WPS6843 Policy Research Working Paper 6843 Are There More Female Managers in the Retail Sector? Evidence from Survey Data in Developing Countries Mohammad Amin Asif Islam The World Bank Development Economics Global Indicators Group April 2014 Policy Research Working Paper 6843 Abstract This paper uses firm-level data for 87 developing from other service sectors, such as wholesale, countries to analyze how the likelihood of a firm having construction, and other services. The analysis also finds female vs. male top manager varies across sectors. The that the higher presence of female managers in the retail service sector is often considered to be more favorable sector vs. manufacturing is much higher among the toward women compared with men vis-à-vis the relatively small firms and firms located in the relatively manufacturing sector. Although the exploration of the small cities. These findings could serve as useful inputs data confirms a significantly higher presence of female for the design of optimal policy measures aimed at managers in services vs. manufacturing, the finding is promoting gender equality in a country. entirely driven by retail firms, with little contribution 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://econ.worldbank.org. The author may be contacted at aislam@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 Are There More Female Managers in the Retail Sector? Evidence from Survey Data in Developing Countries Mohammad Amin* and Asif Islam** Keywords: Gender, Labor market, Managers, Retail sector JEL: J16, G30 _______________________________ *Enterprise Analysis Unit, Development Economics (DECEA), World Bank, Washington DC, 20433. Email: mamin@worldbank.org. ** Corresponding author, Enterprise Analysis Unit, Development Economics (DECEA), World Bank, Washington DC, 20433. Email: asif.m.islam@gmail.com 1. Introduction A number of studies have documented gender-based disparities favoring men over women in labor market participation rates and wages (Duflo, 2005; Tzannatos, Zafiris, 1999). Studies have also found that men and women workers and owners tend to be concentrated in different sectors and occupations. In other words, relative to men, certain sectors or jobs seem to be more favorable towards females than other sectors or jobs. For example, one view is that relative to men, women have a comparative advantage in occupations that require less use of brawn and greater use of brains (Rendall, 2010). Another view is that in a majority of countries, females are the primary caregivers in the family affecting their labor market participation, wages and also the types of jobs they can undertake (Becker, 1985; Bielby and Bielby, 1988; Hundley, 2000 and Lombard, 2001). Of course, in addition to self-selection by females, employer discrimination against females could be another explanation for the observed occupation segregation along gender lines. The present paper contributes to the above literature by focusing on the gender of the top managers of private firms in developing countries and how the gender composition of the managers varies across sectors. For the sectors, we first compare the services sector as a whole with the manufacturing sector and then single out the retail sector as unique and different from other services sectors. A greater presence of female workers and female entrepreneurs in the services sector compared to manufacturing has been noted in the literature (World Bank, 2012; Dolado et al., 2004; ILO, 2012). The retail sector has also been singled out as especially important for female entrepreneurs as for example, in the Sub-Saharan Africa and Eastern Europe and Central Asia regions (Bardasi et al., 2011). Nevertheless, our focus on the gender 2 composition of top managers using cross-country comparable firm-level data across a wide spectrum of 87 developing countries is a rarity in the existing literature. Presence of female top managers is not uncommon in our sample. Projecting to the private non-agricultural part of the economies that is targeted by the survey, 18.8 percent of the firms have a female top manager. Our exploration of the data confirms that the presence of female vs. male managers is significantly higher in the service sectors relative to manufacturing. However, this difference between service and manufacturing sectors is entirely due to firms in the retail sector with little contribution from the remaining service sectors such as wholesale, construction services and the residual category of other services. In other words, for gender composition of top managers, the retail sector is special not just vis-a-vis the manufacturing sector but also the rest of the service sectors. For example, our findings show that depending on the specification used, the likelihood of having a female manager is higher for a retail firm compared with a firm in the manufacturing sector by 4.5 to 8.1 percentage points. The difference is statistically significant and also economically large given that in the full sample only 18.8 percent of the firms have a female manager. We also find strong evidence that the proportion of female managers in the retail sector vis-à-vis manufacturing or the rest of the economy is particularly high among the relatively small firms and among firms located in the relatively small cities in our sample. Hence, use of firm-level or micro data is helpful in order to explore heterogeneity across data points within a sector – a possibility that is not available using macro or industry level data. The findings from this study could be useful from the policy point of view. Given a greater presence of top female managers in the retail sector, policies aimed at improving the livelihood of females may find it attractive and optimal to expand the retail sector through 3 appropriate measures such as reducing entry barriers, providing better access to finance and improving the business climate. Furthermore, policies that benefit small firms and improve public amenities in small cities may also increase female presence in top managerial presence in retail firms. While having more female managers of firms contributes directly to the cause of gender equality and better opportunities for women, other indirect effects cannot be ruled out. For example, anecdotal evidence suggests that females in top positions tend to serve as role models for other women encouraging them to seek education and jobs; females in top managerial positions are also less likely to discriminate against other females seeking jobs (the so called “Revolving door hypothesis”). In fact, in our sample, we do find that female employment is significantly higher among firms with female managers than male mangers. 1 The broader positive effects of gender equality on the economy have been well documented in the literature (Klasen and Lamanna, 2009; Dollar and Gatti, 1999). To summarize, this study contributes to the literature in the following ways: (i) it investigates the relationship between a firm’s sector and the likelihood of having a female top manager complementing the literature on female firm ownership and female employment, (ii) uses a sample of unique cross-country firm level datasets for developing countries that follow the same sampling methodology and design allowing for cross-country comparisons, and (iii) shows that female presence in top managerial positions is much higher among retail firms when compared with the other sectors and that this difference between retail firms and the rest is magnified among the relatively smaller firms and firms located in the relatively smaller cities. The structure of the remaining sections is as follows. In section 2, we introduce the data and discuss the variables along with the estimation method used for the regression analysis. Our 1 This result holds with or without controlling for firm-size (total number of employees at the firm). 4 main regression results are provided in Section 3. In section 4, we discuss some extension of the main results. The concluding section summarizes the main findings of the paper. 2. Data and Main Variables The main data source we use consists of firm-level surveys for 87 developing countries conducted by the World Bank’s Enterprise Surveys between 2007 and 2009. Table 1 contains the list of countries in our sample along with the sample size (number of firms) for each survey. These surveys were conducted in some cases across the whole region (such as Latin America and Eastern Europe and Central Asia) and sometimes in individual countries. However, a common sampling methodology – stratified random sampling – was followed in all the surveys along with a common questionnaire. The sample for each country was stratified by firm-size, sector of activity and location within the country. Weights are provided in the survey to ensure that the sample is representative of the non-agricultural private sector of the economy. We note that Enterprise Surveys cover manufacturing as well as services sectors but certain services such as education and health are not covered. Also, the primary sector, which encompasses agriculture, mining, forestry, etc., is also excluded from the survey. Finally, the sample we use is a pure cross section of firms with one observation per firm. We also use the latest round of the survey conducted in each country. Enterprise Surveys does provide panel data with more than one observation per firm but currently, these data are limited in size and are available for only a handful of countries. The regression results discussed below are obtained from logit estimation with Huber- White robust standard errors and clustered on the country. Marginal effects from the logit estimation discussed below are calculated at the mean value of the all the explanatory variables. 5 We note that all the variables used in the regressions below are taken from Enterprise Surveys. Summary statistics of all the variables used are provided in Table 2. Correlations between the various explanatory variables used are provided in Table 3. 2.1 Dependent variable Our dependent variable is a dummy variable equal to 1 if the top manager of the firm is female and 0 otherwise (Female). In the full sample, 18.8 percent of the firms have a female top manager. The percentage figure varies between 0.16 (Yemen) and 37.1 (St. Vincent and the Grenadines). In the remainder of the paper, we will use the term female managers interchangeably with the likelihood of a firm having a female manger or the proportion of managers that are female for a specified set of firms. 2.2 Main explanatory variable Our main explanatory variables include dummy variables indicating the sector to which the firm belongs. We begin by analyzing the presence of female managers in services as a whole vs. manufacturing sector. To this end, we define a dummy variable equal to 1 if the firm belongs to the service sectors and 0 otherwise (Services). Next, we analyze how the individual service sectors compare with manufacturing and retail sectors as far as the proportion of female managers are concerned. For this purpose, we introduce a dummy variable equal to 1 if the firm belongs to the retail sector and 0 otherwise (Retail), a dummy variable equal to 1 if the firm is engaged in wholesale activity and 0 otherwise (Wholesale), a dummy variable equal to 1 if the firm belongs to the construction services sector and 0 otherwise (Construction) and a dummy variable equal to 1 if the firm belongs to any of the remaining service sectors (i.e., service sectors 6 other than retail, wholesale and construction) and 0 otherwise (Other services). In our sample, about 22.3 percent of the firms belong to the retail sector, 12.2 to the wholesale sector, 8.8 percent to the construction sector, 27.4 percent belong to the other services sector and the remaining 29.3 belong to the manufacturing sector. Our main focus in this paper is to document how the presence of female managers differs across sectors and not to explain why this is so. That is, the paper is descriptive rather than analytical, an approach dictated by data limitations rather than by choice. Nevertheless, we control for some important firm and country level characteristics which helps in two ways. First, the controls help eliminate some possible sources of omitted variable bias problem. That is, some firm or country level covariates of the likelihood of a firm having a female manager may happen to vary systematically across the various sectors listed above even though these characteristics cannot be given a sector specific interpretation. For example, it is possible that richer countries have a more developed and larger service sector; at the same time, richer countries may have better education opportunities for females and therefore more female managers relative to males. The structure of arguments here implies a positive correlation between service sector and the presence of female managers spuriously driven by differences in income levels across countries. Second, the controls help to eliminate and therefore narrow down some of the possible explanations or channels through which sector specific features may affect the presence of female managers. For example, controlling for firm-size (as we do below) implies that the sector specific differences, if any, in the proportion of female managers is not due to differences in firm-size across sectors. The search for the underlying causal factors for our main results can now focus on factors other than firm-size. 7 For country characteristics, we control for all country specific factors using country fixed effects or dummy variables indicating the country to which a firm belongs. We note that country fixed effects filter out all country specific factors that are common to all firms within a country but vary across countries such as the level of overall development, quality of institutions, socio- cultural factors affecting female employment, etc. With country fixed effects in place, the identification of our result (likelihood of a firm having a female manager) comes from differences across sectors within a country rather than from differences across countries. This is helpful in that cross-country differences are known to be particularly sensitive to the omitted variable bias problem. For firm-level variables that vary within a country, we control for basic firm characteristics, known to be important for various aspects of the firm’s structure and conduct. We also control for some proxy measures of the quality of the business environment as experienced by the firms. For example, concerns about crime and security could deter female participation in labor markets and more so than for men. If the level and severity of crime is localized or varies within a country, as it appears to be the case, the country fixed effects are not enough to eliminate crime from spuriously affecting our results. Briefly, the firm-level controls we use are as follows: the (log of) number of permanent full-time employees at the end of the fiscal year prior to the survey (Employment); (log of) age of the firm; and a dummy variable equal to 1 if the city where the firm is located is the capital city or has a population of 1 million or more and 0 otherwise (Large city), percentage of firm’s output during the last year that was exported either directly or indirectly through a third party (Exports), a dummy variable equal to 1 if the firm experienced losses due to crime during the last year and 0 otherwise (Crime), and a 8 measure of the regulatory burden at the firm level captured by the percentage of senior management’s time spent in dealing with government regulations (Time tax). 3. Estimation Our baseline regression results are provided in Table 4. These results pertain to the linear model hence excludes any interaction terms. The estimates shown in the table are the log odds ratios obtained using a logit specification with Female as the dependent variable. We focus on the results in log odds ratios rather than marginal effects; the marginal effects are discussed but only briefly. We do so to be consistent with the non-linear specification that we consider in the next section. As is well known, computing marginal effects in a non-linear model is complicated and the plethora of the estimates available can easily become unwieldy (Ai and Norton, 2003; Puhani, 2004). To begin with, we check how the likelihood of having a female manager varies between the service and manufacturing sectors without any additional controls. As shown in column 1 of table 4, moving from the manufacturing to the service sector increases the likelihood of a firm having a female manager in a statistically significant way. The log odds ratio is 0.176, significant at less than the 5 percent level. The associated marginal effect is an increase of 2.6 percentage points against the sample mean value of 18.8 percent of the dependent variable. As discussed in the introduction, we would like to check if the higher presence of female managers in the service sector is common to all the service sub-sectors or is it just due to a particular sector such as retailing. To this end, in column 2, we provide results for the proportion of female mangers in the retail sector vs. the rest of the economy. The estimated log odds ratio of Retail is positive, statistically significant at less than the 1 percent level. Specifically, a move 9 from the rest of the economy to the retail sector is associated with an increase in the estimated log odds ratio from 0.176 for Services (column 1) to 0.494 for Retail (column 2). The associated marginal effect implies a large 8.1 percentage point increase in the likelihood of a firm having a female manager when we move from the rest of the economy to the retail sector. In fact, the statistically significant effect we mentioned above for the Services dummy (column 1) disappears completely (becomes insignificant and small in magnitude) once we control for the Retail dummy (not shown). What this suggests is that much of the difference in the proportion of female managers between manufacturing and service sectors as a whole is driven by firms in the retail sector with little contribution from the remaining service sectors. We confirm this hypothesis below. We now proceed to add the dummy variables for the remaining sub-sectors within services (column 3). We note that the omitted category here is the manufacturing sector so the results shown for the various service sectors are relative to the omitted manufacturing sector. Column 3 reveals two sets of results. First, the estimated log odds ratio of Retail remains positive and statistically significant at less than the 1 percent level. This magnitude of Retail relative to the rest of the economy is almost unchanged from the estimated log odds ratio in column 2 discussed above. Second, relative to the manufacturing sector, the probability of a firm having a female manager is significantly lower in construction services, not too different (statistically) in the wholesale sector and only weakly higher (at the 10 percent level) in the other services sector. However, this weak result for the other services sector is not robust as it disappears (becomes statistically insignificant) when we controls for factors such as firm-size (discussed below). In short, the higher proportion of female managers we found above for the services as a whole relative to manufacturing appears to be entirely driven by the retail sector; the remaining service 10 sectors do not show any robust and significantly higher proportion of female managers when compared with the manufacturing sector. Below, we argue that retail is special for female managers not just when compared with manufacturing but also when compared with the remaining service sectors. The various firm-level controls mentioned above are added to the specification in column 4. Adding these controls causes the estimated log odds ratio of Retail to decrease from 0.497 in column 3 to 0.341 in column 4. The decrease is largely due to the control for firm-size. However, it is still positive, economically large and statistically significant at less than the 1 percent level. The associated marginal effect here implies an increase of 5.3 percentage points in the likelihood of a firm having a female manager when we move from manufacturing to the retail sector. This is a large effect even though it is lower than the 8.1 percentage point increase we found above in column 3. There is not much change from above in the results for the remaining service sectors except that the estimated log odds ratio of other services sector dummy is now statistically insignificant (at the 10 percent level or less) and much smaller in magnitude than what we found above (0.182 in column 3 versus 0.080 in column 4). For the firm-level controls, we find two significant results. First, the likelihood of a firm having a female manager is significantly lower among the relatively large firms (log odds ratio of -0.260) and among firms in the relatively large cities (log odds ratio of -0.192). There is also a weak (significant at the 10 percent level) positive correlation between the incidence of crime and Female but this correlation is not too robust as it disappears when we control for country specific factors (see below). Country fixed effects are added to the specification above in column 5. The estimated log odds ratio of Retail decreases only slightly from 0.341 (column 4) to 0.324 (column 5), still 11 significant at less than the 1 percent level. The associated marginal effect is an increase of 4.5 percentage points in the likelihood of a firm having a female manager when we move from manufacturing to the retail sector. This is a large effect even though it is lower than the corresponding marginal effect of 5.3 percentage point we found in column 4. As above, the construction services dummy and Employment continue to show a statistically significant negative correlation with the likelihood of having a female manager; however, none of the service sectors or the firm-level controls, including city size, shows any significant correlation with the dependent variable. We also checked for any significant difference in the proportion of female managers between retail and the remaining service sectors. We find that for all the specifications discussed above, the likelihood of having a female manager is much higher, economically and statistically (significant at less than the 1 percent level) among retail firms compared with any of the remaining service sectors (not shown), and this holds irrespective of the set of controls used above. Hence, retail is special not just vis-à-vis manufacturing but also when compared with the other services sectors. 4. Extensions In this section, we explore how the strength of the relationship between Retail and Female varies, if at all, across firms of different types. Specifically, we consider how the strength of the relationship varies between small vs. large firms (that is, with Employment) and for firms located in small vs. large cities (that is, with Large city). To this end, we interact Retail with Employment and Retail with Large city and add these interaction terms to the specifications discussed above. 12 Regression results with the interaction terms are provided in Table 5. These results are the estimated log odds ratios obtained using a logit specification. Results for the interaction term with firm-size are provided in columns (1) to (3) and for city size in columns (4) to (6); column (7) contains the results controlling for both the interaction terms simultaneously. In column 8 we add a triple interaction term between firm-size, employment and retail dummy to check if the difference in the proportion of female managers between retail and other sectors is stronger among smaller firms and more so when the firm is located in a large or small city. 2 These results show that both the interaction terms of the retail dummy with firm-size and location are negative and statistically significant at less than the 5 percent level. In other words, the estimated positive relationship between Retail and Female we found above is significantly stronger (more positive) among the relatively smaller firms (lower values of Employment) and in the relatively smaller cities (lower value of Large city). For example, the estimated log odds ratio equals a large 0.815 at the smallest value of Employment but it declines sharply by 0.58 for each standard deviation increase in the value of Employment (based on results in column 3). Similarly, based on the estimates in column 6, the estimated log odds ratio between Female and Retail equals 0.501 in the relatively smaller cities and a much lower 0.151 in the relatively larger cites. The stated heterogeneity is important not just for academic reasons but also for the appropriate design and targeting of policies aimed at improving female participation in top managerial positions. We note that the above results for the interaction term continue to hold even if we include the interaction terms for employment and large simultaneously in the specification (column 7). The same holds when we add the triple interaction term to the specification above (column 8). The log odds ratio for the triple interaction term is positive but statistically weak and insignificant at the 10 percent level or less (p-value of .109). 2 We would like to thank an anonymous referee for suggesting the triple interaction term. 13 5. Conclusion Using firm-level data for 87 developing countries and focusing on the proportion of female top managers of firms, we find that the percentage of female managers is much higher in the service sector than the manufacturing sector. However, this result is entirely due to firms in the retail sector, which has more female managers when compared with the manufacturing sector and also when compared with the remaining service sectors. We also find that the higher percentage of female managers in the retail sector vis-à-vis manufacturing is not uniform - it is much larger for the relatively smaller firms and for firms located in the relatively smaller cities. We hope that the present paper motivates future work to better understand the determinants and consequences of the gender composition of the top managers of private firms. 14 References Ai, Chunrong and Edward C. Norton (2003), “Interaction Terms in Logit and Probit Models,” Economics Letters 80(1):123-129. Bardasi, Elena, Shwetlana Sabarwal, Katherine Terrell (2011), "How do Female Entrepreneurs Perform? Evidence From Three Developing Regions," Small Business Economics 37:417-441. Becker, Gary S. (1985), “Human Capital, Effort, and the Sexual Division of Labor,” Journal of Labor Economics, 3 (Supplement): S33-S58. Bielby, D. Denise and William T. Bielby (1988), “She Works Hard for the Money: Household Responsibilities and the Allocation of Work Effort,” American Journal of Sociology, 93(5): 1031-1059. Dolado, Juan J., Florentino Felgueroso, and Juan F. Jimeno (2004), “Where Do Women Work? Analysing Patterns in Occupational Segregation by Gender.” Annals of Economics and Statistics 71/72: 293–315. Dollar, David and Roberta Gatti (1999), “Gender Inequality, Income, and Growth: Are Good Times Good for Women?” Working Paper Series No. 1, Policy Research Report on Gender and Development, World Bank, Washington DC, USA. Duflo, Esther (2005). Gender Equality in Development. BREAD Policy Paper No. 11, 2005. Hundley, G. (2000), “Male/Female Earnings Differences in Self-Employment: The Effects of Marriage, Children and the Household Division of Labor,” Industrial and Labor Relations Review, 54(1): 95-114. International Labor Organization (ILO). Global Trends For Women 2012.Geneva, Switzerland, 2012 ISBN: 978-92-2-126657-0. Klasen, Stephan and Francesca Lamanna (2009), “The Impact of Gender Inequality in Education and Employment on Economic Growth: New Evidence for A Panel of Countries,” Feminist Economics, 15(3): 91-132. Lombard, Karen V. (2001), “Female Self-Employment and Demand for Flexible, Nonstandard Work Schedules,” Economic Inquiry, 39: 214–237. Rendall, Michelle (2010), “Brain versus Brawn: The Realization of Women’s Comparative Advantage,” Working Paper No. 491, Working Paper Series, Institute for Empirical Research in Economics, University of Zurich. Puhani, Patrick A. (2012), “The Treatment Effect, the Cross Difference, and the Interaction Term In Nonlinear ‘Difference-in-Differences’ Models,” Economics Letters 115(1):85-87. Tzannatos, Zafiris (1999), "Women and Labor Market Changes in the Global Economy." World Development, 27(3):551-70. World Development Report 2012: Gender equality and development. Washington, DC: The World Bank 2011, 426 pp., ISBN: 978-0821388259. 15 Table 1: List of countries and sample size Country Number of Country Number of firms firms Afghanistan 487 Kyrgyz Republic 227 Angola 172 Latvia 243 Antigua and Barbuda 147 Lesotho 68 Argentina 995 Liberia 29 Armenia 344 Lithuania 236 Azerbaijan 286 Macedonia, FYR 301 Bahamas, The 133 Madagascar 199 Barbados 135 Malawi 136 Belarus 230 Mali 215 Belize 150 Mauritius 120 Benin 74 Mexico 1,420 Bolivia 324 Moldova 336 Bosnia and Herzegovina 300 Mongolia 345 Botswana 252 Montenegro 80 Brazil 1,697 Nepal 363 Bulgaria 249 Nicaragua 308 Burkina Faso 283 Niger 101 Cameroon 339 Panama 318 Cape Verde 85 Paraguay 342 Chad 120 Peru 992 Chile 1,006 Philippines 1,210 Colombia 929 Poland 281 Congo, Dem. Rep. 305 Romania 395 Congo, Rep. 32 Russian Federation 858 Costa Rica 524 Serbia 358 Cote d'Ivoire 144 Sierra Leone 45 Czech Republic 191 Slovak Republic 174 Dominica 150 Slovenia 257 Dominican Republic 325 Sri Lanka 546 Ecuador 358 St. Kitts and Nevis 142 El Salvador 336 St. Lucia 150 Eritrea 117 St. Vincent & the 136 Grenadines Estonia 240 Suriname 152 Gabon 46 Tajikistan 327 Georgia 275 Togo 104 Grenada 146 Trinidad and Tobago 354 Guatemala 566 Turkey 984 Guyana 158 Ukraine 676 Honduras 319 Uruguay 577 Hungary 272 Uzbekistan 355 Indonesia 1,289 Venezuela, RB 287 Jamaica 268 Vietnam 944 Kazakhstan 438 Yemen, Rep. 435 Kosovo 187 Total (all firms) 31,549 16 Table 2: Summary statistics Variable Mean Standard deviation Minimum Maximum Female 0.19 0.39 0 1 Services 0.71 0.46 0 1 Retail 0.22 0.42 0 1 Wholesale 0.12 0.33 0 1 Construction 0.09 0.28 0 1 Other services 0.27 0.45 0 1 Employment (logs) 2.76 1.16 0 11.07 Large city 0.53 0.50 0 1 Age of the firm (logs) 2.48 0.81 0 5.83 Time tax 12.49 19.07 0 100 Crime 0.25 0.43 0 1 Exports 7.44 21.98 0 100 Number of observations (firms): 31,549. 17 Table 3: Correlations between explanatory variables Services Retail Wholesale Construction Other Employment Large Age of Time Crime services (logs) city the firm tax (logs) Services 1 Retail 0.345 1 Wholesale 0.240 -0.200 1 Construction 0.200 -0.166 -0.116 1 Other services 0.396 -0.329 -0.229 -0.191 1 Employment (logs) -0.152 -0.138 -0.042 0.067 -0.038 1 Large city -0.016 -0.023 0.080 -0.013 -0.046 0.035 1 Age of the firm (logs) -0.088 -0.015 -0.025 -0.043 -0.031 0.275 -0.046 1 Time tax 0.001 -0.030 0.049 0.015 -0.016 0.061 0.066 0.023 1 Crime 0.080 0.068 -0.019 0.044 0.004 0.129 0.032 0.070 0.065 1 Exports -0.127 -0.121 -0.011 -0.079 0.041 0.181 -0.070 0.019 -0.002 -0.029 Number of observations (firms): 31,549. 18 Table 4: Base regression results (Logit specification, log odds ratios) Dependent variable: Female (1) (2) (3) (4) (5) Services 0.176** (0.080) Retail 0.494*** 0.497*** 0.341*** 0.324*** (0.081) (0.092) (0.093) (0.092) Wholesale -0.093 -0.182 -0.201 (0.149) (0.155) (0.160) Construction -0.545*** -0.581*** -0.649*** (0.153) (0.158) (0.161) Other services 0.182* 0.080 0.077 (0.103) (0.099) (0.092) Employment (logs) -0.260*** -0.288*** (0.035) (0.034) Large city -0.192** 0.079 (0.096) (0.089) Age of the firm (logs) -0.021 -0.052 (0.049) (0.048) Time tax 0.001 0.001 (0.002) (0.002) Crime 0.129* 0.082 (0.074) (0.073) Exports 0.000 -0.001 (0.002) (0.002) Country fixed effects Yes Pseudo R-squared 0.001 0.007 0.012 0.025 0.078 Observations 31,549 31,549 31,549 31,549 31549 Brackets contain standard errors that are Huber-White robust and clustered on the country. Significance level is denoted by *** (1%), ** (5 percent) and * (10%). Estimates shown are log odds ratios obtained from Logit estimation. All regressions use a constant term (not shown). 19 Table 5: Results using Interaction terms (Logit specification, log odds ratios) Dependent variable: Female (1) (2) (3) (4) (5) (6) (7) (8) Retail 0.815*** 0.781*** 0.815*** 0.707*** 0.552*** 0.501*** 0.979*** 1.281*** (0.225) (0.228) (0.212) (0.103) (0.105) (0.113) (0.205) (0.259) Employment (logs) -0.212*** -0.217*** -0.242*** -0.257*** -0.285*** -0.242*** -0.197*** (0.041) (0.044) (0.039) (0.035) (0.034) (0.039) (0.048) Large city -0.191** 0.083 0.094 0.083 0.269** 0.270** 0.503* (0.096) (0.089) (0.114) (0.117) (0.118) (0.117) (0.291) Retail*Employment -0.187*** -0.183** -0.204*** -0.201*** -0.324*** (0.072) (0.074) (0.071) (0.071) (0.093) Wholesale*Employment 0.008 0.012 -0.011 -0.011 0.203 (0.109) (0.112) (0.126) (0.130) (0.128) Construction*Employment 0.103 0.108 0.104 0.102 0.147 (0.121) (0.121) (0.131) (0.128) (0.152) Other services*Employment -0.053 -0.045 -0.031 -0.023 0.012 (0.095) (0.096) (0.087) (0.088) (0.094) Retail*Large city -0.422*** -0.410*** -0.350** -0.343** -0.915** (0.147) (0.146) (0.155) (0.157) (0.424) Wholesale -0.192 -0.198 -0.158 0.051 -0.082 -0.141 -0.095 -0.610 (0.375) (0.388) (0.427) (0.175) (0.185) (0.200) (0.374) (0.441) Construction -0.844* -0.888** -0.947** -0.660*** -0.699*** -0.764*** -1.051** -1.185** (0.432) (0.437) (0.452) (0.235) (0.245) (0.234) (0.479) (0.599) Other services 0.248 0.207 0.167 0.451*** 0.342*** 0.259** 0.329 0.260 (0.296) (0.302) (0.270) (0.117) (0.117) (0.125) (0.252) (0.293) Wholesale*Large city -0.242 -0.197 -0.127 -0.132 0.841 (0.260) (0.265) (0.279) (0.286) (0.843) Construction*Large city 0.215 0.236 0.222 0.213 0.427 (0.301) (0.312) (0.315) (0.310) (0.893) Other services*Large city -0.574*** -0.539*** -0.376* -0.378* -0.135 (0.198) (0.206) (0.204) (0.207) (0.512) Employment*Large city -0.082 (0.079) Retail*Employment*Large city 0.232 (0.145) Wholesale*Employment*Large city -0.404 (0.269) Construction*Employment*Large city -0.073 (0.266) Other services*Employment*Large city -0.107 (0.150) Age of the firm (logs) -0.020 -0.052 -0.024 -0.054 -0.054 -0.053 (0.049) (0.047) (0.048) (0.047) (0.047) (0.046) Time tax 0.001 0.001 0.001 0.001 0.001 0.001 (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Crime 0.129* 0.082 0.133* 0.083 0.083 0.087 (0.075) (0.074) (0.074) (0.073) (0.074) (0.073) Exports -0.000 -0.001 0.000 -0.001 -0.001 -0.001 (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Country fixed effects Yes Yes Yes Yes Pseudo R-squared 0.245 0.026 0.079 0.016 0.028 0.08 0.08 0.082 See the note at the bottom of Table 4. Sample size for all the regressions is 31,549. 20