guard against this possibility, we interact the maternity leave variable with all the control variables discussed 21    above (as shown in table 4) and use these as additional controls. 12 We also include here the controls for who pays for maternity leave (Employer pays, Gov-Employer pays) and their interaction terms with maternity leave. Adding all these additional controls to the final specification above (i.e., specification in column 6 of table 8) did not change the qualitative nature of the results discussed in the previous paragraph. 4.3 Excluding Socialist legal origin countries The results above show that controlling for Socialist legal origin countries has a significant impact on the estimated strength of the relationship between maternity leave and female workers. To address any lingering concerns, we repeat our baseline regressions in tables 3 and 4 after excluding the Socialist legal origin countries from the sample. However, this did not change the qualitative nature of the results discussed above for maternity leave (table 9). As table 9 shows, the relationship between maternity leave and female workers is still positive, large and statistically significant at the 5 percent level in one specification and at the 1 percent level in all the other specifications. Quantitatively, the estimated coefficient value of maternity leave declines when we exclude Socialist legal origin countries. For example, with all the controls discussed above in place, the estimated coefficient value of maternity leave equals 3.8 (column 6, table 9) compared with 5.0 in the full sample and after controlling for the Socialist legal origin dummy (column 6, table 4). 5. Conclusion                                                              12  For industry, we use the interaction term Maternity Leave and a dummy variable equal to 1 if the firm is in the  manufacturing sector and 0 otherwise.  22    The relationship between the length of paid mandated maternity leave and the presence of female relative to total workers in the formal private sector is explored. Using firm-level survey data for a large cross-section of mostly developing countries, a large positive relationship is found between maternity leave and the proportion of workers at a firm that are females. The relationship survives several robustness checks. The empirical results suggest that this relationship is far from uniform, being much stronger in countries where maternity leave is entirely funded by the government compared to the employer. Several issues remain to be explored. We provide a few examples here. First, there is little in the theoretical literature on how the relationship between maternity leave and female employment depends on firm and country characteristics. Understanding the theoretical basis for this heterogeneity in the maternity leave and female employment relationship can help better understand the impact of maternity leave and may ultimately lead to better policy design. Second, due to data limitations, we were unable to say anything on how other outcome measures may be influenced by maternity leave. Examples of outcome measures include wages of women vs. men, occupational segregation of men and women workers, etc. Future work along the lines in this paper can help shed light on how such outcomes are related to maternity leave and if the relationship is the same or varies with firm and country characteristics such as firm-size, age of the firm, GDP per capita, etc. Last, implications of maternity leave on income equality in the country is another important area for future research. 23    24    References Acs, Zoltan J. and David Audretsch (1988), “Innovation in Large and Small Firms: An Empirical Analysis,” American Economic Review, 78 (4): 678-690. Amin, Mohammad and Asif Islam (2014), “Are There More Female Managers in the Retail Sector? Evidence from Survey Data in Developing Countries,” Journal of Applied Economics, 17(2): 213-228. Amin, Mohammad, Asif Islam and Alena Sakhonchik (2016), “Does paternity leave matter for female employment in developing economies? Evidence from firm-level data.” Applied Economics Letters 23(16): 1145-1148. Becker, G.S. (1991), A Treatise on the Family, Enlarged edition. Harvard University Press. Berger Lawrence M. and Jane Waldfogel (2004), “Maternity Leave and the Employment of New Mothers in the United States,” Journal of Population Economics, 17(2): 331-349. Bertrand, Marianne, Claudia Goldin and Lawrence Katz (2010), “Dynamics of the Gender Gap for Young Professionals in the Financial and Corporate Sectors,” American Economic Journal: Applied Economics, 2(3): 228-55. Branisa, Boris, Stephan Klasen, and Maria Ziegler (2013), “Gender Inequality in Social Institutions and Gendered Development Outcomes.” World Development 45:252–68. Calvó-Armengol, A. and M. O. Jackson (2004), “The Effects of Social Networks on Employment and Inequality,” American Economic Review 94 (3): 426-454. Cohen, Wesley M. and Steven Klepper (1996), “A Reprise of Size and R&D,” Economic Journal, 106(437): 925-951. Contreras, Dante and Gonzalo Plaza (2010), “Cultural Factors in Women's Labor Force Participation in Chile,” Feminist Economics, 16(2): 27-46. Dahl, Gordon B., Katrine V. Løken, Magne Mogstad, and Kari Vea Salvanes (2016), “What is the Case for Paid Maternity Leave?” The Review of Economics and Statistics, 98(4): 655-670. Do, Quy-Toan, A. Levchenko and C. Raddatz (2011), “Engendering Trade,” Policy Research Working Paper No. 5777, World Bank, Washington DC. Drange, Nina and Mari Rege (2013), “Trapped at Home: The Effect of Mothers' Temporary Labor Market Exits on their Subsequent Work Career,” Labour Economics, 24: 125-136. Elson, D. (1999), “Labor Markets as Gendered Institutions: Equality, Efficiency and Empowerment Issues”, World Development, 27(3): 611-627. 25    Phipps, Shelley, Peter Burton, and Lynn Lethbridge (2011), “In and out of the Labour Market: Long-Term Income Consequences of Child-Related Interruptions to Women's Paid Work,” The Canadian Journal of Economics 34(2): 411-29. Glaeser, E.L. and B.I. Sacerdote (1999), “Why Is There More Crime in Cities,” Journal of Political Economy, 107(S6): S225-S258. Goldin, C. (2006), “The Quiet Revolution That Transformed Women’s Employment, Education, and Family,” American Economic Review 96(2): 1–21. Gonzales, Christian, Sonali Jain-Chandra, Kalpana Kochhar, and Monique Newiak (2015), “Fair Play: More Equal Laws Boost Female Labor Force Participation,” IMF Staff Discussion Note SDN/15/02 Hajj, M. and U. Panizza (2009), “Religion and Education Gender Gap: Are Muslims Different?” Economics of Education Review, 28: 337-344. Hanratty, M. and Trzcinski, E. (2009), “Who Benefits from Paid Family Leave? Impact of Expansions in Canadian Paid Family Leave on Maternal Employment and Transfer Income,” Journal of Population Economics, 22(3): 693-711. Hashimoto, Masanori, Rick Percy, Teresa Schoellner and Bruce A. Weinberg (2004) “The Long and Short of it: Maternity Leave Coverage and Women’s Labor Market Outcomes”, IZA Discussion Paper No. 1207, Bonn. Islam, Asif (2013), “Is There a Gender Bias in Crime Against Firms in Developing Economies?” Women’s Studies International Forum, 37: 1-15. Iqbal, Sarah, Asif Islam, Rita Ramalho and Alena Sakhonchik (2016), “Unequal Before the Law: Measuring Legal Gender Disparities Across the World,” Policy Research Working Paper 7622, World Bank, Washington, DC. Islam, Asif, Silvia Muzi and Mohammad Amin (2018), “Unequal Laws and the Disempowerment of Women in the Labor Market: Evidence from Firm-Level Data,” Forthcoming Journal of Development Studies, DOI: 10.1080/00220388.2018.1487055 Juhn, Chinhui, Gergely Ujhelyi, Carolina Villegas-Sanchez (2014), “Men, Women, and Machines: How Trade Impacts Gender Inequality,” Journal of Development Economics, 106: 179-193. Klerman, Jacob Alex and Arleen Leibowitz (1997), “Labor Supply Effects of State Maternity Leave Legislation,” in Francine D. Blau and Ronald G. Ehrenberg (eds.) Gender and Family Issues in the Workplace, New York: Russell Sage Foundation. Klerman, Jacob A. and Arleen Leibowitz (1999), “Job Continuity among New Mothers,” Demography 36(2): 145-155. 26    Kluve, Jochen and Schmitz, Sebastian, (2018), “Back to Work: Parental Benefits and Mothers Labor Market Outcomes in the Medium Run,” International and Labor Relations Review, 71(1): 143-173. La Porta, Rafael, Florencio Lopez-de Silanes, Andrei Shleifer and Robert Vishny (1999), “The Quality of Government,” Journal of Law, Economics and Organization, 15(1): 222-79. Lalive, R. and J. Zweimüller (2009), “How Does Parental Leave Affect Fertility and Return-to- Work? Evidence from Two Natural Experiments,” Quarterly Journal of Economics, 24: 1363- 1402. Lalive, Rafael, Analia Schlosser and Josef Zweimüller (2010) “How Do Employment Protection and Maternity Leave Benefits Affect Mother's Post-Birth Careers?”, Mimeograph. Available online at: https://www.researchgate.net/publication/228426576_How_Do_Employment_Protection_and_P arental_Leave_Benefits_Affect_Mother's_Post-Birth_Careers Ludsteck, J. and U. Schönberg, (2007) “Maternity Leave Legislation, Female Labor Supply, and the Family Wage Gap” IZA Discussion Paper No. 2699, Bonn. Mincer, J. and S. Polachek (1974), “Family Investments in Human Capital: Earnings of Women,” Journal of Political Economy, 82 (2): 76–108 (Part 2). Nelson, Robert L. and William P. Bridges (1999), Legalizing Gender Inequality: Courts, Markets, and Unequal Pay for Women in America, Cambridge University Press. Ondrich, J., C. K. Spiess, Q. Yang and G. G. Wagner (2003), “The Liberalization of Maternity Leave Policy and the Return to the Work of Childbirth in Germany,” Review of Economics of the Household, 1, 77-110. Ozler, S. (2000), “Export Orientation and Female Share of Employment: Evidence from Turkey,” World Development, 28(7): 1239–48. Penner, Andrew M. and Harold J. Toro-Tulla (2010), Women in power and gender wage inequality: the case of small businesses, in Christine L. Williams, Kirsten Dellinger (ed.) Gender and Sexuality in the Workplace (Research in the Sociology of Work, Volume 20), Emerald Group Publishing Limited, pp.83-105. Phipps, Shelly, Peter Burton, and Lynn Lethbridge (2001), “In and Out of the Labour Market: Long-term Income Consequences of Child-Related Interruptions to Women’s Paid Work,” Canadian Journal of Economics, 34(2): 411-429. Pounov, Caroline 92016), “Corruption’s Asymmetric Impacts on Firm Innovation,” Journal of Development Economics 118: 216-231. 27    Ragins, B. R. and E. Sundstrom (1989), “Gender and Power in Organisations,” Psychological Bulletin, 105: 51-88. Rowley, Chris (2013), “What Might Influence Female Manager Careers Success?” Mimeograph, Cass Business School, City University London. Ruhm, Christopher J. (1998) “The Economic Consequences of Maternity Leave Mandates: Lessons from Europe,” Quarterly Journal of Economics, 113(1): 285-317. Schöneberg, U. and J. Ludsteck (2014), “Expansions in Maternity Leave Coverage and Mothers’ Labour Market Outcomes after Childbirth,” Journal of Labor Economics, 32(3): 469-505. Seguino, Stephanie (2000), “Accounting for Gender in Asian Economic Growth,” Feminist Economics, 6:3, 27-58. Söderbom, Måns, and Francis Teal (2004), “Size and Efficiency in African Manufacturing Firms: Evidence from Firm‐Level Panel Data,” Journal of Development Economics, 73 (February): 369– 94. Waldfogel, Jane, Yoshio Higuchi and Masahiro Abe (1999) “Family Leave Policies and Women's Retention After Childbirth: Evidence from the United States, Britain, and Japan,” Journal of Population Economics, 12(4): 523-545. Wamboye, Evelyn F. and Stephanie Seguino (2015), “Gender Effects of Trade Openness in Sub- Saharan Africa.” Feminist Economics, 21:3, 82-113. World Bank (2011), Gender Equality and Development, World Development Report 2012, World Bank, Washington DC. World Bank (2018), Women, Business and the Law 2018, World Bank, Washington DC. 28    Table 1: Summary statistics Std. Variable Mean Minimum Maximum deviation Firm-level variables Female workers 32.4 27.3 0 100 Maternity Leave (logs) 2.6 0.3 1.9 4 Manufacturing (dummy) 0.3 0.5 0 1 Employment (logs) 2.8 1.1 0 10.3 Firm age (logs) 2.5 0.8 0 5.4 Exporter (dummy) 0.2 0.4 0 1 Foreign ownership 7 23.8 0 100 Female top manager 0.2 0.4 0 1 Training 0.4 0.5 0 1 Crime 0.2 0.4 0 1 Labor laws obstacle 1 1.1 0 4 Skills obstacle 1.3 1.3 0 4 Country-level variables: GDP per capita (logs) 8.7 0.9 6.6 10.1 Growth rate 3.8 2.3 -0.6 10.9 Female population 1 0 0.9 1.2 Education 1 0.1 0.8 1.15 Manufacturing VA 15 5.6 4 32.1 Agriculture VA 15.1 10 2.5 46.2 Muslim 21.4 32.5 0 99.4 Catholic 35.9 39.5 0 96.6 Protestant 6.4 12.6 0 66 Socialist 0.3 0.4 0 1 WBL disparity 17.4 7.1 5.9 40 Paternity Leave 0.4 0.5 0 1 Employer Pays 0.27 0.44 0 1 Gov.-Employer Pays 0.22 0.42 0 1 Government Pays 0.51 0.5 0 1 Sample size: 48,647 firms located in 66 countries 29    Table 2: Correlations Maternity GDP per Leave capita (logs) (logs) Maternity Leave (logs) 1 GDP per capita (logs) 0.26 1 Manufacturing (dummy, firm-level) 0.02 0.00 Employment (logs) 0.03 -0.01 Firm age (logs) -0.10 0.07 Exporter (dummy) 0.01 0.05 Foreign ownership -0.01 -0.06 Female population 0.18 -0.09 Female top manager (dummy) 0.04 0.03 Training 0.00 0.04 Crime -0.05 0.06 Labor laws obstacle -0.16 0.07 Skills obstacle -0.10 0.04 Growth rate 0.10 0.01 Education 0.11 0.34 Manufacturing VA 0.10 0.43 Agriculture VA -0.23 -0.87 Muslim -0.05 -0.14 Catholic -0.23 0.14 Protestant 0.00 -0.02 Socialist 0.69 0.30 WBL disparity -0.27 -0.12 Paternity Leave 0.12 0.11 30    Table 3: Base regression results Dependent variable: Female (1) (2) (3) (4) (5) (6) workers Maternity Leave (logs) 16.593*** 17.693*** 15.871*** 15.642*** 15.713*** 13.279*** (0.921) (0.885) (0.914) (0.929) (0.930) (0.959) GDP per capita (logs) 2.559*** 2.612*** 2.618*** 3.078*** (0.363) (0.368) (0.371) (0.375) Employment (logs) -0.708** -0.995*** -0.945*** (0.279) (0.291) (0.291) Firm age (logs) -0.827** -0.740* -0.642 (0.409) (0.408) (0.402) Exporter 1.852** 1.949** (0.859) (0.859) Foreign ownership 0.044*** 0.040*** (0.012) (0.012) Female population 67.117*** (5.422) Trend (year) -0.843*** -0.591*** -0.601*** -0.611*** -0.601*** (0.152) (0.152) (0.153) (0.153) (0.154) Industry fixed effects Yes Yes Yes Yes Yes Constant -11.13*** 1672*** 1147*** 1170*** 1,189*** 1102*** (2.402) (305.234) (306.606) (309.205) (308.904) (309.969) Observations 48,647 48,647 48,647 48,647 48,647 48,647 R-squared 0.034 0.172 0.179 0.180 0.182 0.196 Standard errors in brackets. Significance is denoted by *** (1%), ** (5%), * (10%) 31    Table 4: Robustness of the main results Dependent variable: (1) (2) (3) (4) (5) (6) Female workers Maternity Leave (logs) 12.962*** 12.603*** 12.449*** 12.283*** 6.054*** 5.021*** (0.922) (0.930) (0.938) (0.970) (1.192) (1.264) GDP per capita (logs) 2.574*** 2.635*** 1.712*** 1.314** 0.569 0.694 (0.369) (0.369) (0.644) (0.657) (0.662) (0.669) Employment (logs) -0.819*** -0.791*** -0.866*** -0.598** -0.641** -0.622** (0.291) (0.291) (0.292) (0.287) (0.287) (0.286) Firm age (logs) -0.588 -0.508 -0.457 -0.586 -0.328 -0.369 (0.383) (0.383) (0.388) (0.384) (0.377) (0.377) Exporter 2.037** 2.029** 2.405*** 2.454*** 2.189*** 2.136*** (0.818) (0.821) (0.818) (0.799) (0.804) (0.810) Foreign ownership 0.039*** 0.039*** 0.043*** 0.032*** 0.035*** 0.034*** (0.011) (0.011) (0.011) (0.011) (0.011) (0.011) Female population 57.265*** 56.153*** 54.849*** 28.471*** 21.463*** 19.855*** (5.215) (5.238) (5.435) (5.774) (5.773) (5.785) Female top manager 18.105*** 18.055*** 17.743*** 16.651*** 16.352*** 16.343*** (0.849) (0.851) (0.856) (0.852) (0.848) (0.847) Training 3.652*** 3.761*** 3.388*** 2.768*** 2.534*** 2.488*** (0.650) (0.659) (0.669) (0.665) (0.662) (0.662) Crime 0.526 0.730 0.127 0.410 0.451 (0.788) (0.784) (0.792) (0.784) (0.785) Labor laws obstacle -0.606** -0.648** -0.748*** -0.404 -0.332 (0.287) (0.288) (0.286) (0.285) (0.286) Skills obstacle -0.201 -0.136 -0.103 -0.106 -0.118 (0.275) (0.277) (0.277) (0.277) (0.277) Growth rate 0.483*** -0.007 0.050 0.050 (0.143) (0.162) (0.162) (0.173) Education 15.858*** 9.732 6.639 7.869 (6.083) (6.050) (6.033) (5.912) Manufacturing VA 0.155*** 0.103* 0.104* 0.076 (0.053) (0.053) (0.053) (0.056) Agriculture VA -0.031 -0.045 -0.038 -0.041 (0.059) (0.060) (0.060) (0.060) Muslim -0.145*** -0.108*** -0.093*** (0.012) (0.013) (0.013) Catholic -0.027** 0.019 0.002 (0.011) (0.012) (0.012) Protestant -0.035 0.008 -0.013 (0.027) (0.028) (0.030) 32    Socialist legal origin (dummy) 9.184*** 8.398*** (1.028) (1.030) WBL disparity -0.177*** (0.056) Paternity Leave 1.655** (0.702) Trend (year) -0.664*** -0.690*** -0.470*** -0.276* -0.238 -0.170 (0.152) (0.153) (0.159) (0.157) (0.156) (0.160) Industry fixed effects Yes Yes Yes Yes Yes Yes Constant 1241*** 1297*** 845*** 500*** 449*** 317*** (306.602) (308.411) (321.145) (317.167) (316.117) (323.654) Observations 48,647 48,647 48,647 48,647 48,647 48,647 R-squared 0.262 0.263 0.267 0.283 0.290 0.292 Standard errors in brackets. Significance is denoted by *** (1%), ** (5%), * (10%) 33    Table 5: Split sample results for employer vs. government funded maternity leave Employer pays for entire maternity leave Government pays for entire maternity leave Dependent variable: Female (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) workers Maternity Leave 11.632*** 13.151*** 10.274*** 10.847*** 12.974*** -9.248** 17.296*** 15.057*** 13.268*** 13.924*** 8.806*** 8.899*** (logs) (1.503) (1.552) (1.573) (1.840) (2.096) (3.780) (1.295) (1.283) (1.308) (1.706) (1.932) (1.928) GDP per capita 2.836*** 4.940*** 8.290*** 9.597*** 11.879*** 5.522*** 4.551*** -1.031 -1.183 -1.310 (logs) (0.487) (0.531) (1.089) (1.263) (1.304) (0.702) (0.710) (1.170) (1.168) (1.166) Employment (logs) -1.356*** -1.063** -0.085 0.024 0.118 -0.732* -0.881** -0.632 -0.704* -0.634 (0.511) (0.508) (0.473) (0.472) (0.466) (0.401) (0.398) (0.387) (0.387) (0.386) Firm age (logs) -1.925*** -1.791*** -1.776*** -1.867*** -2.112*** -1.012 -0.604 -0.108 0.065 0.000 (0.599) (0.597) (0.572) (0.576) (0.579) (0.630) (0.617) (0.570) (0.561) (0.568) Exporter 3.681*** 4.271*** 2.591* 2.499* 2.303* -1.128 -0.912 0.655 0.369 0.288 (1.376) (1.381) (1.389) (1.396) (1.357) (1.106) (1.093) (1.002) (1.001) (1.002) Foreign ownership 0.046*** 0.031* 0.009 0.005 -0.011 0.044** 0.046*** 0.040** 0.044** 0.041** (0.017) (0.017) (0.018) (0.018) (0.017) (0.018) (0.017) (0.017) (0.017) (0.017) Female population 84.21*** 64.28*** 84.77*** 28.978 72.901*** 26.593*** 30.715*** 23.730*** (9.564) (11.678) (15.048) (18.535) (7.696) (8.220) (8.343) (8.825) Female top manager 17.647*** 17.735*** 17.715*** 17.342*** 16.970*** 16.983*** (1.361) (1.357) (1.357) (1.123) (1.113) (1.109) Training 1.730* 1.644 2.123** 3.146*** 3.204*** 3.189*** (1.006) (1.006) (0.990) (0.958) (0.956) (0.955) Crime 0.554 0.465 0.405 0.945 1.337 1.428 (1.212) (1.212) (1.215) (1.111) (1.100) (1.094) Labor laws obstacle -0.182 -0.247 0.048 -0.837** -0.619 -0.457 (0.510) (0.510) (0.508) (0.398) (0.398) (0.395) Skills obstacle -0.864** -0.904** -0.767* -0.159 -0.096 -0.058 (0.409) (0.410) (0.403) (0.395) (0.397) (0.397) Growth rate 0.182 0.227 2.972*** -0.197 -0.000 -0.075 (0.368) (0.372) (0.541) (0.280) (0.279) (0.290) Education 34.819** 37.360** 164.11*** 6.097 0.744 4.991 (16.507) (16.395) (22.999) (8.498) (8.492) (8.756) 34    Manufacturing VA 0.143 0.130 -0.817*** 0.420*** 0.298** 0.350*** (0.140) (0.140) (0.189) (0.123) (0.124) (0.126) Agriculture VA 0.225** 0.254*** 0.504*** -0.085 -0.055 -0.058 (0.093) (0.095) (0.103) (0.118) (0.118) (0.118) Muslim -0.116*** -0.105*** -0.054* -0.184*** -0.117*** -0.098*** (0.019) (0.019) (0.030) (0.017) (0.021) (0.021) Catholic 0.221*** 0.224*** -0.010 -0.054*** 0.015 -0.008 (0.032) (0.032) (0.047) (0.016) (0.019) (0.020) Protestant 0.070 0.051 0.308*** -0.040 0.016 -0.026 (0.053) (0.054) (0.064) (0.034) (0.036) (0.039) Socialist legal origin (dummy) -7.161** 5.040 8.653*** 7.322*** (2.823) (3.605) (1.516) (1.525) WBL disparity -0.236 -0.291*** (0.148) (0.076) Paternity Leave 19.725*** 0.187 (2.859) (0.938) Trend (year) -0.465* -0.144 1.010*** 1.002*** 1.003*** -0.076 -0.487* -0.615* -0.684** -0.485 (0.279) (0.276) (0.331) (0.333) (0.314) (0.279) (0.286) (0.334) (0.336) (0.335) Industry fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Constant -0.132 894.805 154.261 -2225*** -2247*** -2288*** -12.93*** 96.891 866.878 1198* 1347** 955.883 (3.646) (560.995) (555.802) (664.332) (665.742) (627.860) (3.546) (563.105) (577.739) (675.148) (679.029) (676.310) Observations 19,601 19,601 19,601 19,601 19,601 19,601 21,068 21,068 21,068 21,068 21,068 21,068 R-squared 0.013 0.133 0.148 0.240 0.241 0.257 0.037 0.223 0.240 0.341 0.346 0.349 Standard errors in brackets. Significance is denoted by *** (1%), ** (5%), 35    Table 6: Interaction term between maternity leave and who pays for it Dependent variable: (1) (2) (3) (4) (5) (6) Female workers Maternity Leave (logs) 17.296*** 49.232*** 54.820*** 82.690*** 86.726*** 86.742*** (1.282) (11.024) (10.468) (10.826) (10.811) (10.801) Maternity Leave * -5.664*** -5.768*** -7.395*** -7.546*** -4.808** -8.364*** Employer pays (2.014) (2.086) (2.013) (2.286) (2.283) (2.369) Maternity Leave * -2.482 11.809** 4.596 9.451* 2.134 1.421 Gov-Employer pays (4.718) (4.638) (4.491) (5.160) (5.307) (5.329) Employer pays 12.796** 15.583*** 20.458*** 20.727*** 15.776*** 24.694*** (5.150) (5.381) (5.212) (6.070) (6.059) (6.282) Gov.-Employer pays 7.413 -26.326** -8.783 -24.755* -5.130 -2.961 (12.422) (12.274) (11.897) (13.647) (14.037) (14.097) GDP per capita (logs) 13.616*** 14.533*** 21.424*** 23.934*** 23.898*** (3.116) (2.967) (3.227) (3.220) (3.217) GDP per capita*Maternity Leave -3.912*** -4.460*** -7.551*** -8.705*** -8.707*** (1.211) (1.147) (1.191) (1.190) (1.189) Employment (logs) -0.757*** -0.843*** -0.558* -0.629** -0.595** (0.277) (0.289) (0.285) (0.285) (0.285) Firm age (logs) -0.739* -0.492 -0.588 -0.325 -0.389 (0.399) (0.381) (0.382) (0.373) (0.374) Female population 80.034*** 67.359*** 40.623*** 34.491*** 33.450*** (6.061) (5.824) (6.210) (6.205) (6.208) Exporter 1.927** 2.091*** 1.822** 1.713** (0.818) (0.797) (0.802) (0.809) Foreign ownership 0.038*** 0.031*** 0.034*** 0.034*** (0.011) (0.011) (0.011) (0.011) Female top manager 17.884*** 16.648*** 16.369*** 16.377*** (0.842) (0.847) (0.838) (0.838) Training 3.788*** 2.740*** 2.596*** 2.585*** (0.664) (0.665) (0.660) (0.661) Crime 0.742 0.129 0.458 0.522 (0.787) (0.791) (0.780) (0.782) Labor laws obstacle -0.556* -0.685** -0.304 -0.241 (0.285) (0.284) (0.284) (0.284) Skills obstacle -0.229 -0.186 -0.192 -0.192 (0.274) (0.275) (0.276) (0.275) Growth rate -0.178 -0.181 -0.160 (0.176) (0.176) (0.185) Education 5.683 -0.339 1.664 (6.190) (6.153) (6.054) Manufacturing VA 0.146** 0.185*** 0.161** (0.063) (0.063) (0.064) Agriculture VA 0.018 0.049 0.030 (0.063) (0.063) (0.063) 36    Muslim -0.148*** -0.108*** -0.089*** (0.013) (0.014) (0.014) Catholic -0.007 0.046*** 0.028** (0.012) (0.013) (0.013) Protestant -0.010 0.039 0.019 (0.028) (0.029) (0.030) Socialist legal origin (dummy) 10.334*** 9.246*** (1.109) (1.107) WBL disparity -0.191*** (0.059) Paternity Leave 1.842*** (0.707) Trend (year) -0.680*** -0.770*** -0.302* -0.339* -0.264 (0.167) (0.166) (0.175) (0.175) (0.177) Industry fixed effects Yes Yes Yes Yes Yes Constant -12.928*** 1,151.658*** 1,332.480*** 354.224 425.893 276.160 (3.513) (333.625) (330.883) (349.749) (348.458) (353.687) Observations 48,647 48,647 48,647 48,647 48,647 48,647 R-squared 0.035 0.200 0.267 0.287 0.296 0.298 Standard errors in brackets. Significance is denoted by *** (1%), ** (5%), * (10%) 37    Table 7: Split sample results for female population Below median female population Above median female population Dependent variable: Female (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) workers Maternity Leave (logs) 13.903*** 13.695*** 8.199*** 5.199*** -0.674 20.446*** 20.735*** 21.639*** 20.317*** 12.120*** (1.049) (1.054) (1.135) (1.328) (1.495) (1.800) (2.273) (2.163) (2.249) (3.424) GDP per capita (logs) 3.156*** 3.732*** 3.722*** 4.095*** 2.467*** 1.611*** -0.214 -0.519 (0.647) (0.630) (1.117) (1.150) (0.594) (0.582) (0.975) (1.061) Employment (logs) -0.356 -0.421 -0.380 -0.258 -1.442*** -1.059** -0.869** -0.847** (0.405) (0.395) (0.383) (0.385) (0.415) (0.415) (0.411) (0.411) Firm age (logs) -1.467*** -1.278** -1.169** -1.007** 0.444 0.456 0.407 0.454 (0.568) (0.526) (0.516) (0.497) (0.562) (0.548) (0.546) (0.557) Exporter 2.151* 2.711** 3.413*** 2.899*** 0.838 1.180 2.049* 1.874* (1.247) (1.163) (1.091) (1.084) (1.186) (1.137) (1.126) (1.129) Foreign ownership 0.053*** 0.043** 0.028 0.022 0.030* 0.029** 0.027* 0.030** (0.018) (0.017) (0.018) (0.018) (0.016) (0.015) (0.015) (0.015) Female population 131*** 144*** 81*** 6.046 -16.453 -23.880 (13.264) (16.896) (18.965) (11.414) (13.507) (14.722) Female top manager 17.307*** 15.619*** 14.524*** 17.570*** 16.475*** 16.522*** (1.217) (1.198) (1.185) (1.170) (1.188) (1.192) Training 5.671*** 4.323*** 3.522*** 2.463*** 1.596* 1.611* (0.936) (0.940) (0.936) (0.905) (0.897) (0.898) Crime 0.216 0.023 0.233 0.948 0.261 0.232 (1.111) (1.097) (1.088) (1.078) (1.072) (1.073) Labor laws obstacle -1.829*** -1.576*** -1.097*** 0.477 -0.073 0.108 (0.371) (0.366) (0.370) (0.433) (0.443) (0.441) Skills obstacle -0.077 0.057 0.031 -0.357 -0.456 -0.469 (0.360) (0.359) (0.358) (0.414) (0.421) (0.421) Growth rate -0.627** -0.614** 0.591*** 0.884*** (0.254) (0.260) (0.216) (0.300) Education 7.575 18.384** 4.000 1.663 (8.070) (8.343) (11.169) (11.897) Manufacturing VA 0.416*** 0.236** 0.047 0.069 (0.080) (0.094) (0.109) (0.117) Agriculture VA 0.288*** 0.346*** -0.267*** -0.308*** 38    (0.102) (0.103) (0.081) (0.083) Muslim -0.208*** -0.120*** -0.117*** -0.084*** (0.023) (0.028) (0.015) (0.020) Catholic -0.082*** -0.019 0.016 0.040 (0.019) (0.021) (0.017) (0.025) Protestant -0.278** 0.123 -0.026 -0.008 (0.122) (0.125) (0.031) (0.040) Socialist legal origin (dummy) 11.987*** 5.934*** (1.480) (2.046) WBL disparity -0.185** -0.039 (0.075) (0.162) Paternity Leave 1.856* 1.281 (1.096) (1.357) Trend (year) -0.212 0.080 0.254 0.210 -0.904*** -1.224*** -0.722** -0.507 (0.200) (0.198) (0.217) (0.215) (0.248) (0.260) (0.315) (0.402) Industry fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Constant -6.760** 416 -294 -662 -515 -18.50*** 1,765*** 2,404*** 1,436*** 1,035*** (2.721) (403.228) (398.535) (435.290) (430.568) (4.728) (499.321) (519.982) (631.274) (807.680) Observations 30,599 30,599 30,599 30,599 30,599 18,048 18,048 18,048 18,048 18,048 R-squared 0.032 0.145 0.235 0.266 0.281 0.036 0.249 0.312 0.331 0.333 Standard errors in brackets. Significance is denoted by *** (1%), ** (5%), * (10%) 39    Table 8: Interaction term between female population and maternity leave Dependent variable: (1) (2) (3) (4) (5) (6) Female workers Maternity Leave (logs) -102.493*** -31.671 -59.283** -41.877 19.776 12.146 (17.990) (25.439) (24.872) (27.524) (27.584) (27.496) Maternity Leave * Female 115.037*** 67.468*** 88.164*** 95.904*** 45.630** 48.941** population (17.698) (19.437) (19.089) (20.914) (21.106) (21.029) Female population -233.399*** -104.044** -169.285*** -212.286*** -88.196 -98.756* (45.664) (51.604) (50.534) (54.905) (55.391) (55.278) GDP per capita (logs) 9.461*** 7.128** 13.926*** 18.638*** 17.726*** (3.177) (3.032) (3.525) (3.516) (3.518) GDP per capita*Maternity Leave -2.625** -1.947* -4.781*** -6.700*** -6.328*** (1.219) (1.155) (1.272) (1.271) (1.274) Employment (logs) -0.996*** -0.894*** -0.674** -0.678** -0.663** (0.290) (0.291) (0.286) (0.287) (0.286) Firm age (logs) -0.556 -0.418 -0.539 -0.293 -0.329 (0.401) (0.382) (0.381) (0.375) (0.375) Exporter 1.810** 1.853** 2.154*** 1.953** 1.914** (0.856) (0.818) (0.796) (0.799) (0.805) Foreign ownership 0.041*** 0.040*** 0.033*** 0.034*** 0.034*** (0.012) (0.011) (0.011) (0.011) (0.011) Female top manager 18.051*** 16.574*** 16.247*** 16.243*** (0.846) (0.846) (0.844) (0.842) Training 4.058*** 2.948*** 2.581*** 2.548*** (0.658) (0.661) (0.658) (0.657) Crime 0.757 0.226 0.439 0.484 (0.787) (0.789) (0.781) (0.782) Labor laws obstacle -0.502* -0.631** -0.306 -0.236 (0.286) (0.285) (0.284) (0.285) Skills obstacle -0.202 -0.147 -0.183 -0.189 (0.274) (0.275) (0.276) (0.275) Growth rate 0.056 0.067 0.061 (0.164) (0.164) (0.174) Education 0.404 -0.418 0.515 (6.297) (6.299) (6.193) Manufacturing VA 0.158*** 0.135** 0.112* (0.055) (0.055) (0.058) Agriculture VA 0.047 0.072 0.064 (0.062) (0.062) (0.062) Muslim -0.146*** -0.114*** -0.099*** (0.013) (0.013) (0.014) Catholic -0.016 0.024** 0.009 (0.011) (0.012) (0.012) Protestant 0.007 0.048* 0.025 (0.027) (0.027) (0.029) Socialist legal origin (dummy) 9.229*** 8.431*** (1.039) (1.036) 40    WBL disparity -0.176*** (0.057) Paternity Leave 1.431** (0.704) Trend (year) -0.696*** -0.772*** -0.468*** -0.430*** -0.364** (0.155) (0.154) (0.160) (0.160) (0.164) Industry fixed effects Yes Yes Yes Yes Yes Constant 231*** 1,408*** 1,649*** 1,021*** 789** 680** (46.284) (314.170) (311.840) (327.015) (327.328) (337.262) Observations 48,647 48,647 48,647 48,647 48,647 48,647 R-squared 0.050 0.198 0.265 0.287 0.294 0.296 Standard errors in brackets. Significance is denoted by *** (1%), ** (5%), * (10%) 41    Table 9: Excluding Socialist legal origin countries Dependent variable: Female (1) (2) (3) (4) (5) (6) workers Maternity Leave (logs) 6.968*** 7.171*** 4.958*** 4.459*** 6.154*** 3.807** (1.537) (1.542) (1.590) (1.473) (1.588) (1.714) GDP per capita (logs) 1.928*** 2.836*** 2.317*** 1.224* 1.723** (0.411) (0.429) (0.426) (0.727) (0.744) Employment (logs) -1.190*** -1.065*** -1.041*** -0.724** -0.689** (0.340) (0.340) (0.343) (0.335) (0.334) Firm age (logs) -0.481 -0.532 -0.610 -0.754* -0.664 (0.434) (0.428) (0.414) (0.412) (0.409) Exporter 2.600*** 2.753*** 2.810*** 3.754*** 3.696*** (0.978) (0.974) (0.948) (0.904) (0.915) Foreign ownership 0.048*** 0.043*** 0.040*** 0.023* 0.021* (0.013) (0.013) (0.012) (0.013) (0.012) Female population 68.573*** 54.911*** 20.441*** 17.104** (6.625) (6.389) (6.959) (6.910) Female top manager 16.711*** 14.794*** 14.821*** (1.029) (1.030) (1.028) Training 3.858*** 2.416*** 2.343*** (0.771) (0.776) (0.777) Crime 1.395 0.346 0.363 (0.856) (0.854) (0.859) Labor laws obstacle -0.126 -0.422 -0.288 (0.312) (0.307) (0.309) Skills obstacle 0.121 0.040 -0.019 (0.308) (0.303) (0.303) Growth rate -0.164 -0.369* (0.205) (0.221) Education 9.932 20.223*** (7.247) (7.311) Manufacturing VA -0.003 -0.125 (0.078) (0.084) Agriculture VA -0.097 -0.082 (0.065) (0.065) Muslim -0.150*** -0.127*** (0.016) (0.016) Catholic -0.001 -0.024 (0.014) (0.015) Protestant 0.024 0.023 (0.034) (0.036) 42    WBL disparity -0.237*** (0.065) Paternity Leave 3.608*** (0.821) Trend (year) -0.697*** -0.651*** -0.680*** -0.095 0.078 (0.153) (0.153) (0.153) (0.159) (0.165) Industry fixed effects Yes Yes Yes Yes Yes Constant 11.478*** 1,383*** 1,216*** 1,291*** 151.700 -196.288 (3.780) (307.869) (308.757) (308.755) (322.639) (335.664) Observations 37,036 37,036 37,036 37,036 37,036 37,036 R-squared 0.003 0.139 0.154 0.216 0.248 0.254 Standard errors in brackets. Significance is denoted by *** (1%), ** (5%), * (10%) 43