Does FDI have a social demonstration effect in developing economies? Evidence based on the presence of women-led local firms

We hypothesize that foreign direct investment (FDI) benefits female entrepreneurs in developing economies through a “social demonstration effect,” namely, by exhibiting norms and practices supporting gender equality and promoting women’s role in business. Results based on data from 44,418 firms in 91 developing economies show that, at the country level, increased employment share of foreign invested firms has a positive association with women’s entrepreneurship, namely, it boosts the presence of female leaders in new ventures and small businesses, as foreign firms help break local conventions unfavorable to female entrepreneurs. This positive impact is more evident when women face greater institutional barriers. However, at the industry level, foreign employment share becomes nonsignificant, and it has a negative association with the presence of female entrepreneurs in countries where women face greater barriers. We reason that this is due to occupational competition: The more disadvantaged women are in a country, the greater the advantage foreign firms possess in attracting women to work for foreign firms instead of pursuing their own entrepreneurial opportunities. The practical implication of our study is that governments can reformulate FDI policies with a view to fostering women’s entrepreneurship.


INTRODUCTION
Foreign direct investment (FDI) has important implications for local firms in developing economies. Whereas multinational enterprises (MNEs) generally possess superior technology and may crowd out local firms due to the technology gap between foreign and local firms, they may also benefit local firms through technology spillovers (Aitken & Harrison, 1999;Caves, 1974;Driffield & Munday, 2000;Feinberg & Majumdar, 2001;Haddad & Harrison, 1993;McGaughey, Raimondos, & la Cour, 2020;Zhang, Li, & Li, 2014). Extant research suggests that a major mechanism of FDI spillovers is the (technology) demonstration effect, through which local firms observe and acquire information on the technology innovations displayed by MNEs (Blomström & Kokko, 1998;Meyer, 2004).
Meanwhile, research in international business suggests that MNEs not only compete on their technological supremacy but also exploit their social and institutional advantages vis-à-vis local firms (Regnér & Edman, 2014;Shi & Hoskisson, 2012). In a recent study, Siegel, Pyun, and Cheon (2019) find that MNEs take advantage of social schisms in their host-country communities and purposefully hire and promote women -a socially excluded group in local businesses -to managerial leadership positions. By so doing, MNEs are able to establish an outsider's advantage over local firms, as the latter are often confined by backward social norms and business practices. We suggest that a meaningful extension of their study, and of the FDI spillovers literature, is to address the following research question, which arises logically from both streams of research: Does FDI have a social demonstration effect in developing economies for women's role in business? By social demonstration effect, we mean the diffusion of new social norms to the local business community to prompt important changes in business practices. Just as local firms can benefit from technology spillovers from MNEs, we wonder whether they could also benefit from a ''social spillover effect.'' To answer our research question, we examine the impact of FDI on women's entrepreneurship in developing economies around the world. We define women's entrepreneurship in terms of the presence of female leaders/owners of new ventures and small businesses. Entrepreneurial endeavor requires not only business resources but also social capital and institutional support (Ellis, 2011;Stephan, Uhlaner, & Stride, 2015;Valdez & Richardson, 2013). Because women overall are economically and socially disadvantaged in society -and even more so in developing economies -their entrepreneurial activity is typically at a lower level than that of their male counterparts (Fang, Goh, Roberts, Xu, & Zeufack, 2022;Jennings & Brush, 2013;Minniti, Arenius, & Langowitz, 2005). We are interested in examining whether the presence of FDI in a host country is positively associated with the presence of female entrepreneurs, as evidence for MNEs' social demonstration effect in developing economies, and whether such an association is contingent on the host countries' institutional conditions for women.
We focus on women's entrepreneurship for our investigation not only because it is a good example of women's role in business but also because prior studies have identified FDI as a significant determinant of the intensity of entrepreneurship in host countries (De Backer & Sleuwaegen, 2003;Goel, 2018;Kim & Li, 2014). Just as FDI can have either a positive or negative impact on local firms in general, so too it can be a double-edged sword for local entrepreneurial firms. On the one hand, there is evidence that FDI fosters entrepreneurial activity in host countries through spillovers (Kim & Li, 2014). On the other hand, FDI may crowd out local entrepreneurs in FDI-concentrated industries through occupational competition between employed jobs with MNEs and entrepreneurship (Danakol, Estrin, Reynolds, & Weitzel, 2017;De Backer & Sleuwaegen, 2003). By shifting the focus to female entrepreneurs, we aim to provide a more complete picture of the effects of FDI on domestic entrepreneurship, in addition to achieving our primary objective, namely gauging the social demonstration role of MNEs.
We tested our hypotheses and obtained confirmatory results using a dataset covering 44,418 local firms in 91 developing economies, with complementary information on the institutional environment women face. Our study contributes to the literatures on FDI spillovers and women's entrepreneurship by articulating the social demonstration effect and identifying, possibly for the first time, a robust empirical association between the presence of FDI in a host country and the presence of women-led local firms. Our study also contributes to these literatures with results suggesting that FDI may crowd out female entrepreneurs through the occupational substitution effect in FDI-concentrated industries. Our findings not only provide early evidence for the social demonstration effect of FDI but also open the door to further research on other manifestations of the social demonstration effect in host countries. the technology demonstration effect, whereby MNEs showcase their advanced technologies and organizing principles to local firms; vertical and horizontal linkages, through which local firms learn from MNEs as buyers, suppliers, and distributors; and employee turnovers, which facilitate knowledge transfer from foreign to local firms (Caves, 1974;Meyer, 2004;Spencer, 2008).
On the other hand, some researchers have shown that FDI has no or a negative effect on local firms (Aitken & Harrison, 1999;De Backer & Sleuwaegen, 2003;Feinberg & Majumdar, 2001;Haddad & Harrison, 1993;Konings, 2001), which supports the conjecture that foreign firms might crowd out local firms in developing economies due to their technological advantages (Blomström, Kokko, & Zejan, 2000;Caves, 1996). Generally, FDI tends to generate net positive spillovers at a more macro level, namely, across industries or regions in a country, and net negative spillovers are more likely at a micro level, namely, within an industry or region (Blalock & Simon, 2009;Chang & Xu, 2008;Javorcik, 2004;Keller, 2002). This is because crowding-out typically occurs among firms competing in the same industry or for the same regional market, whereas the demonstration effect can be more farreaching. A well-known example is that the first four special economic zones in China, established in Shenzhen and other cities in the early 1980s, served as effective showrooms of advanced foreign technologies and management practices for the whole country and played an important role in the improvement of Chinese firms' productivity across the board (Chang & Xu, 2008).
Given these differing spillover effects of FDI, it is necessary to also discuss the social demonstration effect at the country and industry levels separately. At the country level, much of the advanced knowledge that MNEs bring to a developing economy, especially that pertaining to managerial expertise and organizational practices as opposed to pure technological knowhow, is not industryspecific and can benefit all local firms (Meyer, 2004;Spencer, 2008). In a similar vein, FDI may also bring, demonstrate, and diffuse new social norms to local populations and firms across the board. We suggest that MNEs, the majority of which are from developed economies, tend to exhibit activities and practices that are consistent with the laws and social norms of their home countries. For instance, prior studies suggest that MNEs from less corrupt home countries can better resist the pressure to engage in corruption in host countries (Spencer & Gomez, 2011), and thus can help reduce the corruption level of local firms (Kwok & Tadesse, 2006).
Conceivably, among the norms and principles that MNEs bring to host countries are those that endorse gender equality and promote women's role in business to a greater extent than local customs do. Although one may argue that gender inequity exists in most countries around the world, it is probably more severe outside the developed economies due to poorer legal frameworks and more constraints in financial and educational resources for women there (Panda, 2018). Both anecdotal accounts (Shi & Hoskisson, 2012) and case studies (Song, 2021) suggest that MNEs may take advantage of local prejudices against women by hiring qualified women who otherwise cannot find decent jobs with local employers. Empirical research further shows that hiring and promoting local female managers indeed can lead to higher profitability (Siegel et al., 2019). Even if an MNE's own home country does not have the strongest social norms that support women's empowerment, the firm is likely to have been exposed to these new norms because of its multinationality, and advocate these norms in its practice in order to stay competitive in overseas markets (Siegel et al., 2019). Once MNEs demonstrate to local audiences in developing economies that women can be competent firm leaders and contribute to firm success, such practices will quickly spread to local firms across industries because, first, they help break the local customs unfavorable to women in business, and second, they also heighten women's self-confidence in conducting business.
By extension, we expect that women's leadership role in entrepreneurial firms will also be greatly enhanced by MNEs' actions in developing economies in two ways. First, local female entrepreneurs may benefit indirectly from MNEs' employment and promotion policies, because the same local traditions and prejudices that make it difficult for women to play important roles in the workplace also prevent them from leading or owning entrepreneurial ventures. For instance, if women face barriers to employment and promotion because local employers believe women are less competent than men (Oakley, 2000;Ughetto, Rossi, Audretsch, & Lehmann, 2020;Yang & Triana, 2019), then the same local employers will likely refrain from dealing with female entrepreneurs as buyers, suppliers, lenders, and investors. By promoting women to managerial leadership positions, MNEs can dispel the misperceptions that women leaders are less productive (Fischer, Reuber, & Dyke, 1993) and less effective (Paustian-Underdahl, Walker, & Woehr, 2014).
Second, MNEs may extend their women-empowering norms and practices to local female entrepreneurs directly by transacting with them or choosing them as preferred suppliers and business partners. For instance, MNEs can promote the participation of women in supplier training and capacity-building programs (UNCTAD, 2021). Such actions will likely enhance the legitimacy, acceptance, and status of women-led entrepreneurial firms and create a more munificent environment for their entry and survival. This will then produce more spillovers as local firms, especially those in MNEs' supplier networks, follow MNEs' lead to work with female entrepreneurs because MNEs are their important buyers and collaborators. By converging with MNEs' standards and practices, local firms can expect to improve their legitimacy and image in the eyes of the MNEs and other audiences (Guler, Guillén, & Macpherson, 2002;UNCTAD, 2021). Local firms may also imitate the practices as a response to MNEs' strategic initiatives because MNEs may have supported women-led firms as a means of creating their local network advantages over their host country competitors (Liu, 2011). An additional consequence of these actions and reactions is that, seeing the improved prospects of women-led firms, capable male managers and employees will be more willing to work for female entrepreneurs, further strengthening the latter's leadership position and their firms' viability. Combining these indirect and direct benefits for female entrepreneurs, we propose, Hypothesis 1a: In developing economies, FDI has a positive association with women's entrepreneurship at the country level.
In contrast, at the industry level, the social demonstration effect may yield less positive results for female entrepreneurs. FDI is typically unevenly distributed across industries, with some industries receiving more and some industries receiving less FDI. When an industry is populated by MNEs that support women's advancement, both local firms in that industry and those in other industries may follow suit to recruit and promote female managers; and more women with the necessary knowledge and skills will be encouraged to take part in managerial activities, both in that industry and in other industries. For entrepreneurial firms, however, MNEs' action may lead to differing consequences inside and outside that industry. Inside the industry, there may be competing occupational choices at play such that high-caliber women with industry-specific skills and experience are more attracted to the high pay and good benefits provided by MNEs rather than starting or running their own small businesses (Danakol et al., 2017;De Backer & Sleuwaegen, 2003). Because the industry has a high concentration of FDI, a substantial proportion of female talent will be absorbed by MNEs, leading to fewer qualified women to engage in entrepreneurial activity in that industry. In other words, whereas the social demonstration effect benefits female entrepreneurs both inside and outside an industry with a high presence of FDI, crowding-out mostly only occurs inside the industry and not outside the industry. 1 Although we do not have a theoretical basis to predict whether, inside the industry, the social spillover effect will be stronger than, equivalent to, or weaker than the crowding-out effect, it is quite probable that the net effect will be less positive at the industry level than it is at the national level where the demonstration effect dominates. Thus, we suggest, Hypothesis 1b: In developing economies, FDI has a less positive association with women's entrepreneurship at the industry level than it does at the country level.
Further, we suggest that the impact of FDI on women's entrepreneurship depends on the institutional conditions for women in a host country. For technology spillovers, there is a long debate between the technology gap hypothesis and the absorptive capacity hypothesis. The former argues that a substantial technology gap between MNEs and local firms is a prerequisite for spillovers to occur, whereas the latter claims that local companies require some degree of technological readiness to benefit from spillovers (Chang & Xu, 2008;Eapen, 2012). We suggest that a similar question may arise as to whether FDI benefits women to a greater extent in environments more favorable or more unfavorable to women, and we propose that the issue should again be considered at the country and industry levels separately.
At the country level, MNEs bring social norms that differ from local customs. In order to stay competitive, MNEs tend to adopt the most advanced business practices that are supported by these norms (Siegel et al., 2019). If the cultural values, societal traditions, and institutional rules of the host countries are also consistent with these norms, then conceivably, the practices of local firms should be as advanced as those of the MNEs, and there will be no social demonstration effect. It is in institutionally backward countries where the benefits of demonstration effect for women can be maximized. In these countries, women in general and female entrepreneurs in particular are discriminated against in society and unfairly treated by the local establishment because of cultural taboos, social prejudices, and asymmetrical economic conditions (Fang et al., 2022;Shahriar, 2018;Shinnar, Giacomin, & Janssen, 2012;Zhao & Yang, 2021). Particularly, local institutions often discourage women from joining the workforce or running their own ventures (Narula, 2019; Ribes-Giner, Moya-Clemente, Cervelló -Royo, & Perello-Marin, 2018; Song, 2021;Yang & Triana, 2019). As a result, talented women are excluded from good business and career opportunities. Under such circumstances, the MNEs with norms and practices that support women's career and social status will likely obtain a substantial competitive advantage in human resources and supplier networks. The social demonstration effect will take place as local firms imitate these practices and a population of qualified women are encouraged to pursue both managerial careers and venturing opportunities.

Journal of International Business Studies
At the industry level, however, direct competition between foreign and local firms, both for the market and for talent, is a more prominent feature. Extant literature suggests that FDI may crowd out local firms partly because MNEs enjoy various institutional advantages over local firms when dealing with institutional voids in host countries (Delios, Xu, & Beamish, 2008;Doh, Rodrigues, Saka-Helmhout, & Makhija, 2017;Regnér & Edman, 2014;Wang, Cui, Vu, & Feng, 2022). We argue that the more institutional barriers women face in a country, the more advantages MNEs possess in attracting female talent to work for MNEs in FDI-concentrated industries instead of pursuing their own entrepreneurial opportunities. Although FDI still has a social spillover effect for women in such industries, its benefit for women's entrepreneurship is negated by the attractiveness of working for MNEs as an occupational choice for women. The more concentrated FDI is in an industry, the more negative its impact becomes on women's entrepreneurship in that industry compared to other industries with a lower concentration of FDI, which all share the social demonstration effect but no or less crowding-out effect. Thus, we propose two contrasting interaction effects at the country and industry levels, with the institutional environment serving as the moderating factor in each of them.
Hypothesis 2a: In developing economies, FDI has a more positive association with women's entrepreneurship at the country level when the local institutional environment is more unfavorable to women.
Hypothesis 2b: In developing economies, FDI has a less positive or more negative association with women's entrepreneurship at the industry level when the local institutional environment is more unfavorable to women.

Data, Sample, and Dependent Variable
The main data source is the World Bank Enterprise Surveys (WBES) across 144 countries between 2006 and 2019, complemented by country-level data from other sources including the World Development Indicators (WDI), Doing Business dataset, and Women, Business and the Law dataset, which are all provided by the World Bank. The WBES data are collected by the World Bank to benchmark the business climate mainly in developing economies and to understand firm performance. The WBES team has adopted unified standards across countries in terms of questionnaire format, sample representativeness, and sampling methodology. In each country, the survey is based on the universe of eligible non-agricultural firms obtained from the country's census office using stratified random sampling with replacement, and the result is a representative sample of the private economy in the country. Stratification is based on two criteria: the sector of activity and firm size. The manufacturing and service sectors are the primary business sectors of interest, and some industries such as mining and financial services are excluded. Because information on an important control variable in our study, firms' research and development (R&D) activity, is only consistently available from 2010, the timeframe of our data is 2010-2019. This process resulted in an original dataset of 104,175 firm-year observations. From this original dataset, we removed firms in 18 developed or advanced economies to derive a sample purely from developing economies. 2 We also removed firms with 10% or more foreign ownership so that our sample represents domestic firms only (Iurkov & Benito, 2018).

Journal of International Business Studies
Research on entrepreneurship often associates entrepreneurship with new ventures and small businesses (Parker, 2018). Following extant literature, we only included new venture under 7 years of age (Boeker & Wiltbank, 2005;Shepherd & Wiklund, 2009;van Praag & Versloot, 2007) and small businesses with fewer than 100 employees (Elfenbein, Hamilton, & Zenger, 2010;Kang, Matusik, Kim, & Phillips, 2016;Li, Qian, & Qian, 2015) as our sample firms. We then gauged the presence of a female entrepreneur with two alternative measures. The first (referred to as female CEO) is a dummy variable coded as 1 if the chief executive officer (CEO) or highest-ranked manager is a woman, and 0 if otherwise. This step, after dropping observations that have missing values, yields a sample of 44,418 firm-year observations in 91 countries. The second (referred to as female CEO and controlling owner), also a dummy variable, is coded as 1 if the CEO is a woman and at least 50% of the firm's shares are ultimately controlled by women. Because of additional missing values for the genders of the controlling owners, the second sample consists of 40,265 firm-year observations in 81 countries. A list of these countries is provided in Table 3 in the Appendix.

Independent variables
We gauged the presence of FDI with the foreign employment share, the most commonly used measurement for FDI in the spillovers literature (e.g., Caves, 1974;Chang & Xu, 2008;Driffield & Munday, 2000). Foreign employment share in country is the share of all foreign firms in the total employment of a host country. Foreign employment share in industry is the share of all foreign firms in the total employment of an industry in the host country. To capture the relative presence of FDI in industries, we subtracted the value of foreign employment share in country from the original value of foreign employment share in industry, so that this independent variable reflects the presence of FDI at the industry level relative to that at the country level.
We used the original full sample of 104,175 observations to calculate these two variables. A foreign firm is a firm with at least 10% foreign ownership in its total equity, which is a commonly accepted definition (UNCTAD, 2016) and an ''established practice'' in research (Iurkov & Benito, 2018: 1043. Information on both employment and foreign ownership is from the WBES. Joint ventures, as well as wholly owned subsidiaries, are the immediate recipients of foreign technology (Aitken & Harrison, 1999). We adopted the 10% threshold because even in joint ventures where MNEs do not own a majority stake, the local partners will want to import the most advanced business practices to gain a competitive advantage over the purely domestic firms.

Moderating variables
We used two country-level indicators of the institutional environment that women face as moderators. Barriers to employment for women is based on respondents' answers to the following question: ''Can a woman work on an industrial job in the same way as a man?'' The dummy variable is coded as 1 if the answer is no, and 0 if otherwise. The data source is Women, Business, and the Law (wbl.worldbank.org). 3 Barriers to venturing for women is the number of bureaucratic procedures that women have to complete to start their own ventures. These include all the steps women have to take to obtain the necessary licenses and permits and complete any required notifications, verifications, or inscriptions for the company with relevant authorities. The data source is the Doing Business dataset (www.doingbusiness.org). 4

Control variables
We included a range of control variables at the firm, industry, and country levels as well as the industry and year fixed effects. Firm-level controls are firm age, in actual years of age; firm size, in terms of the number of permanent employees; largest share, the ownership share of the largest owner, which is a measure of ownership concentration; labor productivity, calculated as firm sales over the number of permanent employees, in logarithm; and firm R&D, a dummy variable that equals 1 if the firm has spent on R&D in the previous fiscal year and 0 otherwise. At the industry level, we included informal economy competition, the percentage of firms in industry competing against informal (i.e., unregistered) businesses; financial obstacle, the percentage of firms in industry claiming access to finance is an obstacle for them; and corruption obstacle, the percentage of firms in industry claiming corruption is an obstacle. At the country level, we included three variables obtained from the WDI dataset, and they are GDP per capita, in logarithm; national education, measured in terms of literacy rate of adults aged 15 or above; and urbanization rate, the ratio of urban population to total population.

Modeling
We adopted the following Probit model: where i, k, c, and t stand for firm i, industry k, country c, and year t, respectively. Z ckt represents the interactions between each of the two FDI measures and the two moderating variables, respectively. X ikct is a vector of controls at various levels. h k stands for the industry fixed effects, and l t is the year fixed effects. Since our data are largely cross-sectional, we do not control for firm fixed effects. The standard errors are cluster-adjusted at the country-industry-year level. Table 1 provides the descriptive statistics and correlations of variables. We checked variance inflation factors (VIFs) and found that the VIFs ranged from 1.02 to 4.52, well below the generally accepted cut-off point of 10. Therefore, multicollinearity is not a major concern for our models.

Hypothesis Tests
In Table 2, we present results pertaining to both measures of our dependent variable. Models 1 and 7 include all control and moderating variables. Models 2 and 8 enter the two independent variables, foreign employment share in country and foreign employment share in industry. 5 Models 3-6 and models 10-13 enter the various interaction terms between the independent variables and the two moderators, barriers to employment for women and barriers to venturing for women. Models 7 and 14 are the full-specification models. Because the two sets of results using the two different dependent variables are largely consistent, we mainly discuss the first set, pertaining to the female CEO measure of women's entrepreneurship. We discuss the results of hypothesis tests based on the full-specification model. In model 7, foreign employment share in country has a positive and significant effect (b = 0.631, p = 0.000) on the likelihood that a new or small firm's CEO is a woman. In terms of effect size, the marginal effect of foreign employment share in country estimated by Stata is 0.140, suggesting that on average, when foreign employment share in country increases by one standard deviation (i.e., 0.166, or about 72% of the mean), the probability of having a female entrepreneur increases by 0.023 (i.e., 0.166 9 0.140), or about 15% of the mean value of female CEO. This result provides support for hypothesis 1a that FDI has a positive association with women's entrepreneurship at the country level. In contrast, foreign employment share in industry is nonsignificant (b = -0.155, p = 0.157), and a Wald test shows that the difference between the coefficients of the two FDI variables is statistically significant (p = 0.000), which provides support for hypothesis 1b that FDI has a less positive association with women's entrepreneurship at the industry level than at the country level.
Hypothesis 2a predicts that FDI has a more positive association with women's entrepreneurship at the country level when women face a more unfavorable environment. In model 7, the interaction of foreign employment share in country and barriers to employment for women has a positive sign and is statistically significant (b = 0.792, p = 0.009). The interaction of foreign employment share in country and barriers to venturing for women also has a positive sign and is statistically significant (b = 0.080, p = 0.028). These results provide support for hypotheses 2a. We followed prior research (Jandhyala & Weiner, 2014;Schwens, Zapkau, Brouthers, & Hollender, 2018) to illustrate the marginal effect of foreign employment share in country by the moderators. Based on model 7, Figure 1a shows that, when the value of barriers to employment for women is 0 (which accounts for about 31% of the sample), the marginal effect is close to the horizontal, dashed zero-effect line, indicating a nonsignificant effect of FDI on women's entrepreneurship. 6 When the value of barriers to employment is 1, the lower bound of the 90% confidence interval line is far above the zero-effect line, showing a significant and positive relationship of country-level FDI to women's entrepreneurship. Figure 1b shows that, when the level of barriers to venturing for women is below 6 (which roughly accounts for 20% of the sample), the lower bound of the 90% confidence interval intersects with the zero-effect line, indicating that the effect of FDI on women's entrepreneurship is nonsignificant. When the level of barriers to venturing for women is over 6, the lower bound of 90% confidence interval rises above the zero-effect line, indicating that the effect of FDI becomes significantly positive.

Journal of International Business Studies
Hypothesis 2b predicts that FDI has a less positive or more negative association with women's entrepreneurship at the industry level when women face a more unfavorable environment. In model 7, the interaction of foreign employment share in industry and barriers to employment for women has a negative sign, but is nonsignificant (b = -0.179, p = 0.360). The interaction of foreign employment share in industry and barriers to venturing for women also has a negative sign as predicted, and is statistically significant (b = -0.053, p = 0.090). These results provide partial support for hypothesis 2b. Based on model 7, Figure 1c shows that when the level of barriers to venturing is below 10 (which roughly accounts for 55% of the sample), the 90% confidence interval intersects with the zero-effect line, indicating that the effect of industry-level FDI on women's entrepreneurship is nonsignificant. When the level of barriers to venturing is above 10, the upper bound of the 90% confidence interval falls below the zero-effect line, indicating that the effect of industry-level FDI becomes significantly negative.

Robustness and Endogeneity Checks
We conducted a range of robustness checks to assess the sensitivity of our results to alternative firm samples, variable measures, and model specifications. First, because we defined sample firms as firms younger than 7 years of age or smaller than 100 employees in size, and most of the firms are included due to their small size (accounting for 98.5% of the sample) and not their young age (accounting for 18.2% of the sample), we adopted alternative size criteria to derive the sample. Specifically, we included firms of fewer than 20, 50, 150, and 200 employees, respectively, as well as firms of all sizes (that is, effectively removing both age and size constraints). Results show that when firms of up to 150 employees are included, interaction between foreign employment share in industry and barriers to venturing for women turns nonsignificant on female CEOs. And when the maximum employment size reaches 200, the interaction term becomes nonsignificant on both dependent        variables. These results suggest that, whereas other hypotheses are largely upheld across different employment size criteria, hypothesis 2b loses support when larger firms are included in the sample. It seems that occupational substitution between employed jobs with MNEs and entrepreneurship, which underlies hypothesis 2b, only applies to smaller entrepreneurial firms, probably because for a woman to run and own a large business has much greater financial and intrinsic rewards than to work for an MNE. Second, instead of using 10% foreign ownership as a threshold for identifying foreign invested firms, we adopted a stricter criterion and defined foreign firms as firms with 50% or more foreign ownership when calculating the foreign employment share. Third, we substituted foreign employment share in country with an alternative variable, FDI net inflow to country, obtained from the WDI dataset. Fourth, to rule out the possibility of omitting relevant country-level variables driving our results when estimating the effect of foreign employment share in country, we included a range of country-specific factors such as the population density, infrastructure, gender-related social norms, the form of the government, government effectiveness, the rule of law, intellectual property protection, and regulatory quality. We entered these variables one by one as well as collectively. Fifth, we adopted an alternative model specification by using the logit model. Sixth, we clustered the standard errors at country-industry level. Seventh, since some researchers are concerned about the validity of computing and interpreting interaction effects in nonlinear models, we used the Stata ''inteff'' procedure (Marquis & Qian, 2014;Norton, Wang, & Ai, 2004) to verify the interaction effects produced by the Probit model. Throughout these robustness checks, both the main FDI effects and the FDI effects as moderated by institutional barriers for women are largely consistent with those reported in Table 2.
Eighth, certain host country factors may potentially lead to both more inward FDI and more female business leaders simultaneously, which can bias the estimate of the effect of FDI on women's entrepreneurship. To address such potential endogeneity for the country-level FDI, we relied on an instrumental variable measuring pre-existing contract enforcement in host countries for our country-level independent variable. Contract enforcement is important for host countries to attract FDI (Markusen, 2001), but it should be gender-neutral for domestic businesses. Specifically, contract enforcement is measured as the natural logarithm of the average days from filing of the lawsuit in court to the final judgment in a country. The data are obtained from the Doing Business dataset. Rerunning models 2 and 9 in Table 2 with this instrumental variable, the firststage results show that contract enforcement significantly predicts foreign employment share in country. Using the Stata ''weakiv'' procedure, the result of Anderson-Rubin test suggests that we do not have a weak-instrument issue (v 2 = 46.43, p = 0.000). 7 The second-stage results show that the instrumented foreign employment share in country remains positive and significant on both dependent variables (b = 2.048, p = 0.008; b = 2.721, p = 0.006, respectively).
Ninth, Eapen (2013) suggests that when examining FDI spillover effects in incomplete datasets, the under-representation of small firms raises the selection bias issue, and he recommends using a weighted instrumental variables approach, in which the weight can be the inverse of firm employment size. Following this approach, we adopted the inverse of firm total employment size and permanent employment size, respectively, as the weights. The results exhibit similar estimates of foreign employment share in country to those based on the unweighted instrumental variable approach, suggesting that the issue of under-representation of small firms (or firms of other sizes) is not severe in our sample.

Additional Analysis
Prior research has examined the temporal effects of FDI in host countries (Aitken & Harrison, 1999;Kosová, 2010). Our cross-sectional data structure does not allow us to conduct such analyses formally. However, we attempted some temporaleffect analyses using the alternative independent variable we used in robustness tests, FDI net inflow to country. Data on this variable are available annually from the WDI dataset. We estimated our baseline model specification (models 2 and 9 in Table 2), substituting lagged values of FDI net inflow to country for foreign employment share in country. We included lags of up to 7 years. Results show that FDI has positive effects on the two dependent variables in 4 and 6 of the 7 years, respectively ( Table 4 in the Appendix). Then, in another analysis, we calculated the mean values of lagged FDI net inflows to country in the previous 2, 3, 5, and 7 years, respectively, and used these mean lagged values to predict the dependent variables. Results show that the effect of FDI on women's entrepreneurship is positive and significant for all these different time lags (Table 5 in the  Appendix).
Prior research also suggests that female entrepreneurs tend to concentrate in certain industries (Bardasi, Sabarwal, & Terrell, 2011). We therefore conducted a subgroup analysis by separating manufacturing-and service-sector samples to see if the impact of FDI differs across these two sectors. The mean values of female CEO are 0.138 for manufacturing and 0.177 for service, and the mean values of female CEO and controlling owner are 0.087 and 0.120, respectively. The between-sector differences in means are statistically significant (b = -0.039, p = 0.000; b = -0.033, p = 0.000). Rerunning our main-effect regression models on the two samples, the results show that at the country level, foreign employment share has a positive effect on the presence of female CEO in both sectors (b = 0.622, p = 0.002 for manufacturing; b = 0.557, p = 0.015 for service). Yet, on the presence of female CEO and controlling owner, foreign employment share is only significant in the service sector (b = 0.348, p = 0.128; b = 0.540, p = 0.020). At the industry level, foreign employment share has a negative impact on both female CEO and female CEO and controlling owner in the manufacturing sector (b = -0.236, p = 0.091; b = -0.261, p = 0.084), but is nonsignificant in the service sector (b = 0.078, p = 0.651; b = 0.244, p = 0.174). 8 It appears that FDI benefits female entrepreneurs more in the female concentrated service sector, and it tends to crowd out female entrepreneurs in the manufacturing sector.

DISCUSSION
Using a worldwide firm-level dataset, we examine how FDI affects women's entrepreneurship in developing economies around the world, with contrasting findings at country and industry levels. At the country level, our results show a significant positive association between FDI and the likelihood that local firms are led by women; these results are robust across alternative variable measures, model specifications, and time periods, and after having employed weighted and unweighted instrumental variable checks. We find that this positive effect of FDI is stronger in countries with a more unfavorable institutional environment for women. However, at the industry level, the effect of FDI becomes nonsignificant; moreover, it appears that the association turns negative, in other words, FDI crowds out local firms led by women, when women face greater barriers to venturing in the corresponding host country. Our study has several important research implications.
First, our study contributes to research on FDI spillovers by explicitly proposing and providing evidence for a social demonstration effect of FDI at the country level. Such a social demonstration effect of FDI has only been briefly alluded to in extant research. For instance, Kwok and Tadesse (2006) suggest that the presence of FDI may lead to decreasing corruption in host countries because MNEs demonstrate to local firms how a cleaner way of conducting business is viable. Yet, the corruption case may be controversial because other researchers find that FDI can increase local corruption levels (Robertson & Watson, 2004). Social demonstration is a plausible explanation for our major finding, namely, the across-the-board positive effect of FDI on female entrepreneurs, because women are less likely than men to conduct business in technologybased sectors (Bardasi et al., 2011), and hence less likely to benefit from the technology demonstration effect of FDI. Findings from our additional analysis in separate sectors corroborate this argument by showing a more consistent positive effect of FDI on female entrepreneurs in the female concentrated service sector. Moreover, previous research suggests that FDI should be expected to crowd out informal economy, which is closely associated with women and their firms (Narula, 2019). Therefore, it is reasonable to attribute the positive relationship we have found between FDI and women-led local firmsas opposed to those led by men -to the spillovers of new social norms from MNEs. Second, we contribute to the emerging literature on the relationship between FDI and entrepreneurship by focusing on women's entrepreneurship. Whereas the effects of FDI on entrepreneurship in general are mixed and inclusive (De Backer & Sleuwaegen, 2003;Kim & Li, 2014), the impact of FDI on women's entrepreneurship is even more difficult to assess because female entrepreneurs tend to be at a disadvantage vis-à-vis their male counterparts. Whether such a disadvantage means FDI will have a stronger crowding-out effect on female entrepreneurs cannot be taken for granted, because it is also possible that FDI may create a more munificent environment for all firms (Li, 2008) or reduce the negative impact of underdeveloped local institutions on businesses through spillovers (Kim & Li, 2014;Meyer, 2004). A prior study that has investigated the association between FDI and women's entrepreneurship did not find a significant correlation between net FDI inflows to a country and the country's score on a Female Entrepreneurship Index, although it did find that FDI had a negative effect on a country's score on the General Entrepreneurship Index (Goel, 2018). Our study thus fills an important gap in this research.
Third, our study highlights the critical roles of the local institutional environment that women face in determining the social demonstration effect of FDI for women's entrepreneurship. The different results for the main effects of FDI at the country and industry levels, and the contrasting interaction effects of FDI and institutional barriers for women at these levels, indicate that the social demonstration effect, just like technology demonstration effect, has contingent values for local firms and local entrepreneurs. In particular, our findings suggest that FDI indeed can serve as a double-edged sword. On the one hand, our results indicate that institutionally underdeveloped countries, which are usually understudied in extant research (Fainshmidt, Judge, Aguilera, & Smith, 2018), are perhaps the biggest beneficiaries of FDI spillovers. Since entrepreneurship is at the core of capitalism, FDI, with its linkage to women's entrepreneurship, plays a critical role in fomenting capitalism with ''varieties of institutional systems'' (Fainshmidt et al., 2018: 307). On the other hand, within these countries, some female entrepreneurs may be negatively impacted, depending on the industry distribution of MNEs. The situation of these female entrepreneurs is perhaps even more understudied Our study delivers several messages that have practical implications for governments and female entrepreneurs in developing economies. First, FDI may potentially bring important changes in social norms and business practices, and so governments aimed at effecting such changes could use FDI as a tool. Second, more specifically, governments may be able to leverage FDI as a social-economic force to foster domestic women's entrepreneurship, especially in the service sector where women are concentrated. Third, however, if FDI is heavily concentrated in some industries, women's entrepreneurship in those industries could be depressed, and therefore governments should carefully balance the industry distribution of FDI to avoid such consequences, especially in the manufacturing sector. Fourth, for women with an ambition to start and run their own businesses, if the country's institutional environment is particularly hostile to women, they had better not enter industries where FDI is heavily concentrated, unless they hope to gather some industry experience in MNEs first.

Journal of International Business Studies
Our study also has a few limitations. First, female entrepreneurs were identified in terms of female leaders/owners of new ventures and small businesses, not based on their entrepreneurial activity, as we have traded depth for breadth in this study. As such, our dependent variable reflects the combined rates of both new entry and survival of female entrepreneurs. Although our theory is that FDI has positive effects on both, we have not empirically differentiated these effects, Second, as Eapen (2013) points out, the WBES may represent an incomplete dataset, based on which correctly identifying spillover effects might be difficult. Therefore, despite our effort to address potential identification problems through the weighted instrumental variable approach, the interpretation of our results requires caution. Third, our findings -though having gone through some endogeneity checks and often couched in terms of ''effects'' for ease of exposition -represent associations and not causality.
Despite these limitations, we hope that our study constitutes a useful, early attempt to detect the social demonstration effect of FDI and to investigate the relationship between FDI and women's entrepreneurship around the world, and that the findings form a coherent narrative of how FDI may affect domestic female entrepreneurs and what roles the local institutional environment can play in making FDI work better for developing economies. Future research should continue to investigate other forms of social spillovers from FDI in host countries. no. 71902195). Jingyu Gao (currently at Asian Infrastructure Development Bank) helped with the datasets early in this project.

FUNDING
Open Access funding enabled and organized by CAUL and its Member Institutions. NOTES a ''Review of Data Irregularities in Doing Business'' document (https://thedocs.worldbank.org/en/doc/ 791761608145561083-0050022020/original/ DBDataIrregularitiesReviewDec2020.pdf) and provided corrected data on the countries concerned. None of the country-year data at issue was included in or affected our analysis.

5
To save space, we did not enter the two independent variables separately. But if we had done that, the results would have been consistent with those presented here.

6
In Table 2, all the independent variables and moderator are mean-centered, so that the main effects of independent variables are comparable across models. However, when plotting the figures, the original values of these variables were used. 7 We also tested the validity of the instrument by conducting linear first stage regression. The instrumental variable passed both the under-identification test (Kleibergen-Paap rk LM statistic = 8.463, p = 0.004) and the weak-identification test (Cragg-Donald F statistic = 1929.463, above the rule of thumb of 10).
8 Detailed results of this additional analysis and the robustness/endogeneity checks are available upon request from the first author.

APPENDIX
See Tables 3, 4, 5.    Heba Shams is Vice President for global public policy and head of multilateral engagement at Mastercard, and a Fellow of Mastercard Policy Center for the Digital Economy. She was formerly lead investment policy specialist at the World Bank. Her research interests include payment, digital policy, and financial and digital inclusion.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Accepted by Becky Reuber, Area Editor, 21 December 2022. This article has been with the authors for three revisions.