Policy Research Working Paper 11198 International Activity and Female Labor Participation New Evidence for Services Firms in Developing Countries Luis Aguilar Luna Deborah Winkler Economic Policy Global Department September 2025 Policy Research Working Paper 11198 Abstract Using a cross-section of more than 33,000 services firms COVID-19, and the gap for global value chain participants in 104 low- and middle-income countries from the World is no longer significant after COVID-19. Controlling for Bank’s Enterprise Surveys, this paper examines whether sectoral relative wages between men and women does not the female labor share premium of international firms change the findings in a smaller subsample of economies. relative to non-international firms in manufacturing also Controlling for female top management and ownership holds for services firms. Unlike the manufacturing sector, reveals a female labor share gap for exporters, global value the paper finds a negative relationship between exporting chain participants, and importers. Using an alternative esti- and global value chain participation and the female labor mator and data set confirms the female labor share gap in share for services firms, while no relationship is found for services firms. This may be attributed to the sectoral segre- importing or foreign ownership status, controlling for firm gation between women and men, with women tending to output, productivity, technology intensity, and fixed effects. pursue work opportunities in less skill- and export-intensive The female labor share gap for exporters was larger before services sectors compared to men. This paper is a product of the Economic Policy Global Department. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at dwinkler2@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 International Activity and Female Labor Participation: New Evidence for Services Firms in Developing Countries 1 Luis Aguilar Luna 2 and Deborah Winkler 3 Authorized for distribution by Sébastien Dessus, Practice Manager, Economic Polity Global Department, World Bank Group Keywords: Services firms, trade, global value chains, women, female labor share, labor force participation. JEL codes: F1, F2, F6 1 This paper builds on Rocha and Winkler (2019) and Rocha, Shepherd and Winkler (2020) who found that trading firms in manufacturing have a female labor share premium. The authors thank Sébastien Dessus, Roberto Echandi, Anirudh Shingal, Pierre Sauvé and Ben Shepherd for their valuable comments on the services firm assessment, as well as Ana Fernandes, Caroline Freund, Claire Hollweg and other participants of the “Trade and Gender Authors’ Workshop” at the World Bank on March 7, 2019, and the participants of the “Conference on Trade and Gender” at the World Trade Organization on December 6-7, 2018, for their helpful suggestions on the manufacturing firm analysis. The authors also thank Marlon Amorim for creating the Trade, Women and Welfare dashboard and Ankriti Singh and Maria Reyes Retana Torre for verifying the reproducibility of the analysis. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the World Bank Group and its affiliated organizations, its Executive Directors or the governments they represent. 2 Consultant, Trade Unit, Economic Policy, The World Bank, Washington, DC 20433. 3 Senior Economist, Trade Unit, Economic Policy, The World Bank, Washington, DC 20433. Email: dwinkler2@worldbank.org. 1. Motivation Several studies have linked trade participation in developing economies to higher female employment and wages (World Bank-World Trade Organization 2020). Empirical evidence indicates a positive correlation between export growth in developing countries during the 1960s and mid- 1980s and increased female labor force participation (e.g., Wood 1991). The Heckscher-Ohlin model suggests that trade-induced sectoral reallocation can impact female workers. In developing countries with abundant low-skill labor, trade may increase female labor force participation because women are often concentrated in low-value-added, labor-intensive export sectors like apparel, where these countries have a comparative advantage. This sectoral reallocation mechanism is also emphasized in other studies like Aguayo-Tellez et al. (2013) and Busse et al. (2006). Focusing on manufacturing firms across a wide sample of low- and middle-income countries (LMICs), Shepherd (2018), Rocha and Winkler (2019), and Rocha, Shepherd and Winkler (2020) assess whether firms that engage in trade and foreign direct investment (FDI) have a higher proportion of female employees compared to non-international firms. Their findings confirm a female labor share premium for exporters, importers, firms participating in global value chains (GVCs), and foreign firms. These results remain consistent even after controlling for factors such as firm output, capital intensity, productivity, female wage gaps, female ownership and management, and fixed effects. In this paper, we build on this earlier work but focus on services firms instead – an area which remains underresearched – using a cross-section of more than 33,000 services firms in 104 low- and middle-income countries (LMICs) from the World Bank’s Enterprise Surveys. 4 Cross-country evidence also suggests that even across developing economies the relationship between trade and female labor force participation varies substantially (Heath et al. 2024, Wacker et al. 2017, Tejani and Milberg 2016, Aguayo-Tellez 2012), which can depend on several factors. Besides sector reallocation, other channels through which trade participation affects female labor participation include trade-induced increases in competition in export markets, which can foster technological upgrading within firms, as well as changes in firms’ discriminatory behavior (Heath and others 2024, World Bank-World Trade Organization 2020). These channels can work in favor of 4 One caveat of using the Enterprise Surveys is that they classify firms into manufacturing or services based on the ISIC Rev. 4 classification of the firm’s main product or service. As a result, firms can be ‘mixed’ producers, so it is not clearly identifiable if exports are aligned with the main activity of the firm. 1 or against women, depending on several factors including the relative magnitude of channels in the country, the country’s comparative advantage, and sector and workforce characteristics. Studies also show that trade liberalization can reduce the female wage gap and raise female labor demand by reducing discrimination, as seen in Uruguay (Ben Yahmed 2017) and Colombia (Ederington et al. 2009). Oostendorp (2009) found that greater economic development and trade are associated with lower female wage discrimination across a large sample of countries. Junh et al. (2013, 2014) demonstrated that firms entering export markets due to NAFTA, and those which also imported computerized equipment, experienced increased female employment and wages relative to men in blue-collar jobs. In contrast, a study on Norway using matched employer-employee data, for instance, found that the female wage gap is higher in exporting compared to non-exporting firms, but only for higher-skilled jobs. Explanations include that exporting firms demand highly flexible employees (for working peculiar hours, taking late night phone calls and engaging in international travel arranged at short notice) and therefore discriminate against women as they are perceived to be less flexible (Bøler et al. 2018). In this paper, we focus on the relationship between firms’ international status and the female labor share for services firms. Only few studies have emphasized the role of trade in services for female labor participation. Wacker et al. (2017) find in a panel of 80 developing countries over the period 1980-2005 that the effect of services trade on the female labor force participation rate is negative and significant, although small. The authors explain this finding with the fact that tradable services may require higher skills, leading young women to pursue education and delay entering the labor market. These findings align with Bussmann (2009) who found that globalization correlates with fewer women employed in services in non-OECD countries. This paper contributes to the research on trade and labor in several ways: (1) We adapt the cost share functions of Feenstra and Hanson (1996) to examine trade’s impact on female versus male workers, rather than high- versus low-skill workers. (2) We use a comprehensive, harmonized micro- level dataset covering numerous developing and emerging economies, unlike many studies that rely on aggregate data. 5 (3) We focus specifically on services firms across all service sectors, an area with limited research, and compare findings with manufacturing firms. (4) We examine the role of female entrepreneurship, particularly ownership and management, in the female labor share within firms. 6 5 Exceptions include Ederington et al. (2009) for Colombia, Chen et al. (2012) for China, and Juhn et al. (2013, 2014) for Mexico. 6 For an overview of female entrepreneurship and a discussion of different drivers, see, e.g., Carranza et al. (2018). 2 (5) We use an alternative dataset which confirms the broad findings of a female labor share gap in services sectors. Unlike the manufacturing sector, our study finds a negative relationship between exporting and GVC participation and the female labor share for services firms, while no relationship is found for importing or foreign ownership status, controlling for firm output, productivity, technology intensity, and fixed effects. The female labor share gap for exporters is larger before Covid-19, while the gap for GVC participants is no longer significant after Covid-19. Controlling for female top management and ownership reveals a female labor share gap for exporters, GVC participants, and importers. Controlling for sectoral relative wages between men and women does not change the overall findings in a smaller sub-sample of economies for which relative wage data are available. The findings also indicate that higher relative wages between male and female workers – or a higher female wage gap – are associated with a lower female wage share which is in line with expectations. Using an alternative estimator and dataset confirms the female labor share gap in services firms. The female labor share gap for exporters and GVC participants in services sectors may be attributed to the sectoral segregation between women and men, with women tending to pursue work opportunities in less export-intensive and skill-intensive sectors (such as retail and hospitality), and men being employed in more export-intensive sectors (such as transportation) and skill-intensive services sectors (such as information and communication technologies). Given that global firms in modern services sectors are more likely to hire more skill-intensive or specialized labor, the results would reflect the skill disadvantage that women have relative to men which precludes their employment in such jobs. Since skill-intensive modern services jobs are likely to pay more, this also explains the negative link between the female wage gap and female wage share. In contrast, in manufacturing, women tend to have the “right” skill-set for jobs in GVC-intensive sectors like textiles or computers and electronics, as confirmed by our findings covering manufacturing firms. Our paper is organized as follows. Section 2 outlines the underlying model and derived descriptive regressions, as well as the data and measurements. Section 3 presents stylzed facts of female labor force participation, focusing on variations across international status, sectors, and countries. Section 4 displays the descriptive regression results for the female labor share. Updated results for manufacturing firms are provided in the Appendix to compare with the services firms’ results. Robustness checks, including sample splitting by time period, additional control variables, an alternative fractional probit estimator, and a different dataset, are also performed. Section 5 concludes the paper. 3 2. Descriptive Model, Data, and Measures 2.1 Model A variable unit cost function CV is specified as follows 7: CV=CV(Y, wF, wM, k, T) where Y denotes the output and wF and wM are the exogenous wages for the variable input factors female labor LF and male labor LM. Capital is considered a quasi-fixed input factor in the form of capital intensity k. The technology shifter T=T(prod, int) is defined as a function of productivity and international activity. Using the transcendental logarithmic (translog) form of the variable unit cost function as introduced by Brown and Christensen (1981) and applying Shephard’s Lemma, 8 the following factor demand function can be derived: = + + � ⁄ � + + + SF is the cost share of LF in variable costs CV. Since wF and wM are the only variable costs, CV is determined by the sum of the products of the variable factor costs with their respective factors, CV = wFLF + wMLM = wL, where w designates the average wage per labor input L, regardless of the sex. We refer to the composite term SF = wFLF / wL as the female wage share. A decrease of SF can reflect both a fall in LF and/or a fall in wF, which for non-declining CV, implies a rise in SMand thus an increase in LM and/or in wM. The control variables include output, Y. The coefficient sign of output Y, ß1, is not unambiguously predictable. An increase in Y can be expected to lead to a higher total wage bill. If the cost increase is equally distributed between female and male labor, there should be no influence on SF. If the wage bill of male labor increases more than proportionally, e.g., due to better bargaining power, this results in a higher LM and/or wM, and SF is expected to fall (ß 1 < 0). One expects a lower SF and thus a lower cost share of LF in total wages when relative wages for male labor, wM/wF, as part of CV rise (ß 2 < 0). An increase in the capital intensity, k, will increase SF, if 7 This is a modified version of the unit cost functions by Feenstra and Hanson (1996), Geishecker (2006) and others that differentiate between skilled and unskilled labor. 8 According to Shephard’s Lemma (1953), factor demand is determined by the first partial derivative of the cost function with respect to the corresponding factor price, regardless of the kind of production function. 4 capital is a substitute for male labor (ß 3 > 0), but reduce SF, if capital is a substitute for female labor (ß 3 < 0). A higher productivity, prod, could increase labor demand in favor of female (γ2 > 0) or male workers (γ2 < 0). The influence of int on SHS is not easily predictable, as international activity could increase the relative demand for female (γ1 > 0) or male labor (γ1< 0). 2.2 Descriptive Regressions In the absence of sex-specific wage data at the firm-level, the model is modified in two ways to obtain the following desriptive regressions. First, the analysis on international activity and the female wage share combines the Enterprise Survey dataset with sectoral wage data by sex, available from Household Surveys. This assumes equal relative wages between women and men across all firms within a given sector in a specific country. While this is a strong assumption, one could argue that wage-setting often occurs at the sectoral level where sectoral trade unions and employer associations negotiate collective bargaining agreements that cover most firms in that sector. This allows to estimate the following female wage share equation: = + + � ⁄ � + + + + + + + (1) Subscripts i, c, s, r, and t refer to firm, country, sector, sub-national region, and survey year, while Dcs, Dr, Dt denote country-sector, sub-national region and time fixed effects, and the idiosyncratic error term. We allow for heteroskedasticity depending on the full set of explanatory variables (other than fixed effects), and correct all standard errors for clustering by country-sector. International activity, int, is a binary variable measuring a firm’s participating status. Second, since sex-disaggregated sectoral wage data for services are only available for 31 countries in the Enterprise Surveys, and only 17 of those countries have available data for all the explanatory variables, the baseline analysis assumes equal wage rates for female and male workers, wF = wM = w. Since SF = wFLF / wL, the dependent variable becomes the female labor share in total employment, femsh= LF / L. Similarly, the model no longer controls for relative wages, since we assume ln(wM/wF) = ln1 = 0. This assumption allows us to include over 100 countries in the services sample. The female labor share equation takes the following form: = + + + + + + + + (2) 5 While assuming equal wage rates between female and male workers is a strong assumption, we address possible concerns as follows: (i) We argue that the fixed country-sector effects included in the model partially correct for wage differences. (ii) While equation (2) becomes our baseline specification, allowing us to include all available countries in the analysis, we perform a robustness check based on equation (1) to confirm if key findings continue to hold when controlling for female wage differences. (iii) We also run the descriptive regressions and robustness checks for manufacturing firms – which are based on the same database and assumptions – to identify if the results for services firms are a function of sectoral characteristics rather than specification. Both descriptive regressions consider four types of international firms: (1) Exporters are firms with an export share (direct or indirect) of at least 10% of total sales. (2) Importers are firms with an imported input share of at least 10% of total inputs. (3) GVC participants are firms that are classified as both exporters and importers. (4) FDI refers to firms with a foreign ownership share of at least 10%. The international variables are dummies taking a value of 1 if the firm trades, and 0 if it does not. 2.3 Data Our dataset draws on two underlying datasets published by the World Bank Enterprise Analysis Unit, namely the Enterprise Surveys Global Database and the Firm-level TFP Estimates and Factor Ratios. The Enterprise Surveys Global Database covers 242 surveys in 160 countries over the period 2006 to 2024. The Enterprise Surveys represent a comprehensive source of firm-level data in emerging markets and developing economies. One major advantage of the Enterprise Surveys is that the survey questions are the same across all countries. Moreover, the Enterprise Surveys represent a stratified random sample of firms using three levels of stratification: sector, firm size, and region. Sectors are determined based on the ISIC Rev. 4 classification of the main product or service. One caveat of using the Enterprise Surveys is therefore that firms can be ‘mixed’ and engage in both services and manufacturing activities, so it is not clearly identifiable if exports are aligned with the main activity of the firm. The Enterprise Surveys Global Database covers a wide range of indicators on firm characteristics, the business environment, innovation and technology, and workforce and skills among others. We merged this dataset with data on firm-level output, value added, and capital stock obtained from the Firm-level Total Factor Productivity (TFP) Estimates and Factor Ratios dataset. It is important to note 6 that value added, capital stock, and TFP data are only available for manufacturing firms (whose results are shown in the Appendix) but not for services firms in the Enterprise Surveys. In our descriptive regressions, we therefore use technology intensity and labor productivity as alternative control variables, respectively, which are described in section 2.4. All local currencies have been converted into US dollars and deflated using a GDP deflator in USD (base year 2009). Exchange rates and GDP deflators have been obtained from the World Development Indicators. We aggregated ISIC Rev. 4 sectors up to TiVA 2018 sectors, as the number of services firms in several cases was very spotty, to allow for greater firm coverage per sector. We apply the following rules to create the services firm dataset: (i) We include only the most recent Enterprise Surveys for each country (excluding EU15 countries and Singapore to focus on emerging markets and developing economies). (ii) We cover the years 2010 to 2024. (iii) We exclude manufacturing firms from the sample, i.e. for which the ISIC Rev. 4 classification of the main product or service is lower than 45 and those that have been classified as manufacturing firm by the Enterprise Surveys. (iv) We drop countries for which female employment data were spotty and firms with missing observations for female employment or total employment. (v) We exclude countries with fewer than 100 services firms after applying the previous steps. (vi) Finally, we exclude TiVA 2018 sectors with only one firm (D64T66, D68, D84, and D86T88). 9 We applied similar rules as described in Rocha and Winkler (2019) to obtain a separate dataset for manufacturing firms which allows for comparisons with the services firm sample. The procedure above results in more than 33,000 services firms in 104 countries and 34,000 manufacturing firms in 81 countries. The list of countries, year of the most recent survey and number of firms by country and broad sector can be found in Appendix 1. The distribution of firms across TiVA 2018 sectors is shown in Appendix 2. We combine the firm-level data with Global Employment Sex-disaggregated Statistics (GESS) at the sector level from the World Bank Household Surveys and other public resources. GESS uses harmonized classifications of economic activities and occupation categories. It fills an important information gap in global sex statistics by providing more detailed accounts on education, employment levels, wages, labor income, and employment status at very disaggregated economic activity levels and occupation categories than is usually available. The data are available for 53 9 Steps (v) and (vi) remove 2.4% of services firms. 7 countries at the ISIC Rev. 4 level, but only 31 countries overlap with the Enterprise Survey data sample. These 31 countries are listed under the country list of Appendix 1. 2.4 Measures This study focuses on two measures of female labor participation: • Female labor share, fem_sh = number of permanent full-time female employees as % of total number of permanent full-time employees. 10 • Female wage share, SF = total compensation of female employees as % of total compensation of all employees. Since the Enterprise Survey dataset has only information on average wage rates, 11 we approximate female wage rates at the firm level by assuming the same ratio of female-to-average wages in a given sector as in the sex-disaggregated wage data from the World Bank Household Surveys. Total female compensation is then obtained by multiplying the female wage rate with the number of female employees from the Enterprise Surveys. The analysis examines two measures that describe trading firms: • Exporter, exp = 1 if direct plus indirect export share as % of sales >= 10%, and 0 otherwise. 12 • Importer, imp = 1 if share of imported inputs as % of total inputs >= 10%, and 0 otherwise. We additionally include two measures that describe global firms: • GVC participant, gvc = 1 if both exp = 1 and imp = 1, and 0 otherwise. • Foreign ownership, fdi = 1 if foreign private ownership >= 10% and 0 otherwise. Finally, we include the following control variables in all regressions: • Output, Y = total sales (in natural logarithms). • Labor productivity, lp, is measured as output per worker (in natural logarithms). • Technology intensity, tech, is captured as tech = iso + tech_for ∈ {0, 1, 2}, where iso = 1 if firm owns internationally-recognized quality certification and 0 otherwise, and tech_for = 1 if firm uses technology licensed from foreign firms and 0 otherwise. 10 For manufacturing firms, the variable can be computed for permanent production workers, non-production workers and total workers (sum of both). Non-production workers include managers, administration and sales. This is not feasible for services firms. 11 These are obtained by dividing a firm‘s total compensation by its total number of employees. 12 As stated above, firms in the Enterprise Surveys can engage in both services and manufacturing activities, so it is not clearly identifiable if exports are aligned with the main activity of the firm. 8 Due to data constraints for services firms, we are not able to use TFP or capital intensity as was done for the manufacturing sample in Rocha and Winkler (2019). In addition, their measure of labor productivity for manufacturing firms used value added as numerator which is unavailable for services firms in the Firm-level Total Factor Productivity (TFP) Estimates and Factor Ratios dataset. Appendix 1 shows the countries included in the services and manufacturing samples, as well as their survey year. Appendix 2 shows the sector coverage at the TiVA 2018 sector level for both the services and manufacturing sample separately. Similarly, Appendix 3 shows the summary statistics for both samples separately, while Appendix 4 shows updated results for manufacturing firms following the specification suggested by Rocha and Winkler (2019). 3. Stylized Facts 3.1 Differences by International Status To gain an initial understanding of the relationship between the female labor share and international firms, this section first outlines the average share of female workers in a firm’s total workforce across various types of international firms. At this stage, no other factors that could influence the female worker share, as discussed in section 2.1, are controlled for. Figure 1 (left panel) indicates that international manufacturing firms have a higher average share of female permanent workers compared to non-international firms, corroborating the findings of Rocha and Winkler (2019). Appendix Figure 4.1 presents results separately for production and non-production workers, demonstrating that the higher female labor share in international manufacturing firms is largely driven by production workers. Figure 1: Female labor share, manufacturing vs. services, by international status Manufacturing Services exporting 43.2% exporting 33.5% 28.8% 37.7% importing 41.5% importing 40.8% 30.4% yes 40.8% yes 46.9% no 29.4% no gvc-participant 30.0% gvc-participant 37.6% fdi firm 50.7% 41.4% 32.4% fdi firm 35.8% Note: The female labor shares are averages using firm employment as weights. 9 However, the patterns are less straightforward for services firms. Specifically, the proportion of women in the average permanent workforce of services firms is higher only for importers (40.8%) and FDI firms (41.4%) compared to non-importing (31.2%) and domestically-owned firms (35.8%), respectively. Conversely, services firms that export (33.5%) or participate in GVCs (29.4%) have a lower share of women in their permanent workforce compared to non-exporters and non- participants, respectively. To better understand the reasons behind the observed differences between services and manufacturing firms, section 3.2 delves into the trends at the disaggregated sub-sector level. 3.2 Sectoral Differences This section describes the average female labor share across various services sectors. The average share of female permanent workers in all services sectors is 36.8%, compared to 36.6% in manufacturing. The proportion of female permanent workers surpasses 40% in wholesale and retail trade, as well as in accommodation and food services, which are recognized as female-intensive sectors (Figure 2). These are followed by other business services sectors (35.8%), telecommunications (33.9%), and publishing, audiovisual, and broadcasting activities (31%). Notably, the average female labor share in most services sectors falls between 28% and 40%. The transportation and storage sector stands out with a significantly lower female labor share of 19.1%. Figure 2: Female labor share in services firms, by sub-sector Wholesale and retail trade; repair of motor vehicles 40.4% Accommodation and food services 40.3% Other business sector services 35.8% Telecommunications 33.9% Publishing, audiovisual and broadcasting activities 31.0% Arts, entertainment, recreation and other service… 29.1% IT and other information services 28.3% Transportation and storage 19.1% Note: For the sectoral distribution of firms, see Appendix 2. For female labor shares in manufacturing sub-sectors, see Appendix Figure 4,2. The female labor shares are averages using firm employment as weights. This is in contrast to manufacturing firms, which exhibit a much wider variation in female labor share across different sectors. Textiles, apparel, and leather (57.9%) and computer and electronics 10 (42.5%) are highly female-intensive. On the other hand, several other sectors, such as coke and petroleum, mineral and metal products, as well as motor vehicles and machinery and equipment, have female labor shares below 25% (Appendix Figure 4.2). The previous figure does not distinguish between international and non-international firms. Figure 3 (top panel) therefore displays the average female labor share of permanent workers for exporters on the x-axis and non-exporters on the y-axis by sector. Many sectors are positioned around or above the dotted y = x line, indicating lower female labor shares for exporting firms compared to non-exporting firms. The exceptions are arts, entertainment, recreation, and other service activities (D90T96) and telecommunications (D61), where exporters are more female- intensive. This sharply contrasts with manufacturing firms, where exporters tend to have a higher female labor share than non-exporters (Appendix Figure 4.3, top panel). Figure 3: Female labor share, trading vs. non-trading firms, by services sub-sector Exporting vs non-exporting firms Female labor share (%), non- 60% 50% D45T47 exporting 40% D55T56 D69T82 D61 30% D62T63D58T60 D90T96 20% D49T53 10% 0% 0% 10% 20% 30% 40% 50% 60% Female labor share (%), exporting Importing vs non-importing firms Female labor share (%), non- 60% 50% 40% importing D69T82 D45T47 D55T56 30% D58T60 D62T63 D61 20% D90T96 D49T53 10% 0% 0% 10% 20% 30% 40% 50% 60% Female labor share (%), importing Source: The female labor shares are averages using firm employment as weights. Sector numbers relate to TiVA 2018 sectors. For sector names and the sectoral distribution of firms, see Appendix 2. 11 In contrast, Figure 3 (bottom panel) indicates that importing firms generally have a higher female labor share compared to non-importers across most services sectors—a pattern also observed in manufacturing firms (Appendix Figure 4.3, bottom panel). Similar to exporters, the female labor share premium is notably high in entertainment, recreation, and other service activities (D90T96), telecommunications (D61), as well as accommodation and food services (D55T56). GVC participation exhibits mixed patterns across services sectors (Figure 4, top panel). Several services sectors display higher female labor shares for firms that both import inputs and export compared to firms that do not participate in GVCs, especially in entertainment, recreation, and other service activities (D90T96) and telecommunications (D61). In contrast, GVC-participating firms show lower female labor shares in some sectors, particularly wholesale and retail trade and repair of motor vehicles (D45T47). This pattern contrasts sharply with the manufacturing sample, where GVC firms demonstrate a female labor share premium (Appendix Figure 4.4, top panel). Figure 4: Female labor share, global vs. non-global firms, by services sub-sector GVC participants vs non-participants Female labor share (%), non- 50% 40% D45T47 D55T56 participant D69T82 D61 30% D58T60 D62T63 D90T96 20% D49T53 10% 0% 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% Female labor share (%), GVC participant FDI vs non-FDI firms 60% Female labor share (%), non-FDI 50% 40% D45T47 D61 D69T82 D55T56 30% D90T96 D58T60 D62T63 20% D49T53 10% 0% 0% 10% 20% 30% 40% 50% 60% Female labor share (%), FDI Source: The female labor shares are averages using firm employment as weights. Sector numbers relate to ISIC Rev. 3 sectors. For sector names and the sectoral distribution of firms, see Appendix 2. 12 Finally, focusing on the role of foreign ownership, the scatterplot in Figure 4 (bottom panel) suggests that FDI firms generally have a higher female labor share compared to domestic firms across nearly all services sectors. This pattern is also confirmed for manufacturing sectors (Appendix Figure 4.4, bottom panel). 3.3 Differences across Countries This section shifts the focus to the average female labor share of services firms by country. The number of surveyed services firms varies significantly across the country sample, ranging from over 3,400 firms in India, to over 1,000 firms in Indonesia and Nigeria, to around 100 in Guinea, Niger, and Sierra Leone, among others. Therefore, the findings of the country analysis should be interpreted with caution. However, they provide an initial insight into female labor participation across different countries. The main observation from Figure 5 is that the female labor share ranges from over 58% in Eastern European and Central Asian countries, particularly Serbia and Belarus, to less than 10% in some Middle Eastern and South Asian economies, such as the Republic of Yemen, Iraq, Pakistan, Saudi Arabia, and Bangladesh. Figure 5: Female labor share in services firms, by country 72.3% 69.1% 62.5% 59.1% 58.3% 57.3% 56.7% 56.6% 55.8% 55.5% 55.3% 53.8% 53.5% 52.9% 52.5% 52.1% 50.9% 50.9% 50.2% 49.9% 49.3% 49.2% 48.5% 48.1% 47.4% 46.8% 46.0% 45.3% 45.0% 44.8% 44.5% 43.7% 43.6% 43.5% 43.4% 42.9% 42.9% 42.6% 41.5% 41.1% 41.0% 40.9% 40.4% 39.4% 39.2% 38.5% 38.4% 38.1% 37.9% 37.6% 37.5% 37.5% 37.5% 37.0% 36.9% 36.0% 35.7% 35.3% 35.2% 34.7% 34.1% 33.7% 33.6% 33.5% 33.3% 32.5% 32.2% 32.2% 31.2% 30.9% 30.4% 30.4% 30.2% 29.2% 29.1% 29.0% 28.2% 27.4% 27.0% 26.8% 26.5% 26.2% 26.2% 25.0% 24.3% 23.4% 22.2% 21.6% 20.7% 20.5% 19.8% 18.6% 18.4% 18.2% 17.1% 15.8% 14.4% 12.2% 9.9% 9.6% 8.4% 6.7% 4.5% 1.8% bih dom jor ner pse hrv lva khm rus mkd cze tto dma phl ecu uzb sau atg est kgz mar mex civ tls pry ssd yem irq mng egy cod cog bwa hnd slv cmr moz srb idn kaz col ago zaf sur gtm mmr aze svk rou svn pan tza Note: See Appendix 1 for full country names. The female labor shares are averages using firm employment as weights. Since the previous analysis does not distinguish between international and non-international firms, Figure 6 (top panel) plots the average female labor share of permanent workers for exporting services firms on the x-axis and non-exporting firms on the y-axis by country. More countries are located below the dotted y = x line, indicating higher female labor shares for exporting firms compared to non-exporting firms. Notable exceptions include East Asian and Pacific countries such 13 as Cambodia, the Lao People’s Democratic Reublic, and the Philippines, as well as North African countries like Morocco and Tunisia. In Eastern European and Central Asian economies, female labor shares in services firms tend to be higher for non-exporters. Regarding importer status, more countries fall below the dotted y = x line, suggesting higher female labor shares for importers compared to non-importing firms (Figure 6, bottom panel). Figure 6: Female labor share in services firms, trade participants vs. non-participants, by country Exporting vs non-exporting firms 80% srb Female labor share (%), non-exporting 70% blr czeest svk hrv 60% ltu lva geo zwe bih arm pol ukr grd bgr hun jam mng atg 50% rus mkd chl bwa khm rou hkg tjk chn kaz cyp vnm svn ury nam idn mus kgz tto tza ven col rwa 40% alb zaf hnd mex nzl pan mys civ bol ago mdg nic pry tur tls arg gtm sur lao ken zmbslv dma per aze npl gha mmr uga 30% ecu jor cmr cri lbn mwi ssdlka dom tun phl egy cog uzb nga 20% eth sen cod mozdji sle mar ner sdn ind pse 10% irq yemgin sau pak 0% bgd 0% 10% 20% 30% 40% 50% 60% 70% 80% Female labor share (%), exporting Importing vs non-importing firms 80% Female labor share (%), non-importing 70% blr ltu srb 60% est hkg arm 50% nzl bwa pol ukr mng bol rus geo hrv tza hndbgr idn hun tjk mus lva 40% mdg cyp col mkd mex zaf lao rwa vnm svk tur cri nic civmys pry alb sur rou cze ken arg tlsmmr svn zwe 30% sle phl gha cmr slv per kaz ury aze tun kgz uzb dom zmb mar 20% cogmozlbn egy jor gtm bih npl khm ind 10% pse ner ecu sau pak gin irq 0% bgd nga 0% 10% 20% 30% 40% 50% 60% 70% 80% Female labor share (%), importing Source: The female labor shares are averages using firm employment as weights. See Appendix 1 for full country names. 14 Shifting the focus to cross-country patterns of global services firms reveals a similar trend for GVC participation as seen with exporting status (Figure 7). Although there is no distinct pattern for foreign ownership status, in several cases, domestic firms have higher female labor shares compared to foreign firms. This is observed in East Asian countries like Myanmar and Cambodia, as well as Eastern European countries such as Serbia and the Slovak Republic. Figure 7: Female labor share in services firms, global vs. non-global firms, by country GVC participant vs non-participants 90% Female labor share (%), non-participants 80% srb blr 70% svk hrv 60% est lva geo zwe ltu arm mng rus mkd hun ukr pol 50% bwa hkg bih khm kaz bgrrou svn ury mus kgzcze idn tza cyp rwa vnm 40% mex tjk mdgalb zaf hnd nzl col nicpry mys civ tls tur bol sur lao gtm azemmr cri 30% ecuzmb arg ken jor slv phl cmr per gha tun npl nga lbn uzb dom 20% egy cog moz sle mar ner ind pse 10% irq sau pak gin 0% bgd 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% Female labor share (%), GVC participants FDI vs. non-FDI firms 100% Female labor share (%), non-FDI 90% mmr srb 80% cze khm svk 70% blr zwe arm atg bih grd hrv 60% ury hun jam hkgltu pol 50% mar arg alb rwa tto mkdbwa geo jor rou cyp est lva phl per tun mdg chn kaz vnmnam mus chl 40% dom col kgz pan svn mng ner egy lbn ecu slv cri civ dma uga mys bol lao zaf sur hnd tjkago tza ukr 30% ngalka uzb ssd nic tls mex idn sdn ghanpl ken tur bgr rus cog ind sledji mwi cmr zmb aze pry gtm vennzl 20% cod pse sen moz 10% pak gin eth yem 0% bgd sau 0% 10% 20% 30% 40% 50% 60% 70% 80% Female labor share (%), FDI Source: The female labor shares are averages using firm employment as weights. See Appendix 1 for full country names. In summary, comparing the results for services firms with those in manufacturing reveals that international firms in manufacturing generally exhibit higher female labor shares across countries 15 relative to the rest of the sample (Appendix Figures 4.6 and 4.7). This is particularly true for exporting and GVC participation status, and to a lesser extent for importing and foreign ownership status. In contrast, the results for services firms are more mixed, with no clear trends in the female labor share, except for importing status, which tends to benefit the female labor share. Interestingly, countries in East Asia and Eastern Europe show a female labor share premium for global manufacturing firms but a male labor share premium for global services firms (Appendix Figure 4.7). 4. Descriptive Regression Results 4.1 Baseline Results This section presents the results of the regression analysis based on the specification of equation (1). Table 1 displays results on mean differences in female labor share between trading and non- trading firms, while Table 2 focuses on global versus non-global firms. Regarding the control variables, labor productivity is negatively associated with the female labor share, suggesting that women are more likely to be employed in less-productive services firms. The coefficient for technology intensity is also negative, indicating a higher female labor share in less technology- intensive services firms. Both factors might be influenced by the lower average skills of female workers. Output is negatively correlated with the female labor share, suggesting that smaller services firms have higher female labor shares, while it loses statistical signifiance once productivity differences across firms are accounted for. These findings are consistent with those for the manufacturing sample, although the relationship between higher female labor share and smaller firm size is observed for manufacturing firms in the exporter-status regressions even when controlling for productivity differences (Appendix Table 5.1). The results in Table 1 indicate that exporting is negatively correlated with the female labor share, which is statistically significant after controlling for firm size and technology intensity (column 3) and when also including productivity (column 4). The results in column 4 suggest that the female labor share is 3.1 percentage points lower for exporting firms compared to non-exporting firms, after accounting for firm-level factors and fixed effects. This gap is smaller than the 4.2 percentage points difference shown in Figure 1, which can be attributed to the inclusion of control variables and fixed country-sector, subnational region, and year effects. These results contrast sharply with those for manufacturing firms, where exporters have a female labor share premium of 4.6 percentage points, after controlling for firm-level factors and fixed effects (Appendix Table 5.1). 16 Table 1: Female labor share difference of trading vs. non-trading services firms, OLS Dependent Exporter, exp Importer, imp variable: (1) (2) (3) (4) (5) (6) (7) (8) fem_shisrt lnYisrt -0.109 -0.156 0.222 -0.187** -0.212* 0.174 (0.158) (0.205) (0.254) (0.037) (0.099) (0.387) lntechisrt -1.187*** -1.357*** -1.880*** -2.053*** (0.007) (0.002) (0.000) (0.000) lnlpisrt -0.637** -0.661** (0.012) (0.012) globalisrt -0.502 -0.635 -3.112*** -3.161*** 0.564 0.594 0.588 0.597 (0.248) (0.170) (0.000) (0.000) (0.125) (0.125) (0.241) (0.233) Obs. 33105 29509 14591 14591 25383 23415 13792 13792 R2 0.36 0.36 0.32 0.32 0.37 0.38 0.31 0.31 Note: P-values are in parentheses. p*<0.1, p**<0.05, p***<0.01 (robust standard errors corrected for clustering by country- sector in parentheses). All regressions include country-sector, subnational region and year fixed effects. Regarding importing firms, no significant differences are found between importers and non- importers for services firms (Table 1, columns 5 to 8), while there is a female labor share premium for manufacturing firms, although it is smaller than for exporters (Appendix Table 5.1). Table 2 examines the mean differences between global and non-global services firms. The female labor share is 3.2 percentage points lower for GVC participants compared to non-participants (columns 1 to 4). In other words, GVC participation accounts for approximately 39% (3.2 out of 8.2 percentage points) of the female labor share gap relative to non-participants, as shown in Figure 1. Similar to the findings for exporting, the results for services firms contrast sharply with those for the manufacturing sample, where there is a significant female labor share premium of 4.57 percentage points for GVC participants compared to non-participants (Appendix Table 5.2). Finally, we do not find any difference in the female labor share between foreign-owned and domestically-owned services firms (Table 2, columns 5 to 8). This contrasts with the manufacturing firm sample, where foreign firms have a female labor share premium of 1.8 percentage points compared to domestic firms, after controlling for firm-level factors and fixed effects (Appendix Table 5.2). In summary, we find a female labor share gap for exporters and GVC participants, whereas there is no relationship for importers and foreign firms in services sectors. This may be attributed to the sectoral segregation between women and men resulting in a selection bias with women tending to pursue work opportunities in less export- and skill-intensive sectors and men being employed in more export- and skill-intensive sectors. 17 Possible explanations for such a selection bias include longer working hours, higher travel requirements, or less flexible working conditions in international firms, which can be challenging for women, especially those with family responsibilities (see, e.g., Bøler et al. 2018). Business services (D69T82), for example, show a lower female labor share for exporting firms relative to non-exporting firms. However, flexibility in the workplace has been slower to increase within the business and financial sectors (Goldin and Katz 2011). In addition, tradable services may require higher skills, leading young women to pursue education and delay entering the labor market (Wacker et al. 2017). Supporting this prior, our analysis confirms that the female labor share is smaller in sectors such as IT and other information services (section 3.2). Table 2: Female labor share difference of global vs. non-global services firms, OLS Dependent GVC participant, gvc Foreign ownership, FDI variable: (1) (2) (3) (4) (5) (6) (7) (8) fem_shisrt lnYisrt -0.102 -0.165 0.203 -0.136* -0.223* 0.130 (0.187) (0.178) (0.296) (0.082) (0.072) (0.507) lntechisrt -1.238*** -1.405*** -1.464*** -1.620*** (0.005) (0.002) (0.001) (0.000) lnlpisrt -0.623** -0.594** (0.014) (0.019) globalisrt -1.466** -1.635*** -3.283*** -3.320*** 0.724 0.663 0.926 0.842 (0.013) (0.009) (0.000) (0.000) (0.127) (0.195) (0.233) (0.278) Obs. 33105 29509 14591 14591 33105 29509 14591 14591 R2 0.36 0.36 0.32 0.32 0.36 0.36 0.32 0.32 Note: P-values are in parentheses. p*<0.1, p**<0.05, p***<0.01 (robust standard errors corrected for clustering by country- sector in parentheses). All regressions include country-sector, subnational region and year fixed effects. On the other hand, exporters in entertainment, recreation, and other service activities (D90T96) and telecommunications (D61) are more female-intensive. The nature of the work including direct interactions with foreign customers, and possibly more flexible work arrangements, could attract more women. Note that other sectors such as health or education, which are known to be female- intensive, are not part of the services firm sample. 4.2 Robustness Checks This section carries out a number of robustness checks. First, we divide the sample into two periods, 2010-2019 and 2020-2024, to determine if the findings remain consistent before and after Covid-19. Second, we add the ratio of male to female sectoral wages as an additional control measure for a sub-sample of countries to check if findings hold when accounting for relative wage differences. Third, we additionally control for female firm ownership and management, as these could shape the relationship between international status and the female labor share. Fourth, we implement the 18 fractional probit estimator as a final robustness check to address the fact that the dependent variables are bounded between zero and unity. Table 3 divides the sample into two periods, 2010-2019 and 2020-2024. The correlations between exporting and GVC participation and the female labor share are mostly statistically significant only in the pre-Covid period; only exporting has a statistically significant and smaller coefficient in both periods. This indicates that the female labor share gap for services exporters has narrowed after Covid-19. The sharp decline in services trade due to the pandemic and slower recovery relative to manufacturing could also explain this finding. This finding may partly also be due to the different sample sizes, so it should not be over-interpreted. However, it is possible that the increased adoption of digital technologies across all firms during Covid-19 (as suggested by the reduced coefficient on technology intensity) might have played a role (Avalos Almanza et al. 2023). Table 3: Female labor share difference of international vs. non-international services firms, 2010- 2019 vs. 2020-2024, OLS Dependent 2010-2019 2020-2024 variable: (1) (2) (3) (4) (5) (6) (7) (8) fem_shisrt exp imp gvc fdi exp imp gvc fdi lnYisrt 0.103 0.0980 0.0778 0.00652 0.250 0.194 0.236 0.182 (0.764) (0.790) (0.821) (0.985) (0.291) (0.419) (0.318) (0.445) lntechisrt -2.332*** -4.609*** -2.563*** -2.992*** -0.859* -1.159** -0.876* -1.004* (0.007) (0.000) (0.003) (0.001) (0.098) (0.029) (0.092) (0.054) lnlpisrt -1.271*** -1.488*** -1.251*** -1.207*** -0.392 -0.402 -0.384 -0.361 (0.004) (0.003) (0.005) (0.007) (0.201) (0.198) (0.211) (0.240) tradeisrt / -6.931*** -0.283 -8.124*** 0.0223 -1.553** 0.953 -1.409 0.845 globalisrt (0.000) (0.767) (0.000) (0.990) (0.043) (0.106) (0.118) (0.330) Obs. 4311 3789 4311 4311 10280 10003 10280 10280 R2 0.38 0.35 0.38 0.38 0.29 0.29 0.29 0.29 Note: P-values are in parentheses. p*<0.1, p**<0.05, p***<0.01 (robust standard errors corrected for clustering by country- sector in parentheses). All regressions include country-sector, subnational region and year fixed effects. The findings for the manufacturing sample are in sharp contrast, with exporting and GVC participation showing a higher female labor share in the post-Covid period, whereas importing shows a reduced premium and foreign ownership status is no longer statistically significant (Appendix Table 5.3). One potential source of omitted variable bias could arise from the absence of sex-specific wage data in the Enterprise Surveys. We address this issue by using sex-specific wage data available at the TiVA 2018 sector level for 31 countries in our dataset. We calculate the ratio of male to female wages as our control measure of relative wages at the sector level in a country. Another important modification to the model is the change in the dependent variable. Instead of estimating the female labor share (see equation 2), the regressions in this section estimate the female wage share (see 19 equation 1). Importantly, once all the independent variables are included in the descriptive regressions the country sample falls to 17 for services firms. Table 4 presents results for female wage share for different firm categories, where the dependent variable is calculated using all workers. The findings indicate that higher relative wages between male and female workers – or a higher female wage gap – are associated with a lower female wage share, which aligns with expectations (see section 2.1). Regarding the international variables of interest, we now observe a premium of 1.6 percentage points for importers and 3.4 percentage points for foreign firms. The female wage share gap for exporters and GVC participants is not evident in columns 1 and 3. The results contrast sharply with those for manufacturing firms where exporting, GVC participation and foreign ownership status – but not importing – remain linked to female labor share premia (Appendix Table 5.4). Table 4: Female wage share difference of international vs. non-international firms, reduced country sample, OLS Dependent Total Dependent Total variable: (1) (2) (3) (4) variable: (5) (6) (7) (8) exp imp gvc fdi fem_shisrt exp imp gvc fdi lnYisrt -0.659* -0.733** -0.721** -0.766** lnYisrt -0.934** -0.994*** -1.002*** -1.053*** (0.063) (0.042) (0.042) (0.031) (0.013) (0.009) (0.007) (0.005) lntechisrt 1.905** 1.576* 1.788** 1.610** lntechisrt 1.950** 1.598* 1.829** 1.637* (0.020) (0.067) (0.029) (0.048) (0.024) (0.077) (0.034)) (0.058) lnlpisrt 0.917** 0.967** 0.949** 0.965** lnlpisrt 1.130** 1.138** 1.165** 1.184** (0.048) (0.042) (0.041) (0.037) (0.022) (0.025) (0.018) (0.016) ln(wM/wF)cst -4.593*** -4.261** -4.537** -4.542** (0.010) (0.019) (0.011) (0.010) tradeisrt / -1.257 1.638* 1.843 3.414** tradeisrt / -1.709 1.656 1.531 3.429** globalisrt (0.296) (0.080) (0.230) (0.012) globalisrt (0.187) (0.102) (0.343) (0.017) Obs. 3782 3624 3782 3782 Obs. 3782 3624 3782 3782 R2 0.26 0.26 0.26 0.26 R2 0.23 0.23 0.23 0.23 Note: P-values are in parentheses. p*<0.1, p**<0.05, p***<0.01 (robust standard errors corrected for clustering by country- sector in parentheses) for the sample of 17 countries for which relative wage rates at the sector level and other control variables are available. All regressions include sector, subnational region and year fixed effects. While these results provide important indicative information, it is crucial not to over-interpret them for two reasons. First, the results for the reduced sample of 17 countries only cover a sixth of the full country sample for services firms, so observed trends may be country-specific. To test for this possibility, we estimate the baseline regressions (equation 2) just for these 17 countries (columns 5 to 8). Only foreign ownership status is associated with a higher female labor share, with coefficients of similar magnitude compared to the alternative specification in equation (1) (column 4). Additionally, the regressions include sectoral (not country-sector) fixed effects. This is a key difference from the other descriptive regressions presented previously, due to the inclusion of wage 20 data, which vary only by sector within countries, not by firm. Overall, the findings do not change when including relative wage rates as additional control variable. As an additional robustness check, we revert to the original specification in equation (2) and incorporate two more potential influences on the female labor share to address concerns about omitted variables. Specifically, we include the female ownership share of each firm and a dummy variable for firms with a female top manager. Both variables are expected to have a positive association with the female labor share, based on the assumption that women are less likely to discriminate against other women in the hiring process compared to male owners or managers (see, e.g., Song 2022). The estimation equation therefore becomes: = + + + + + + + + + (2a) where the female vector includes variables measuring the female ownership percentage and a dummy for a female top manager. Table 5 presents the results. As expected, female ownership and female management are both associated with higher female labor shares in services firms, with all effects being statistically significant at the 1% level. Importantly, the negative coefficients on exporting and GVC participation remain similar to those reported in Tables 1 and 2 above and are statistically significant at the 1% level. Foreign ownership status still does not impact the female labor share of services firms. The only difference is that importers now also exhibit a statistically significant female labor share gap. Therefore, we conclude that integration with the global economy is an important determinant in explaining female labor share gaps in services firms, even after accounting for female involvement in firm ownership and management. This finding, however, does not apply to foreign-owned firms. The results for manufacturing firms suggest that the female labor share premium continues to hold for exporting, GVC participation and foreign ownership status, but not for importing firms (Appendix Table 5.5). Finally, from an econometric point of view, it is important to take account of the fact that the dependent variables in equations (1) and (2) are shares, so they are bounded between zero and unity. Papke and Wooldridge (1996) derive the fractional logit estimator as a solution to exactly this kind of problem. We implement the closely related heteroskedastic fractional probit model of Bluhm (2013), which we prefer to the in-built Stata command for fractional response regression because it 21 is significantly more rapid in terms of computation time. We report coefficient estimates as average marginal effects, with standard errors computed using the delta method. The results in Appendix 6 broadly confirm the overall findings. Table 5: Female labor share difference of international vs. non-international services firms, role of female ownership and top manager, OLS Dependent Ownership Management variable: (1) (2) (3) (4) (5) (6) (7) (8) fem_shisrt exp imp gvc fdi exp imp gvc fdi lnYisrt 0.183 0.445** 0.165 0.113 0.247 0.392** 0.212 0.159 (0.357) (0.029) (0.405) (0.571) (0.205) (0.048) (0.276) (0.416) lntechisrt -1.530*** -2.216*** -1.604*** -1.798*** -1.360*** -2.005*** -1.419*** -1.597*** (0.001) (0.000) (0.000) (0.000) (0.002) (0.000) (0.001) (0.000) lnlpisrt -0.613** -0.845*** -0.606** -0.582** -0.666*** -0.771*** -0.633** -0.622** (0.017) (0.001) (0.018) (0.024) (0.008) (0.003) (0.012) (0.014) femaleisrt 0.0703*** 0.185*** 0.112*** 0.0957*** 7.893*** 15.19*** 7.936*** 10.27*** (0.000) (0.000) (0.000) (0.009) (0.000) (0.000) (0.000) (0.000) tradeisrt / -4.279*** -3.100*** -5.025*** -0.854 -4.576*** -2.521*** -4.725*** -1.148 globalisrt (0.000) (0.000) (0.000) (0.336) (0.000) (0.000) (0.000) (0.167) Obs. 14172 13417 14172 14172 14564 13766 14564 14564 R2 0.32 0.33 0.32 0.32 0.32 0.33 0.32 0.32 Note: P-values are in parentheses. p*<0.1, p**<0.05, p***<0.01 (robust standard errors corrected for clustering by country- sector in parentheses). All regressions include sector, subnational region and year fixed effects. 4.3 Alternative Trade in Employment Dataset This section examines whether the female labor share gaps or premia can also be detected at the sector level within countries. The OECD Trade in Employment (TiE) and Trade in Value-added (TiVA) databases provide value-added and employment measures for 45 sectors over the period 1995-2020 for 76 countries. Employment data are available separately for men and women as part of the extended Trade in Employment by Workforce Characteristics (TiMBC) database. Importantly, the dataset publishes these measures for direct export activities, i.e., value added and employment in exports. By taking the difference (total minus export-linked), one can obtain measures for value added and employment for the rest of the sector. These measures allow for the computation of the female labor share and value added for export activities and the rest of the sector. This alternative dataset also addresses the main caveat of using the Enterprise Surveys, namely that is not clearly identifiable if exports are aligned with the main activity of the firm. Domestic employment by industry statistics are drawn from various sources such as OECD’s Annual National Accounts and Structural Analysis (STAN) databases, official national statistics and, in a very few cases, from research projects such as, for example, India KLEMS (Das et al., 2017). Employment is defined as persons engaged in production activity within the National Accounts 22 boundary of the resident institutional unit (domestic concept) and includes both employees and self- employed. If there are no estimates of employment by industry in official National Accounts statistics, then Labor Force Survey (LFS) statistics are exploited. National Accounts are preferred to LFS as a source for employment by industry since LFS are usually based on residential households and thus exclude non-resident workers while including resident workers commuting abroad (national concept). To generate insights, Figures 8 and 9 illustrate the female labor share for export and non-export activities in our sample of 76 countries for the years 2010 and 2020, distinguishing between services sectors and manufacturing industries. The initial observation for manufacturing sectors suggests a female labor share premium of export activities compared to non-export activities for several countries which was more pronounced in 2010 compared to 2020 (Figure 8). In contrast, the scatterplot suggests a female labor share gap in services sectors across most countries, which has persisted over time (Figure 9). Although female labor force participation varies across countries, a wage premium in the manufacturing sector is observed in several economies like Viet Nam, Bulgaria, Philippines, Cyprus, and Türkiye. These countries have actively participated in GVCs, particularly in female-dominated sectors such as apparel and electronics. The results are less pronounced than for the Enterprise Survey manufacturing sample, possibly because the OECD sample includes high-income economies where manufacturing jobs can be more male-dominated (e.g., machine operators, technicians) compared to those in low- and middle-income countries. To rule out that unobserved factors at the country and sector level do not influence the results, we specify the following female labor share regressions: _ = + + + + (3°) where femsh designates the female labor share, c a country, s a sector, and a the type of economic activity (direct export activity or rest of the sector). Exp is a dummy taking the value of 1 if a worker is linked to an export activity a within sector s, and 0 if the worker is linked to non-export activities in the sector. We rely on a vector of unobserved country and sector fixed effects. We also specify an augmented regression model that is more closely related to equation (2) in section 2.2 as follows: _ = + + + + + (3b) 23 where Y denotes output for export and non-export activities. Figure 8: Female labor share difference by exporting activity, manufacturing sectors, 2010 vs. 2020 Manufacturing sectors, 2010 60 Female labor share (%) non-exporting lva ltunzl bgr vnm col est rus per rou prt 40 twn chn hrvkazhun svn sgp cyp lux jpn cze svk mex cri pol deu irl canfra dnk aut fin che chl usa ita gbr sweaus belesp grc isl nor nld tur 20 mlt kor egy 0 0 20 40 60 Female labor share (%) exporting Manufacturing sectors, 2020 60 tha Female labor share (%) non-exporting nzl bgr lva ltu rou vnm est prt rus phl 40 ukrcoltwn per kaz sgp hun luxjpn bra svk mys hrv cze svn pol chl kor fraisr che cri arg grc deu cyp aut fin gbr esp irl aus dnk indusa can swe mex ita mlt isl nldbel nor tur 20 egy sau 0 0 20 40 60 Female labor share (%) exporting Source: OECD Trade in Employment by Workforce Characteristics and Trade in Value added data. See Appendix 2 for the types of manufacturing sectors included in the database. 24 Figure 9: Female labor share difference by exporting activity, services sectors, 2010 vs. 2020 Services sectors, 2010 est ltu lva isl irl fin 60 nor swe dnk lux hun prtsvn polsvk bgr cyp fra aut hrv can cze deu nld bel che gbr esp aus usa rus Female labor share (%) non-exporting rou kor chl vnm per ita col jpn sgp kaz twn grc nzl mex chn cri mlt 40 tur egy 20 0 0 20 40 60 Female labor share (%) exporting Services sectors, 2020 ltu est lva irl fin lux 60 nor ukr svn pol prt hrv aut svk bgr dnk swe islnldisr bel cze fra hun can deu cyp arg rou chegbr esp aus phl thavnm rus Female labor share (%) non-exporting bra chl ita kor sgpkaz per twn usamlt jpn grc col nzl mys cri 40 mex tur ind egy 20 sau 0 0 20 40 60 Female labor share (%) exporting Source: OECD Trade in Employment by Workforce Characteristics and Trade in Value added data. See Appendix 2 for the types of services sectors included in the database. Our findings in Table 6 suggest that exporting activities show a significantly higher female labor share relative to non-exporting activities for manufacturing sectors, while we find a female labor share gap for services sectors. After controlling for output, only the female labor share gap for services remains statistically significant. Overall, this robustness check confirms the female labor share gap for services exporters at the firm level using the Enterprise Survey data. 25 Table 6: Female labor share by exporting activity of exporting vs. non-exporting activity, services vs. manufacturing, 2010-2020 Dependent Services Manufacturing variable: OLS Fractional Probit OLS Fractional Probit fem_shcsat (1) (2) (3) (4) (5) (6) (7) (8) lnYcst -0.00335 -0.00309 0.00334 0.00314 (0.142) (0.233) (0.480) (0.498) expcsat -0.0113*** -0.0191*** -0.0111*** -0.0257*** 0.00843* 0.00992 0.00850** 0.00980 (0.004) (0.000) (0.003) (0.000) (0.074) (0.125) (0.036) (0.110) Obs. 7503 7227 7503 7228 2504 2412 2504 2412 R2/Pseudo-R2 0.922 0.931 0.629 0.671 0.953 0.952 0.0922 0.0969 Note: P-values are in parentheses. p*<0.1, p**<0.05, p***<0.01 (robust standard errors corrected for clustering by country- sector in parentheses). The fractional probit results present average marginal effects. All regressions include country- sector, country-time and sector-time fixed effects. 5. Conclusions Using a cross-section of more than 33,000 services firms in 104 LMICs from the World Bank’s Enterprise Surveys, this paper assessed whether the female labor share premium of international relative to non-international firms in manufacturing also holds for services firms. It focused on four types of international firms: exporters, importers, GVC participants and FDI firms. The stylized facts found important differences across broad sectors and countries. In manufacturing, international firms, especially exporters and GVC participants, generally have higher female labor shares than other firms. This trend is less pronounced for importers and foreign-owned firms. However, in services, the results are less clear, with no consistent trends except for a positive relationship between importing and the female labor share. Notably, East Asia and Eastern Europe show a female labor share premium in global manufacturing but a male premium in global services. A negative relationship exists between exporting and GVC participation and the female labor share, while no relationship is found for importing or foreign ownership, controlling for firm output, productivity, technology intensity and fixed effects. The negative correlations for exporting and GVC participation are mainly significant pre-Covid-19, with only exporting remaining significant (though smaller) in both periods. Including female top management and female ownership confirms the negative relationship for exporting and GVC participation and reveals a negative relationship for importing services firms. Controlling for sector-level relative wages between men and women does not change the findings in a smaller sub-sample of 17 economies. The findings also indicate that higher relative wages between male and female workers – or a higher female wage gap – are 26 associated with a lower female wage share which is in line with expectations. Using an alternative estimator and dataset confirms these findings, highlighting a female labor share gap for services traders and a premium for manufacturing traders. The female labor share gap for exporters and GVC participants in services sectors may be attributed to the sectoral segregation between women and men, with women tending to pursue work opportunities in less export-intensive and skill-intensive sectors (such as retail and hospitality), and men being employed in more export-intensive sectors (such as transportation) and skill-intensive services sectors (such as information and communication technologies). Given that international firms in modern services sectors are more likely to hire more skill-intensive or specialized labor, the results would reflect the skill disadvantage that women have relative to men which precludes their employment in such jobs. Possible explanations include longer working hours, higher travel requirements, or less flexible working conditions in trading firms, which can be challenging for women, especially those with family responsibilities (see, e.g., Bøler et al. 2018). In addition, tradable services may require higher skills, leading young women to pursue education and delay entering the labor market (Wacker et al. 2017). Since skill-intensive modern services jobs are likely to pay more, this also explains the negative link between the female wage gap and female wage share. In contrast, in manufacturing, women tend to have the “right” skill-set for jobs in GVC-intensive sectors like textiles or computers and electronics and these sectors also tend to be more outward-oriented, as confirmed by our findings covering manufacturing firms. While this paper examined the relationship between trade and female labor force participation from multiple perspectives, including various trade types and firm-level female labor participation, several areas warrant further research: (1) The role of policy, such as labor regulations, in shaping the relationship between trade and female labor share. (2) Whether the patterns observed remain consistent when incorporating more direct measures of technology or innovation. (3) Whether the findings hold when considering worker characteristics. (4) Whether a subgroup of countries with panel data can be used to establish a stronger causal link. 27 References Aguayo, E., Airola, J., Juhn, C., C. Villegas-Sanchez (2013). “Did trade liberalization help women? The case of Mexico in the 1990s.” Research in Labor Economics, 38, 1–35. Aguayo‐Tellez, E. (2012), "The impact of trade liberalization policies and FDI on gender inequalities: a literature review." World Development Report 2012, World Bank, available at: https://openknowledge.worldbank.org/server/api/core/bitstreams/6a0d687f-a129-507d- 8247-9446cb976ddc/content. Avalos Almanza, Edgar Itamar; Cirera, Xavier; Vargas Da Cruz, Marcio Jose; Iacovone, Leonardo; Medvedev, Denis; Nayyar, Gaurav; Reyes Ortega, Santiago. “Firms’ Digitalization during the COVID-19 Pandemic : A Tale of Two Stories (English).” Policy Research working paper No. WPS 10284; COVID-19 (Coronavirus) Washington, D.C. : World Bank Group. Ben Yahmed, S. (2017). “Gender Wage Discrimination and Trade Openness. Prejudiced Employers in an Open Industry,” ZEW Discussion Paper No. 17-047, Mannheim. Bluhm, R. (2013). “Fhetprob: A Fast QMLE Stata Routine for Fractional Probit Models with Multiplacative Heteroskedasticity.” Working paper, http://www.richard-bluhm.com/wp- content/uploads/2013/02/fhetprob.pdf. Bøler, E. A., Javorcik, B. and K. H. Ulltveit-Moe (2018). “Working across time zones: Exporters and the gender wage gap,” Journal of International Economics, 111, 122-133. Brown, R. and L. Christensen (1981). “Estimating elasticities of substitution in a model of partial static equilibrium: an application to US agriculture, 1947–1974”, In: Berndt, E. and B. Field (eds.) Modeling and measuring natural resource substitution, MIT Press: Cambridge, MA, 209-229. Busse, M. and C. Spielmann (2006). “Gender inequality and trade,” Review of International Economics, 14(3), 362-79. Carranza, E., C. Dhakal and I. Love (2018). “Female Entrepreneurs: How and Why are they Different?”, Jobs Working Paper, No. 20, Washington, D.C.: The World Bank Group. Chen, Z., Y. Ge, H. Lai and C. Wan (2013). “Globalization and gender wage inequality in China,” World Development, 44, 256-66. Ederington, J., J. Minier and K.R. Troske (2009). “Where the Girls are: Trade and Labor Market Segregation in Colombia,” IZA Discussion Paper No. 4131. Feenstra, R. and G. Hanson (1996). “Globalization, Outsourcing and Wage Inequality,” American Economic Review, 86(2), 240-45. Geishecker, I. (2006). “Does outsourcing to central and eastern Europe really threaten Manual Workers’ Jobs in Germany?”, The World Economy, 29(5), 559-83. Goldin, C., and L. Katz (2011). “The Cost of Workplace Flexibility for High-Powered Professionals.” The Annals of the American Academy of Political and Social Science, 638(1), 45- 67. Heath, R., A. Bernhardt, G. Borker, A. Fitzpatrick, A. Keats, M. McKelway, A. Menzel, T. Molina, G. Sharma (2024). “Female Labour Force Participation,” VoxDevLit, 11(1), February 2024. Juhn, C., G. Ujhelyi and C. Villegas-Sanchez (2013). “Trade Liberalization and Gender Inequality,” American Economic Review, 103(3): 269-73. Juhn, C., Ujhelyi, G. and C. Villegas-Sanchez (2014). “Men, women, and machines: How trade impacts gender inequality,” Journal of Development Economics, 106, 179-93. 28 Oostendorp, R.H. (2009). “Globalization and the Gender Wage Gap,” World Bank Economic Review, 23, 141-61. Papke, L. and J. Wooldridge (1996). “Econometric Methods for Fractional Response Variables with an Application to 401(k) Plan Participation Rates,” Journal of Applied Econometrics 11 (6): 619-32. Rocha, N. and D. Winkler (2019), “Trade and Female Labor Participation: Stylized Facts Using a Global Dataset.” Policy Research Working Paper 9098, World Bank, Washington, DC. Rocha, N., B. Shepherd, and D. Winkler (2020), “Trade and Female Labor Participation: Stylized Facts Using a Global Dataset,” mimeo. Shephard, R. (1953). Cost and Production Functions, Princeton University Press: Princeton. Shepherd, B. (2018). “Global Value Chains and Women's Labor: Firm-Level Evidence,” Paper prepared for the Economic Research Institute for ASEAN and East Asia. Song, Jun Myung (2022). “Female senior managers and the gender equality environment: Evidence from south Korean firms,” Pacific-Basin Finance Journal, 75. Tejani, S., and W. Milberg (2016), “Global defeminization? Industrial upgrading and manufacturing employment in developing countries,” Feminist Economics, 22(2), 24-54 Wacker, K. M., A. Cooray, and I. Gaddis (2017), “Globalization and female labor force participation in developing countries: an empirical (re-) assessment,” Globalization: Strategies and Effects, 545- 583. Wood, A. (1991). “North‐South trade and female labour in manufacturing: An asymmetry,” The Journal of Development Studies, 27(2), 168-89. World Bank-World Trade Organization (2020). Women and Trade: The Role of Trade in Promoting Gender Equality. Washington, DC: World Bank. 29 Appendices Appendix 1: Number of firms by country, services vs. manufacturing Services Manufacturing Economy Year Firms Economy Year Firms Albania 2019 190 Albania 2019 126 Angola 2010 137 Argentina 2017 625 Antigua and Barbuda 2010 104 Armenia 2020 270 Argentina 2017 291 Bangladesh 2022 531 Armenia 2020 242 Belarus 2018 322 Azerbaijan 2019 152 Bolivia (Plurinational State of) 2017 105 Bangladesh 2022 262 Bosnia and Herzegovina 2023 130 Belarus 2018 231 Botswana 2023 163 Bolivia (Plurinational State of) 2017 213 Bulgaria 2023 371 Bosnia and Herzegovina 2023 179 Cambodia 2023 261 Botswana 2023 386 Chile 2010 763 Bulgaria 2023 212 China 2012 1638 Cambodia 2023 242 Colombia 2023 290 Cameroon 2016 228 Congo, Dem. Rep. 2013 235 Chile 2010 238 Costa Rica 2023 106 China 2012 868 Côte d'Ivoire 2023 272 Colombia 2023 503 Croatia 2023 143 Congo, Rep. 2024 219 Czechia 2019 280 Congo, Dem. Rep. 2013 268 Egypt, Arab Rep. 2020 1988 Costa Rica 2023 221 El Salvador 2023 287 Côte d'Ivoire 2023 356 Estonia 2023 129 Croatia 2023 290 Ethiopia 2015 364 Cyprus 2019 125 Georgia 2023 221 Czechia 2019 169 Ghana 2023 193 Djibouti 2013 159 Guatemala 2017 127 Dominica 2010 112 Hungary 2023 391 Dominican Republic 2016 231 India 2022 5212 Ecuador 2017 227 Indonesia 2023 1007 Egypt, Arab Rep. 2020 891 Iraq 2022 335 El Salvador 2023 266 Jamaica 2010 101 Estonia 2023 172 Jordan 2019 242 Ethiopia 2015 331 Kazakhstan 2019 846 Georgia 2023 330 Kenya 2018 440 Ghana 2023 430 Kosovo 2019 109 Grenada 2010 108 Kyrgyzstan 2023 148 Guatemala 2017 177 Lao PDR 2018 139 Guinea 2016 101 Latvia 2019 119 Honduras 2016 223 Lebanon 2019 261 Hong Kong SAR, China 2023 473 Lithuania 2019 121 Hungary 2023 261 Malaysia 2019 659 India 2022 3423 Mauritius 2023 108 Indonesia 2023 1061 Mexico 2023 648 Iraq 2022 382 Mongolia 2019 113 Jamaica 2010 211 Morocco 2023 239 Jordan 2019 174 Mozambique 2018 283 Kazakhstan 2019 394 Myanmar 2016 345 Kenya 2018 489 Namibia 2014 148 Kyrgyzstan 2023 166 Nepal 2023 210 Lao PDR 2018 170 Nicaragua 2016 104 Latvia 2019 182 Nigeria 2014 822 30 Lebanon 2019 231 North Macedonia 2023 116 Lithuania 2019 197 Pakistan 2022 844 Madagascar 2022 221 Panama 2010 107 Malawi 2014 170 Paraguay 2023 102 Malaysia 2019 436 Peru 2023 432 Mauritius 2023 186 Philippines 2023 360 Mexico 2023 495 Poland 2019 743 Mongolia 2019 167 Romania 2023 489 Morocco 2023 205 Russian Federation 2019 739 Mozambique 2018 273 Rwanda 2023 132 Myanmar 2016 224 Saudi Arabia 2022 642 Namibia 2014 285 Senegal 2014 225 Nepal 2023 364 Serbia 2019 114 New Zealand 2023 178 Sierra Leone 2023 101 Nicaragua 2016 204 Slovak Republic 2023 108 Niger 2017 101 Slovenia 2019 153 Nigeria 2014 1017 South Africa 2020 332 North Macedonia 2023 199 Sri Lanka 2011 354 Pakistan 2022 422 Tajikistan 2019 115 Panama 2010 223 Tanzania, United Republic of 2023 193 Paraguay 2023 230 Trinidad and Tobago 2010 113 Peru 2023 478 Tunisia 2020 354 Philippines 2023 507 Türkiye 2019 1020 Poland 2019 223 Uganda 2013 351 Romania 2023 311 Ukraine 2019 886 Russian Federation 2019 353 Uzbekistan 2019 782 Rwanda 2023 222 Viet Nam 2023 556 Saudi Arabia 2022 514 West Bank and Gaza 2023 123 Senegal 2014 316 Yemen, Rep. 2013 107 Serbia 2019 183 Zambia 2019 162 Sierra Leone 2023 101 Zimbabwe 2016 276 Slovak Republic 2023 155 TOTAL 34,221 Slovenia 2019 181 South Africa 2020 611 Note: The following 17 countries with sectoral sex- South Sudan 2014 599 disaggregated wage data are included in the services Sri Lanka 2011 242 firm samples: Botswana, Cambodia, Colombia, Costa Sudan 2014 503 Rica, Djibouti, El Salvador, Georgia, Indonesia, Jordan, Mexico, Mongolia, Mauritius, Nepal, Philippines, Suriname 2018 118 Russian Federation, Sierra Leone, and Viet Nam. Tajikistan 2019 128 The manufacturing firm sample includes the following Tanzania, United Republic of 2023 383 26 countries: Argentina, Bangladesh, Bolivia Timor-Leste 2021 112 (Plurinational State of), Botswana, Chile, Colombia, Trinidad and Tobago 2010 209 Costa Rica, Ethiopia, Georgia, Indonesia, Iraq, Jordan, Tunisia 2020 205 Cambodia, Mexico, Mongolia, Mauritius, Nicaragua, Türkiye 2019 405 Nepal, Peru, Philippines, Russian Federation, Sierra Uganda 2013 340 Leone, El Salvador, Tajikistan, Uganda, Viet Nam. Ukraine 2019 307 Uruguay 2017 206 Uzbekistan 2019 295 Venezuela, RB 2010 216 Viet Nam 2023 300 West Bank and Gaza 2023 203 Yemen, Rep. 2013 207 Zambia 2019 358 Zimbabwe 2016 295 TOTAL 33,284 31 Appendix 2: Number of firms, services vs. manufacturing Services TiVA 2018 Sector name Firms Percent D45T47 Wholesale and retail trade; repair of motor vehicles 20,561 61.8% D49T53 Transportation and storage 2,607 7.8% D55T56 Accommodation and food services 7,287 21.9% D58T60 Publishing, audiovisual and broadcasting activities 278 0.8% D61 Telecommunications 271 0.8% D62T63 IT and other information services 842 2.5% D69T82 Other business sector services 1,109 3.3% D90T96 Arts, entertainment, recreation and other service activities 329 1.0% TOTAL 33,284 100% Manufacturing TiVA 2018 Sector name Firms Percent D10T12 Food products, beverages and tobacco 7,517 22.0% D13T15 Textiles, wearing apparel, leather and related products 6,889 20.1% D16 Wood and products of wood and cork 932 2.7% D17T18 Paper products and printing 1,141 3.3% D19 Coke and refined petroleum products 87 0.3% D20T21 Chemicals and pharmaceutical products 2,044 6.0% D22 Rubber and plastic products 1,980 5.8% D23 Other non-metallic mineral products 2,779 8.1% D24 Basic metals 1,082 3.2% D25 Fabricated metal products 3,406 10.0% D26 Computer, electronic and optical products 455 1.3% D27 Electrical equipment 848 2.5% D28 Machinery and equipment, nec 1,925 5.6% D29 Motor vehicles, trailers and semi-trailers 887 2.6% D30 Other transport equipment 164 0.5% D31T33 Other manufacturing; repair and installation of machinery and equipment 2,085 6.1% TOTAL 34,221 100% Note: TiVA classification based on most important product of firm. 32 Appendix 3: Summary statistics, services vs. manufacturing Services Variable Obs Mean Std. dev. Min Max Female labor participation fem_sh_perm 33,210 33.13996 27.71175 0 100 fem_sh_per~w 6,261 34.69018 24.72865 0 99.5521 lnwage_rat~f 28,741 7.606307 1.794615 -8.2284 22.45576 Trade exporter 33284 0.113508 0.317218 0 1 importer 25501 0.433709 0.495596 0 1 gvc 33284 0.054981 0.227947 0 1 FDI 33284 0.083524 0.276676 0 1 Controls lnsales_def 29660 12.51409 2.376369 -0.58821 23.35165 tech_med 16102 0.234257 0.493444 0 2 lnprod_med 29636 9.559747 1.930512 -4.72144 20.16372 lnrel_wage 6320 4.753926 0.369997 3.458598 6.523321 Characteristics fem_own 31110 17.32414 33.84011 0 100 fem_man 33198 0.191759 0.39369 0 1 33 Manufacturing Variable Obs Mean Std. dev. Min Max Female labor participation fem_sh_perm 34,028 31.58238 31.7928 0 100 fem_sh_prod 33,749 29.94768 35.03853 0 100 fem_sh_nonp 31,605 36.94851 33.59327 0 100 fem_sh_per~w 5,768 33.34079 27.98353 0 100 Trade exporter 34,221 0.272698 0.445353 0 1 importer 33,670 0.403267 0.490561 0 1 gvc 34,221 0.177055 0.381721 0 1 FDI 34,206 0.089721 0.285786 0 1 Controls lnsales_def 30,182 13.37472 2.312589 0.808394 21.12873 lncapint 23,985 -1.1668 1.780617 -15.0195 9.416192 lntfp 21,370 0.586996 0.720478 -6.48438 2.124743 lnrel_wage 5,881 4.786585 0.367216 2.359362 7.570676 Characteristics fem_own 32,054 12.51198 29.01237 0 100 fem_man 34,162 0.128827 0.335014 0 1 34 Appendix 4: Analysis for manufacturing firms Appendix Figure 4.1: Female labor share, services vs. manufacturing, by international status v Manufacturing,production Manufacturing, non-production exporting 43.9% 39.5% 28.1% exporting 30.9% importing 42.6% importing 37.2% 30.1% yes 31.9% yes 48.2% no 40.9% no gvc-participant gvc-participant 31.9% 29.5% fdi firm 52.3% fdi firm 43.4% 32.3% 32.8% Note: The female labor shares are averages using firm employment as weights. Appendix Figure 4.2: Female labor share in manufacturing firms, by sub-sector Textiles, wearing apparel, leather and related… 57.9% Computer, electronic and optical products 42.5% Food products, beverages and tobacco 38.0% Electrical equipment 32.6% Other manufacturing; repair and installation of… 32.4% Other transport equipment 31.2% Rubber and plastic products 29.6% Wood and products of wood and cork 28.9% Chemicals and pharmaceutical products 28.3% Paper products and printing 25.8% Coke and refined petroleum products 24.6% Motor vehicles, trailers and semi-trailers 22.7% Machinery and equipment, nec 22.1% Other non-metallic mineral products 21.0% Fabricated metal products 19.5% Basic metals 15.8% Note: For the sectoral distribution of firms, see Appendix 2. The female labor shares are averages using firm employment as weights. 35 Appendix Figure 4.3: Female labor share, trade participants vs. non-participants, by manufacturing sub-sector Exporting vs non-exporting firms 70% Female labor share (%), non-exporting 60% 50% D13T15 40% D10T12 D26 30% D19 D17T18 D22 D27 D20T21 D16 D31T33 20% D23 D25 D28 D30 D24 D29 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% Female labor share (%), exporting Importing vs non-importing firms 70% Female labor share (%), non-importing 60% 50% D13T15 40% D10T12 30% D16 D27 D22 D31T33 D20T21 D26 D23 D17T18 20% D25 D19 D29 D28 D30 D24 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% Female labor share (%), importing Source: The female labor shares are averages using firm employment as weights. Sector numbers relate to TiVA 2018 sectors. For sector names and the sectoral distribution of firms, see Appendix 2. 36 Appendix Figure 4.4: Female labor share, global firms vs. non-global firms, by manufacturing sub-sector GVC participants vs non-participants 70% Female labor share (%), non-participant 60% 50% D13T15 40% D10T12 30% D31T33 D26 D16 D22 D27 D19D17T18D20T21 20% D23 D25 D28 D29 D30 D24 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% Female labor share (%), GVC participant FDI vs non-FDI firms 70% Female labor share (%), non-FDI 60% 50% 40% D10T12 D26 30% D22 D16 D30 D31T33 D17T18 D20T21 D27 20% D23 D28 D19 D29 D25 D24 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% Female labor share (%), FDI Source: The female labor shares are averages using firm employment as weights. Sector numbers relate to TiVA 2018 sectors. For sector names and the sectoral distribution of firms, see Appendix 2. 37 alb 99.1% xkx 90.1% khm 79.1% vnm 70.1% mng 65.0% mmr 64.7% zaf 63.5% blr 58.9% bgr 57.9% tun 57.0% lao 54.4% geo 54.1% idn 52.9% rou 51.8% rwa 50.6% lka 50.4% ltu 49.6% srb 48.7% pol 48.5% eth 48.2% tza 47.8% arm 47.6% ukr 47.5% bgd 46.1% svn 46.1% bwa 45.8% slv 45.6% est 45.6% tjk 44.7% col 43.7% bih 43.6% phl 43.3% mkd 43.1% uzb 42.9% rus 42.4% jor 42.3% lva 42.1% zmb 41.9% mex 41.5% nic 41.3% 38 kaz 40.1% mys 39.4% gha 38.9% hrv 38.4% hun 38.3% cze 38.2% kgz 37.5% svk 37.4% ken 37.1% jam 36.9% per 36.8% nam 34.8% tur 34.3% cri 33.6% chn 32.6% mus 32.1% mar 31.8% pan 31.1% civ 31.0% tto 27.3% pry 27.2% sle 26.6% gtm 26.3% bol 23.7% moz 23.7% uga 23.2% npl 23.0% Appendix Figure 4.5: Female labor share in manufacturing firms, by country chl 22.5% cod 22.1% zwe 21.6% arg 20.5% lbn 20.0% ind 19.9% sen 16.3% nga 16.2% egy 15.0% pse 12.5% yem 11.7% irq 10.7% Note: See Appendix 1 for full country names. The female labor shares are averages using firm employment as weights. pak 9.0% sau 7.6% Appendix Figure 4.6: Female labor share in manufacturing, trade participants vs. non-participants, by country Exporting vs non-exporting firms 100% alb Female labor share (%), non-exporting 90% xkx 80% 70% zaf 60% mng svk pol ltu idn vnm 50% rou gha bwa rus bih ukr svn srb rwa bgr blr mmr 40% hrv mex tjk col khm civ jam est kaz arm lka lao chncri mar cze kgz myslva nic slv phl uzb eth geo 30% tur hun mkd pan tza tto gtmsle mus nam per zmb tun cod bolchlmozuga pry jorken bgd 20% arg indzwe pse nga lbn npl egy 10% sau pak yem sen irq 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Female labor share (%), exporting Importing vs non-importing firms 100% alb Female labor share (%), non-importing 90% 80% xkx 70% mng 60% vnm zaf bwa rwa rou bgr 50% hrvnic mus rus est ukr pol blr srb lka idn tun mmr khm perghazmb bih arm svn 40% kgz hun svkmex kaz col uzb slv ltu geo sle civ pan mar phl bgd tjk lao cod chn criturcze ken mkd lva eth 30% uganpl jam zwe chl moz tto pry gtm nam mys jor tza 20% pse arg indbol nga lbn 10% egy sen sau irqpak 0% yem 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Female labor share (%), importing Source: The female labor shares are averages using firm employment as weights. See Appendix 1 for full country names. 39 Appendix Figure 4.7: Female labor share in manufacturing, global vs. non-global firms, by country GVC participant vs non-participants 100% alb Female labor share (%), non-participants 90% xkx 80% 70% zaf vnmmng 60% rwa pol rou 50% bgr svkrusbwa ukr ltu srb blr idn lka khm hrvmex zmb nic col est arm tjk mmr 40% kgz jamgha bih svn slv kaz phl uzb tun laogeo civ mar nam per jor hunlva bgd chn tur cze 30% gtm tto cri pan mus pry mkd tzaeth bol zwe cod chl moz sle mys ken 20% arg uga nga indnpl pse lbn 10% egy sen irq sau pak yem 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Female labor share (%), GVC participants FDI vs. non-FDI firms 100% alb 90% xkx Female labor share (%), non-FDI 80% 70% zaf mng 60% vnm bwa tun blr rwa rou bgr 50% bihgha armlka poltza idn lao khm mmr col rus mkd bgd ukr srb ltu kgz hrv lva mex est uzb eth geo 40% kaz phl tjk nam zmb jam mys mar pan tur svk mus cze per ken hun chn slv jor 30% nic pry sle npl moz tto civ gtm cri svn 20% arg chl zwe uga ind lbn bol egy nga 10% pak cod irqyem sen pse sau 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Female labor share (%), FDI Source: The female labor shares are averages using firm employment as weights. See Appendix 1 for full country names. 40 Appendix 5: OLS estimates manufacturing firms Appendix Table 5.1: Female labor share difference of trading vs. non-trading manufacturing firms, OLS Dependent Exporter, exp Importer, imp variable: (1) (2) (3) (4) (5) (6) (7) (8) fem_shisrt lnYisrt -0.475*** -0.370*** -0.383*** -0.277*** -0.164* -0.144 (0.000) (0.000) (0.000) (0.001) (0.095) (0.173) lncapintisrt -0.368*** -0.390*** -0.322*** -0.322** (0.002) (0.002) (0.006) (0.012) lntfpisrt -0.159 -0.0977 (0.724) (0.828) globalisrt 2.596*** 3.902*** 4.229*** 4.626*** 0.750** 1.279*** 1.754*** 1.687*** (0.000) (0.000) (0.000) (0.000) (0.029) (0.000) (0.000) (0.000) Obs. 33921 29922 23777 21309 33399 29543 23560 21144 R2 0.45 0.47 0.48 0.48 0.45 0.47 0.48 0.48 Note: P-values are in parentheses. p*<0.1, p**<0.05, p***<0.01 (robust standard errors corrected for clustering by country- sector in parentheses). All regressions include country-sector, subnational region and year fixed effects. Appendix Table 5.2: Female labor share difference of global vs. non-global manufacturing firms, OLS Dependent GVC participant, gvc Foreign ownership, FDI variable: (1) (2) (3) (4) (5) (6) (7) (8) fem_shisrt lnYisrt -0.408*** -0.295*** -0.298*** -0.284*** -0.151 -0.114 (0.000) (0.003) (0.005) (0.001) (0.122) (0.283) lncapintisrt -0.352*** -0.370*** -0.322*** -0.331*** (0.002) (0.004) (0.006) (0.010) lntfpisrt -0.139 -0.151 (0.757) (0.740) globalisrt 2.959*** 3.780*** 4.197*** 4.573*** 1.341*** 2.054*** 2.398*** 1.832*** (0.000) (0.000) (0.000) (0.000) (0.007) (0.000) (0.000) (0.006) Obs. 33921 29922 23777 21309 33905 29910 23767 21299 R2 0.45 0.47 0.48 0.48 0.45 0.47 0.48 0.48 Note: P-values are in parentheses. p*<0.1, p**<0.05, p***<0.01 (robust standard errors corrected for clustering by country- sector in parentheses). All regressions include country-sector, subnational region and year fixed effects. Appendix Table 5.3: Female labor share difference of trading vs. non-trading manufacturing firms, 2010-2019 vs. 2020-2024, OLS Dependent 2010-2019 2020-2024 variable: (1) (2) (3) (4) (5) (6) (7) (8) fem_shisrt exp imp gvc fdi exp imp gvc fdi lnYisrt -1.128*** -0.928*** -1.030*** -0.917*** 0.0199 0.292** 0.0989 0.328** (0.000) (0.000) (0.000) (0.000) (0.884) (0.030) (0.464) (0.014) lncapintisrt -0.613*** -0.536*** -0.593*** -0.565*** -0.240 -0.176 -0.218 -0.174 (0.002) (0.008) (0.003) (0.005) (0.145) (0.285) (0.186) (0.292) lntfpisrt 1.064 1.030 1.004 0.890 -0.925 -0.769 -0.847 -0.756 (0.103) (0.116) (0.124) (0.181) (0.125) (0.204) (0.161) (0.211) tradeisrt / 4.036*** 1.838*** 3.768*** 2.636*** 5.218*** 1.537*** 5.345*** 1.388 globalisrt (0.000) (0.005) (0.000) (0.003) (0.000) (0.009) (0.000) (0.163) Obs. 8236 8120 8236 8226 13072 13023 13072 13072 R2 0.48 0.48 0.48 0.48 0.47 0.46 0.46 0.46 Note: P-values are in parentheses. p*<0.1, p**<0.05, p***<0.01 (robust standard errors corrected for clustering by country- sector in parentheses). All regressions include country-sector, subnational region and year fixed effects. 41 Appendix Table 5.4: Female wage share difference of trading vs. non-trading manufacturing firms and global vs. non-global manufacturing firms, reduced country sample, OLS Dependent Total Dependent Total variable: (1) (2) (3) (4) variable: (5) (6) (7) (8) exp imp gvc fdi fem_shisrt exp imp gvc fdi lnYisrt -0.812*** -0.257 -0.647*** -0.337 lnYisrt -0.712*** -0.127 -0.552** -0.178 (0.000) (0.224) (0.002) (0.113) (0.004) (0.607) (0.027) (0.480) lncapintisrt -0.581** -0.415 -0.548** -0.448* lncapintisrt -0.740** -0.561* -0.709** -0.591* (0.030) (0.123) (0.041) (0.097) (0.016) (0.068) (0.021) (0.056) lntfpisrt -1.050 -1.089 -1.098 -0.995 lntfpisrt -2.171* -2.229* -2.233* -2.099* (0.261) (0.245) (0.241) (0.287) (0.061) (0.053) (0.054) (0.068) ln(wM/wF)cst -16.10*** -15.97*** -15.99*** -16.38*** (0.000) (0.000) (0.000) (0.000) tradeisrt / 7.557*** 0.774 6.546*** 2.426* tradeisrt / 8.164*** 1.037 7.338*** 2.230 globalisrt (0.000) (0.392) (0.000) (0.066) globalisrt (0.000) (0.310) (0.000) (0.120) Obs. 3517 3492 3517 3517 Obs. 3517 3492 3517 3517 R2 0.52 0.52 0.52 0.52 R2 0.48 0.47 0.48 0.47 Note: P-values are in parentheses. p*<0.1, p**<0.05, p***<0.01 (robust standard errors corrected for clustering by country- sector in parentheses) for the sample of 26 countries for which relative wage rates at the sector level are available. All regressions include sector, subnational region and year fixed effects Appendix Table 5.5: Female labor share difference of trading vs. non-trading manufacturing firms, role of female ownership and top manager, OLS Dependent Ownership Management variable: (1) (2) (3) (4) (5) (6) (7) (8) fem_shisrt exp imp gvc fdi exp imp gvc fdi lnYisrt -0.349*** -0.0729 -0.282** -0.0940 -0.371*** -0.0889 -0.289*** -0.111 (0.002) (0.509) (0.011) (0.393) (0.001) (0.401) (0.007) (0.296) lncapintisrt -0.405*** -0.313** -0.388*** -0.350*** -0.375*** -0.277** -0.357*** -0.319** (0.002) (0.018) (0.003) (0.008) (0.003) (0.031) (0.005) (0.013) lntfpisrt -0.159 -0.0701 -0.122 -0.131 -0.126 -0.0216 -0.114 -0.113 (0.733) (0.881) (0.793) (0.782) (0.779) (0.962) (0.799) (0.803) femaleisrt 0.0528*** 0.0773*** 0.0359** 0.0204 4.408*** 8.326*** 4.873*** 5.288*** (0.000) (0.000) (0.036) (0.507) (0.000) (0.000) (0.000) (0.002) tradeisrt / 4.150*** 0.695 4.326*** 1.508** 4.069*** 0.597 3.954*** 1.206* globalisrt (0.000) (0.138) (0.000) (0.046) (0.000) (0.183) (0.000) (0.084) Obs. 20052 19898 20052 20044 21284 21119 21284 21274 R2 0.48 0.48 0.48 0.48 0.48 0.48 0.48 0.48 Note: P-values are in parentheses. p*<0.1, p**<0.05, p***<0.01 (robust standard errors corrected for clustering by country- sector in parentheses). All regressions include country-sector, subnational region and year fixed effects. 42 Appendix 6: Fractional probit estimates, services firms Appendix Table 6.1: Female labor share difference of trading vs. non-trading services firms, fractional probit Dependent Exporter, exp Importer, imp variable: (1) (2) (3) (4) (5) (6) (7) (8) fem_shisrt lnYisrt -0.000922 -0.00122 0.00241 -0.00183 -0.00211 0.00138 (0.413) (0.454) (0.447) (0.174) (0.198) (0.666) lntechisrt -0.0221** -0.0237** -0.0296*** -0.0311*** (0.029) (0.026) (0.006) (0.006) lnlpisrt -0.00599 -0.00581 (0.210) (0.238) globalisrt -0.00963 -0.0105 -0.0446*** -0.0451*** 0.00655* 0.00560 0.00441 0.00446 (0.227) (0.185) (0.001) (0.001) (0.096) (0.198) (0.466) (0.460) Obs. 33210 29618 14658 14658 25477 23512 13860 13860 Pseudo-R2 0.36 0.37 0.32 0.32 0.37 0.38 0.31 0.31 Appendix Table 6.2: Female labor share difference of global vs. non-global services firms, fractional probit Dependent GVC participant, gvc Foreign ownership, FDI variable: (1) (2) (3) (4) (5) (6) (7) (8) fem_shisrt lnYisrt -0.000972 -0.00147 0.00196 -0.00138 -0.00212 0.00120 (0.382) (0.360) (0.535) (0.223) (0.200) (0.707) lntechisrt -0.0226** -0.0241** -0.0250** -0.0264** (0.028) (0.026) (0.016) (0.016) lnlpisrt -0.00566 -0.00548 (0.237) (0.258) globalisrt -0.0198** -0.0241** -0.0504*** -0.0511*** 0.00795 0.00680 0.00655 0.00580 (0.039) (0.016) (0.000) (0.000) (0.108) (0.206) (0.465) (0.523) Obs. 33210 29618 14658 14658 33210 29618 14658 14658 Pseudo-R2 0.36 0.37 0.32 0.32 0.36 0.36 0.32 0.32 Appendix Table 6.3: Female labor share difference of trading vs. non-trading services firms, 2010-19 vs. 2020-24, fractional probit Dependent 2010-2019 2020-2024 variable: (1) (2) (3) (4) (5) (6) (7) (8) fem_shisrt exp Imp gvc fdi exp imp gvc fdi lnYisrt 0.000774 -0.000116 -0.000106 -0.000694 0.00254 0.00142 0.00227 0.00156 (0.885) (0.983) (0.984) (0.898) (0.520) (0.713) (0.560) (0.693) lntechisrt -0.0656*** -0.0911*** -0.0720*** -0.0760*** -0.00927 -0.0124 -0.00913 -0.0101 (0.000) (0.000) (0.000) (0.000) (0.434) (0.310) (0.434) (0.391) lnlpisrt -0.0119 -0.0132 -0.0113 -0.0118 -0.00349 -0.00313 -0.00332 -0.00293 (0.108) (0.110) (0.127) (0.127) (0.549) (0.586) (0.568) (0.616) tradeisrt / -0.123*** -0.00451 -0.125*** 0.00288 -0.0181 0.00773 -0.0205* 0.00596 globalisrt (0.000) (0.688) (0.000) (0.851) (0.135) (0.277) (0.091) (0.567) Obs. 4345 3821 4345 4345 10313 10039 10313 10313 Pseudo-R2 0.07 0.05 0.06 0.06 0.20 0.21 0.19 0.19 Note: The table presents average marginal effects. P-values are in parentheses. p*<0.1, p**<0.05, p***<0.01 (robust standard errors corrected for clustering by country-sector in parentheses). All regressions include country-sector, subnational region and year fixed effects. The dependent variable now ranges from 0-1 instead of 0-100. 43 Appendix Table 6.4: Female wage share difference of trading vs. non-trading services firms and global vs. non-global services firms, reduced country sample, fractional probit Dependent Total Dependent Total variable: (1) (2) (3) (4) variable: (5) (6) (7) (8) exp imp gvc fdi fem_shisrt exp imp gvc fdi lnYisrt -0.0102 -0.00948 -0.0110 -0.0108 lnYisrt -0.0143* -0.0147* -0.0155* -0.0160** (0.281) (0.257) (0.220) (0.241) (0.078) (0.071) (0.059) (0.049) lntechisrt 0.0244* 0.0217 0.0235* 0.0220 lntechisrt 0.0270* 0.0244 0.0259* 0.0249* (0.070) (0.138) (0.084) (0.112) (0.069) (0.141) (0.075) (0.074) lnlpisrt 0.0125 0.0115 0.0130 0.0125 lnlpisrt 0.0168* 0.0165* 0.0175** 0.0176** (0.206) (0.190) (0.173) (0.196) (0.060) (0.058) (0.049) (0.046) ln(wM/wF)cst -0.0388 -0.0468 -0.0421 -0.0451 (0.650) (0.400) (0.566) (0.461) tradeisrt / -0.0188 0.0148* 0.0216 0.0353* tradeisrt / -0.0311 0.0154 0.0136 0.0393** globalisrt (0.523) (0.086) (0.248) (0.010) globalisrt (0.345) (0.121) (0.547) (0.018) Obs. 3796 3642 3796 3796 Obs. 3796 3642 3796 3796 Pseudo-R2 0.27 0.26 0.27 0.27 Pseudo-R2 0.24 0.24 0.24 0.24 Appendix Table 6.5: Female labor share difference of trading vs. non-trading services firms, role of female ownership and top manager, fractional probit Dependent Ownership Management variable: (1) (2) (3) (4) (5) (6) (7) (8) fem_shisrt exp imp gvc fdi exp imp gvc fdi lnYisrt 0.00195 0.00324 0.00150 0.00114 0.00265 0.00235 0.00198 0.00150 (0.539) (0.333) (0.635) (0.719) (0.404) (0.493) (0.530) (0.637) lntechisrt - -0.0254** -0.0308*** -0.0263** 0.0294** -0.0234** -0.0293*** -0.0241** -0.0269** (0.024) (0.005) (0.021) (0.010) (0.030) (0.009) (0.028) (0.014) lnlpisrt -0.00562 -0.00682 -0.00538 -0.00539 -0.00627 -0.00562 -0.00572 -0.00582 (0.243) (0.165) (0.265) (0.268) (0.190) (0.250) (0.232) (0.228) femaleisrt 0.104*** 0.181*** 0.173*** 0.143*** 0.131*** 0.161*** 0.143*** 0.159*** (0.000) (0.000) (0.000) (0.010) (0.000) (0.000) (0.000) (0.001) tradeisrt / -0.0659*** -0.0373*** -0.0834*** -0.0144 -0.0752*** -0.0337*** -0.0844*** -0.0195* globalisrt (0.000) (0.000) (0.000) (0.185) (0.000) (0.000) (0.000) (0.051) Obs. 14238 13484 14238 14238 14632 13835 14632 14632 Pseudo-R2 0.32 0.34 0.32 0.32 0.33 0.34 0.32 0.32 Note: The table presents average marginal effects. P-values are in parentheses. p*<0.1, p**<0.05, p***<0.01 (robust standard errors corrected for clustering by country-sector in parentheses). All regressions include country-sector, subnational region and year fixed effects. The dependent variable now ranges from 0-1 instead of 0-100. 44