How institutions and gender differences in education shape entrepreneurial activity: a cross-national perspective

Various definitions and forms of entrepreneurship exist (Acs et al. 2014). For example, Schumpeter views entrepreneurs as innovators whose function is to carry out new combinations of means of production. According to Knight’s (1982) seminal writings, an entrepreneur is someone who makes decisions under conditions of uncertainty. Estrin et al. (2013, p. 412) argue that entrepreneurship—“new entry” during efforts to create a viable business—results from an individual’s occupational choice to work on his or her own account.

In this study, we follow Wennekers and Thurik (1999, p. 29), who describe entrepreneurship as an ill-defined and, at best, multidimensional concept that requires decomposition at different levels. They argue that two major stages of entrepreneurship can be identified. The first has to do with “new entry” and the second with “innovativeness” in general. As a result, later research began to make a distinction between different stages of entrepreneurial activity (Reynolds et al. 2005; Baumol and Blinder 2011; Henrekson and Sanandaji 2014). Here, we concentrate on three different stages of entrepreneurial activity. In the first stage, we focus on entrepreneurial awareness, that is, whether an entrepreneur sees an opportunity to start a business. In the second stage, we focus on the sector in which the entrepreneur starts a business, as some sectors are more innovative and “entrepreneurial” than others (Wennekers and Thurik 1999; Marques 2017). In the third stage, we examine entrepreneurs’ growth aspirations (Estrin and Mickiewicz 2011). As such, our strategy in defining entrepreneurship is similar to Dilli et al. (2018). An important motivation to study the role of education in different stages of entrepreneurial activity is Van Der Zwan et al.’s (2011) argument that cross-country gender differences are largest during the conversion of start-up considerations into start-up activities and in business survival rates.

A large body of literature shows that education benefits an entrepreneur’s performance in different ways, such as business survival, firm growth, or the firm’s return on investment (e.g., Van Der Sluis et al. 2008; Millán et al. 2014). For instance, at the national level, De Clercq et al. (2008) find a positive effect of tertiary education on the GEM’s TEA rate. At the individual level, education can enhance managerial ability, which increases the probability of entrepreneurship. However, higher levels of education may also generate better outside options (i.e., more lucrative wage employment under better working conditions) and thus decrease the likelihood of entrepreneurship as the preferred choice (Van Der Sluis et al. 2008, p. 798). Empirical findings confirm this indeterminate effect of education level on advancement in the entrepreneurial process (see Van Der Zwan et al. 2013 for a review of the literature).

Studies that have considered the role of gender in entrepreneurship (e.g., Van Der Zwan et al. 2011; Caliendo et al. 2015; Stefani and Vacca 2015) also show that lower levels of female education are a crucial factor in explaining the gender differences in entrepreneurial activity. However, Fig. 1 shows that this link between education and the gender gap in entrepreneurial activity, captured here in terms of self-employment, is not always straightforward. Since the 1980s, the gender gap in tertiary education has closed substantially and even reversed in some industrialized countries, such as Portugal and Ireland. However, despite this progress toward gender equality, the gender gap in self-employment rates has persisted over time in many European countries, such as Germany and Spain, and even increased in the case of Great Britain and Portugal. This could be because higher levels of female education create better opportunities for women’s wage employment and, therefore, lead to lower levels of self-employment (see Verheul et al. 2006 for evidence of this link). However, the increasing levels of female labor force participation in Germany and Spain since the 1980s (ILO 2017) do not seem to be reflected in the trends of gender gap in self-employment of these two countries presented in Fig. 1. Moreover, this link is expected to be strongest in countries where women are largely engaged in necessity-motivated entrepreneurial activity with low-paid businesses. In summary, because the trends in Fig. 1 vary across countries, the aggregate picture of general education and the overall level of self-employment activity provide limited insight into the link between education and entrepreneurial activity.

We therefore argue, first, that it is important to consider the differing impacts of formal education during various stages of entrepreneurial activity. In their meta-analysis, Van der Sluis et al. (2008) show that education’s impact on entrepreneurial activity differs depending on the stage of entrepreneurial activity. While the impact of education on the first stage of the entrepreneurial process, which is selection into entrepreneurship, is insignificant, the effect on performance, as captured by indicators such as the number of employees, is positive and significant (see also Van Der Zwan et al. 2013). Other studies demonstrate that education impacts selection into some sectors as self-employed, particularly in the so-called “knowledge industries,” such as the information and communication technology (ICT) industries (Bosma et al. 2002). Similarly, according to Bates (1995), increasing levels of women’s education (captured by tertiary education) are the strongest predictor of why women are more likely to enter self-employment in skilled service fields in the USA. Likewise, focusing on GEM data for a single sector (hotels and restaurants), Ramos-Rodriguez et al. (2012) find that women are 50% more likely to enter this sector as entrepreneurs than men, while education has no impact on their choice.

Second, it is important to consider whether the impact of education on entrepreneurial activity differs between men and women. The evidence for this link is ambiguous. For instance, Marques (2017) finds that while education is positively associated with higher participation of both women and men in low-routine sectors, the influence of education level is not gendered. Van der Sluis et al. (2008) show that the effect of college graduation on the probability of selection into an entrepreneurial position is higher in the USA than in European countries and is the same for males and females. According to their analysis, though, the link between education and performance seems to be stronger for women than for men. However, these studies focus on the role of education on those that actually start up a business. A higher educational level can lead to lower entrepreneurial activity among women because traditional gender role attitudes and care duties can discourage women from pursuing entrepreneurship as a career choice in the first place. These gender differences can be relevant in understanding why men and women with similar levels of education would be less likely to participate in knowledge-intensive sectors and in growing their businesses. Therefore, we formulate the following hypothesis:

Third, it is important to consider not only the entrepreneur’s education but also the (type of) education of the population in which entrepreneurs start their businesses. Millán et al. (2014, p. 613) measure educational attainment levels in the population through the share of the population having tertiary education and show that educational attainment at the national level is linked with an individual’s entrepreneurship success in terms of survival, earnings, and job creation by own-account workers. According to Millán et al. (2014, p. 613), there are many reasons why a higher education level in the population matters for entrepreneurial activity. Highly educated populations may be characterized by (i) a higher-quality workforce, (ii) a more sophisticated and diverse consumer market, and (iii) more productivity and innovation. At the individual level, entrepreneurs may benefit from a highly educated population because it makes it easier to find qualified personnel. Additionally, a more highly educated consumer market positively affects the demand for consumer products in a qualitative sense such that the demand for innovativeness and diversity increases. Entrepreneurs may also benefit from more diverse consumer demand because it will create opportunities to enter and exploit niche markets.

Recently, policy makers and scholars have increasingly promoted STEM education as a major focus of enterprise and innovation, based on the belief that these disciplines will guide the development of new businesses and economic growth (Jones 2008). STEM subjects are of particular importance in the creation of scientific knowledge, and the entrepreneurship literature highlights the importance of scientific knowledge for the development of entrepreneurial ventures in general. Caprile et al. (2012) show that there is a skills shortage in STEM fields, creating recruitment challenges for employers in engineering, high tech/IT, and science sectors. Given that women are significantly underrepresented among STEM university graduates (OECD 2015), there is great potential to increase the share of female graduates in STEM fields. Thus, increasing the share of women in STEM fields can contribute directly to the creation of scientific knowledge and, as such, to a more innovative and productive environment. A more innovative and productive environment can create opportunities for both men and women to start businesses in more knowledge-intensive sectors.¹

To understand the gender differences in the choice of study, it is important to consider the role of informal institutions. Informal institutions, or social norms and practices, play a key role in determining the societal position of women (Dilli et al. 2015). More specifically, Flabbi and Tejada (2012) find that gender differences in fields of study are strongly related to expectations about labor market outcomes. They show that women who graduate in STEM fields are significantly less likely than men to pursue a career in those fields: 71% of male graduates work as professionals in STEM fields, while only 43% of female graduates work as professionals in STEM fields (OECD 2015). In comparison, men and women who pursue degrees in the humanities or health sciences make much more similar choices about the kinds of careers they pursue (OECD 2012). Traditional perceptions of gender roles strongly influence societal ideas of what constitute “masculine” and “feminine” vocations, and these ideas are formed early in life (Kane and Mertz 2011). In the 2012 Programme for International Student Assessment (PISA) test, parents were more likely to expect their sons to work in STEM-related fields than their daughters—even if their children performed at the same level in mathematics (OECD 2015). Closing the gender gap in STEM education can change attitudes toward feminine and masculine vocations, thereby stimulating female involvement in more (knowledge-intensive/innovative) entrepreneurial activity. Therefore, we propose the following hypothesis:

While many studies have shown that institutions matter for entrepreneurial activity,² fewer studies have paid attention to how institutions help explain gender differentials in entrepreneurial activity (e.g., Elam and Terjesen 2010; Estrin and Mickiewics 2011; Lewellyn and Muller-Kahle 2016; Marques 2017). In a comparative study of 55 countries, Estrin and Mickiewicz (2011) find that women are less likely to undertake entrepreneurial activity in countries with a larger state sector and show that restrictions on the freedom of movement away from home make it less likely for women to have high aspirations for employment growth, even if their entry into entrepreneurial activities is not affected by these restrictions. Among cultural factors, Baughn et al. (2006) show that when a society has more gender-egalitarian values, women show greater involvement in entrepreneurship. In contrast, Lewellyn and Muller-Kahle (2016, p. 770) argue that in societies where women are expected to fulfill traditional family responsibility functions (child-rearing and housekeeping), entrepreneurial activity may provide greater flexibility than working in the established business sector. Moreover, such institutional structures are also important for understanding the link between education and (gendered) entrepreneurial activity (Estrin et al. 2016). According to Estrin et al. (2016, p. 454), the relationship between human capital and an individual’s occupational choice is sensitive to the institutional context. They show that when the rule of law is strong, it ensures that commercial entrepreneurs benefit more from their human capital in their ventures; however, they do not observe the same effect for social entrepreneurs.

While studying the role of institutions in entrepreneurial activity, it is important to consider the complementarity between them (Dilli et al. 2018). Research on the varieties of capitalism (VoC) approach shows large differences between national economies (e.g., in terms of their innovativeness and sectoral specialization) due to their institutional arrangements related to the supply of knowledge, interfirm relations, finance, and labor, which support each other (Hall and Soskice 2001). Based on these four dimensions of institutions, Hall and Soskice (2001) identify two main clusters among capitalist industrial nations: liberal market economies (LMEs) and coordinated market economies (CMEs). In LMEs, firms coordinate their activities via competitive market arrangements, while in CMEs, firms depend heavily on non-market relationships, such as cooperation among economic actors. Because LMEs are characterized by flexible labor market institutions, the education system supports investments in general skills (Hall and Soskice 2001). In CMEs, because the labor market is more regulated, educational systems and in-house training encourage the development of industry-specific skills. Therefore, the return on investment associated with specific human capital (e.g., field of education) is expected to be higher in CMEs than in LMEs (Jackson and Deeg 2006). As formal education and investment in specific human capital are more important in CMEs, having fewer graduates from STEM subjects can matter more for the three stages of entrepreneurial activity in CMEs than in LMEs. We therefore hypothesize the following:

Within the VoC literature, a number of scholars have called for attention to gender dynamics (Estevez-Abe 2009; Folbre 2009; Mandel and Shalev 2009). For instance, Estévez-Abe (2006, p. 152) shows that in CMEs, strong employment protection exacerbates employers’ discrimination against women and promotes their investments in male human capital because firm-specific skills present high risks for women who are likely to interrupt their careers due to family-related contingencies. Moreover, CMEs typically have more generous social welfare policies, including those related to families, such as maternal leave and childcare. When these welfare benefits are linked with job tenure, it can make it less attractive for women to pursue careers as entrepreneurs. However, the flexibility that self-employment provides can be particularly attractive for women in contexts where there is no formal institutional support for childcare.

In LMEs, on the other hand, while women’s participation in the labor market is usually high, the quality of participation is low because competition is expected to eliminate systematic discrimination. The liberal market approach means that women who wish to combine employment with motherhood are forced into low-paid, part-time jobs. This implies that women can be overrepresented in necessity-based entrepreneurial activity in LMEs (Perrons 1995). Contrarily, the generous welfare environment of CMEs can be supportive of ambitious and opportunity-driven female entrepreneurs because they would be likely to earn enough (in the long run) to afford social security contributions and benefit from them. Following this reasoning, we formulate the following hypothesis:

It is important to note that in terms of their policies toward female integration into the workforce, European countries are characterized by a larger variation among the CME than the LME countries (Estevez-Abe 2009, p. 6). For instance, Perrons (1995) shows that wage and participation differentials between women and men in the social democratic model practiced by the Nordic countries are among the lowest in the world due to the provision of low-cost, high-quality child care, and the system of parental insurance. In Southern European countries, because family often provides the main means of welfare and general financial support, female participation in the labor market is low. As a result, the size of the gender gap in entrepreneurial activity would likely differ among the CMEs depending on the extent to which they follow female-friendly policies. This means that the size of the gender gap is likely to be smaller in social democratic countries than in the Southern European countries.

Figure 2 summarizes our hypotheses. The solid lines in the figure highlight the gendered effect of education on (individual level) entrepreneurial activity; the dashed lines show the direct effect of gender differences in educational attainment for the overall entrepreneurial activity.

To test our hypotheses, we use the well-known GEM database. The GEM database includes data from a representative national sample of at least 2000 respondents and offers comprehensive data on different forms of entrepreneurial activity, providing us a unique opportunity to answer our research questions. However, one limitation of the GEM data is that it does not provide information on entrepreneurs’ choice of study at the university level. To our knowledge, no publicly available individual-level cross-nationally comparable dataset provides this information.

We limit our analysis to 19 European countries and the USA for three reasons. First, we select those countries that have received the most attention in the varieties of capitalism approach. Second, while we focus on the European context, we include the USA because it has received substantial attention in the literature as an example of an entrepreneurial society. Third, data availability plays a role in our selection of countries. Our sample consists of the following 19 countries, with the number of respondents given in parentheses: Austria (91), Belgium (951), the Czech Republic (39), Denmark (1079), Finland (195), France (124), Germany (901), Greece (239), Hungary (217), Ireland (279), Italy (176), the Netherlands (248), Norway (282), Poland (30), Slovenia (244), Spain (1673), Sweden (284), Switzerland (210), the United Kingdom (UK; 1933), and the USA (1051).

To model gender differences during the three stages of the entrepreneurial process across 19 European countries and the USA, we use multilevel probit regression techniques, which are suitable due to the binary nature of the dependent variable (Long 1996). Since the GEM data used in our analyses are taken from 20 different countries and represent surveys taken during various years between 2002 and 2009, our data have a hierarchical structure, with individuals nested in countries and years. We can account for this hierarchal structure with a multilevel model (Hox 2010). While multilevel techniques also allow us to model a random slope for gender, we do not add a random slope for “female” because the likelihood ratio tests show that adding a random slope does not significantly improve the estimation models [LR chi-square (1) perceived opportunity = 0.46, p = 0.49; LR chi-square (1) knowledge-intensive sector = 0.00, p = 0.99; LR chi-square (1) high aspiration = 2.74, p = 0.10]. This finding already supports the European Commission’s view (2013, p. 8) that the main features of female entrepreneurship are similar across these countries.

To test our hypotheses, we follow a similar strategy as Estrin and Mickiewicz (2011, p. 404) and introduce random country-year effects to all our estimates, which accounts for unobserved heterogeneity across countries and for measurement errors and idiosyncrasies that are country-year sample specific. While the introduction of three levels with countries and years as separate levels is an alternative, the unbalanced structure of the GEM database creates estimation problems. Moreover, even when we retest our (base) models with three levels instead of two, the interpretation of the results presented below does not change. Additionally, estimates with country-fixed effects are not possible due to the slowly varying nature of our institutional indicators (e.g., Alesina et al. 2011).

In sum, we estimate the following equation:where i denotes individuals, j denotes countries, and t denotes time. Entrepreneurial activity is a dummy variable denoting whether an individual sees an opportunity to start a business, whether she/he starts a business in knowledge-intensive business sector, and whether he/she is engaged in high-growth start-up activity. First, we estimate the relevance of individual predictors and then add the contextual indicators. To test our hypotheses, we also add the interaction effects of an individual entrepreneur being female and of education and institutional variables. All models include year-fixed effects to control for common shocks. We examine for multicollinearity issues by using variance inflation factor (VIF) tests. While the inclusion of all direct effects does not indicate problematic collinearity, we present the interaction models for each contextual variable separately to avoid biased estimates due to multicollinearity issues (Maas and Hox 2005). Moreover, for simplicity, we present only the interaction effects for contextual variables that are significant. For ease of comparison, all continuous variables (on both the individual and contextual levels) in the regression analyses are mean-centered. We present the results of the estimation model in Table 4 in the following section.

Although care must be taken when discussing causality, two points can partially address this issue. The first is the exogeneity of country-level variables relative to the individual. The second is the use of early-stage entrepreneurship data. Country-level variables of interest represent slow-moving cultural conditions that were already in place when individual entrepreneurs first thought about setting up a business (Marques 2017, p. 14). The same reasoning applies to the variable for an entrepreneur’s education, which he/she (very often) has received before establishing a business.

Table 3 presents mean levels of three stages of the entrepreneurial process broken down by gender for all countries separately. Two important findings are apparent from Table 3. First, women are underrepresented compared to men on average during all three stages of entrepreneurial activity, but the size of the gender gap grows in the later stages in many countries. Additionally, we find no considerable or only small gender differences in the perceived opportunity to start a business in many countries (Austria, France, the Netherlands, Finland, Norway, Sweden, the USA, the UK, Slovenia, and Italy). The gap becomes significant and larger in the later stages of starting a business in these countries. In Spain and Greece, there are no significant gender differences in selection into knowledge-intensive business sectors. In Poland and Ireland, the gender differences are present only in perceived opportunities to start a business but disappear in later stages of the entrepreneurial process.

Second, Table 3 shows the importance of considering the country-level differences in entrepreneurial activity. The level of entrepreneurial activity in all three stages differs substantially across countries. For instance, in countries such as the USA, the UK, and the CME countries, individuals, on average, see more opportunities to start a business and are more likely to start a business in highly knowledge-intensive business sectors. Below, we explore these cross-national differences in entrepreneurial activity in greater detail and test the extent to which they relate to individual characteristics and a country’s level of gender equality at the tertiary level, the choice of study at the tertiary level, and the complementarity between institutional structure and attitudes toward gender equality in business.

The results of our multilevel probit regressions are presented in Table 4. Models 1, 4, and 8 in Table 4, which include only gender as a predictor, show that on average, women are significantly less likely to see opportunities, to be involved in highly knowledge-intensive sectors, and to engage in high-aspiration start-up activity than men, supporting the findings of previous research (e.g., Verheul et al. 2006; Estrin and Mickiewicz 2011; Marques 2017). To obtain a better understanding of the coefficients, we calculate the marginal effects. Accordingly, compared to men, the probability that women will see an opportunity to start a business is on average 3% smaller, the probability that they will engage in knowledge-intensive sectors is 6% smaller, and the probability that they will have growth aspirations is 11% smaller.

Model 2 adds individual characteristics and their interactions with the “female” variable. After including individual-level covariates, the mean gender effect is not significant. This shows that in our sample of countries, gender differences are fully explained by differences in entrepreneurs’ individual characteristics. These findings are consistent with the results of previous studies. For instance, Langowitz and Minniti (2007) show that men and women tend to react to the same set of incentives and that much of the difference across genders disappears after correcting for individual differences in socioeconomic conditions. Similarly, a report from the European Commission (2013) identifies individual characteristics such as women’s care responsibilities and lack of role models, business networks, and representation as the main barriers to female entrepreneurship.

In particular, tertiary education is associated with higher perceived opportunities and higher chances of selection into knowledge-intensive sectors, whereas it has no impact on high-growth aspirations. This provides only partial support for hypothesis 1, which states that entrepreneurs’ education level is not linked with seeing opportunities; it is positively related with the engagement in highly knowledge-intensive sectors and high-growth aspirations. Earlier findings show that education is not linked with whether one starts a business or not (see Van Der Zwan et al. 2013 for a review). The fact that tertiary education does seem to increase the probability of perceiving opportunities to start a business implies that other factors, such as financing or the ease of starting a business, may play a more important role than education in setting up a business. This supports the findings of Van der Sluis et al. (2008), who argue that education has a varying impact during the different stages of entrepreneurship. An explanation for the difference between our findings and those from the previous literature on high-growth aspirations (Van Der Sluis et al. 2008; Estrin and Mickiewicz 2011) may arise from our measurement of entrepreneurial performance. While Van Der Sluis et al. (2008) focus on the entrepreneur’s income as a measure of performance, Estrin and Mickiewicz (2011) define a highly aspirational entrepreneur as someone who aspires for firm growth of more than ten employees. We measure entrepreneurial performance as aspirations for the firm growth of more than five employees. Thus, education begins to matter for entrepreneurial performance above a certain threshold. Notably, the influence of educational level is not gendered. Therefore, we do not find evidence to support H2, which states that the impact of education on entrepreneurial activity is expected to be lower for women than for men during all three stages of entrepreneurial activity due to the social arrangements that discriminate against women.⁷ This finding does, however, support the conclusions of Marques (2017) and Van Der Sluis et al. (2008).

Individuals with the relevant knowledge, skills, and experience to start a new business may also see more opportunities to start a business and are more likely to engage in highly knowledge-intensive sectors and to become involved in highly aspirational entrepreneurial activity. This finding is consistent with earlier studies, which show that in their capacity as “jacks-of-all-trades,” entrepreneurs may require a broad range of skills (Silva 2007). Being acquainted with an entrepreneur also increases the probability of entrepreneurial activity during all stages of the entrepreneurial process. Both skills and network are factors in which the size of the gender gap is the largest among the individual factors (Table 1). Therefore, these areas should be prioritized to close the gender gap in entrepreneurial activity. Consistent with the earlier findings, the probability of seeing opportunities to start a business and of becoming an entrepreneur in knowledge-intensive sectors as well as the possibilities for firm growth are lower for older people. This could be linked to generational constraints and family responsibilities and is especially true for women who are involved in highly knowledge-intensive sectors.

Because individual differences account for the gender gap in entrepreneurial activity, we do not find any support for hypotheses 4 and 6, which argue that cross-national differences in gender equality with regard to education and institutional environment should help to explain the gender differences in entrepreneurial activity. However, we test the role of gender differences at the contextual level in explaining the cross-national differences in overall levels of entrepreneurial activity, as argued in hypotheses 3 and 5. The results of models 3 and 10 in Table 4 show that countries with higher gender equality at the tertiary level also have more individuals who see an opportunity to start and grow their business. While the education field does not matter in determining whether an individual sees an opportunity to start a business, in countries with higher gender equality in science, individuals are more likely to engage in knowledge-intensive business sectors and to see opportunities to grow their businesses (models 6 and 9 in Table 4). On average, in countries that achieve gender equality in science education, the probability of finding entrepreneurs in highly knowledge-intensive sectors is 25% higher, and the probability of finding entrepreneurs with high-growth aspirations is 10% higher relative to countries that do not. This finding provides support for hypothesis 3. Consistent with the conclusions of Dilli et al. (2018), individuals see significantly fewer opportunities to start a business in the Mediterranean and Eastern European market economies than in the liberal market economies. Individuals also have lower growth aspirations in the Mediterranean economies than in the liberal market economies. Interestingly, more individuals are engaged in knowledge-intensive sectors in coordinated/Nordic market economies than in liberal market economies.

Models 7 and 11 in Table 4 test whether the impact of gender equality in science on entrepreneurial activity varies across different institutional constellations. Figure 3 presents the marginal effect of the gender gap in science calculated over four institutional constellations based on models 7 and 11 in Table 4, including both the main and the interaction terms. According to Fig. 3, the benefits of closing the gender gap in science for entrepreneurial activity in knowledge-intensive businesses are highest in the CMEs. This could be due to the fact that the CMEs have moderate employment protection and higher governmental expenditure in education than other institutional contexts, which can stimulate investment in highly specific skills. Moreover, CMEs generally pursue more female-friendly policies, which means that women who pursue education in science subjects are more likely to pursue a career in the same field. This provides partial evidence for our hypothesis 5, which argues that the impact of gender differences in STEM education on entrepreneurship is smaller in LME institutional constellations where investment in general skills is more important. However, we do not find any evidence that the impact of gender equality in science education on perceived opportunities or high-growth aspirations varies substantially among different institutional constellations.