Networking Increases the Probability of Women’s Start-Ups in EMDEs

Abstract In emerging markets and developing economies (EMDEs), entrepreneurship and self-employment, particularly among women, play a pivotal role as a means of livelihood. Unfortunately, the share of women participating in these activities is significantly lower compared to men, which has a negative effect on both the development of these countries and gender equality. Usual programs that support women’s entrepreneurship are costly. We aim to explore the role of social networks for the entrepreneurial entry of women in EMDEs, as women can ‘help themselves’ through their development and activation. Our results, using GEM data for 53 low- and middle-income countries for the period 2011–2018, revealed that an entrepreneur on the women’s social network increases the probability of their entrepreneurial entry by 3.3 percentage points. Our results also suggest that social networks represent an important source of information that enhances the determination of women to take advantage of good business opportunities, increases the positive effect of self-assessed start-up skills on entrepreneurial entry, and raises awareness of business risks. Based on these results (confirmed by robustness and causality checks), we can recommend supporting networking between entrepreneurs and women by targeted policy measures.


Introduction
In emerging markets and developing economies (EMDEs), a minority of individuals work as employees, with only 18% of men and a mere 7% of women engaged in wage labour (World Bank, 2022a).Consequently, entrepreneurship and self-employment, particularly among women, assume a pivotal role as a means of livelihood in these countries.However, the 2018/2019 Women's Entrepreneurship Report (Elam et al., 2019) shows that the average rate of women's participation in nascent and new enterprises is approximately 80% of the rate for men's participation in low-income countries.
Commonly used programs to support women's entrepreneurship, such as developing women's entrepreneurial skills or providing them with better access to credit can be very costly.However, the support of social networks can serve as a catalyst for women's entrepreneurial entry and business aspirations in EMDEs, as its accumulation does not require specific policy measures and women can 'help themselves' with no or only minimal costs.The theory of entrepreneurial behaviour points to the important role of social networks in the decision of whether to become an entrepreneur (Gulati, 1998(Gulati, , 1999;;Uzzi, 1999).Minniti (2004) developed a model of entrepreneurship adoption with social network externalities, where she argued that the presence of other entrepreneurs reduces ambiguity by acquiring information and skills.Throughout this process, social networking becomes important and may be modelled as a network externality in which entrepreneurship is assumed to exhibit increasing returns to adoption.Empirical evidence, in general, brings a consensus on the importance of social and professional networks in the creation and pursuing new opportunities (Leyden, Link, & Siegel, 2014), start-ups (see next section for a review), venture capital (Alexy, Block, Sandner, & Ter Wal, 2012) and firm performance (Stam, Arzlanian, & Elfring, 2014), but there is very limited evidence on its role for entrepreneurial entry in EMDEs (e.g.Ali & Yousuf, 2019).
To address these gaps in the literature, this article aims to explore the role of social networks in women's entrepreneurial entry and business aspirations in EMDEs.We employ data from the Global Entrepreneurship Monitor for 53 low-and middle-income countries in the period 2011-2018 to provide robust and up-to-date evidence on the importance of social networks for entrepreneurial entry and business aspirations.We also pay the necessary attention to the mechanism of this relationship, as well as the issues of endogeneity and causality.
Our results confirm a statistically significant gender gap in both intentions to start a business and involvement in start-up activities in EMDEs, yet reveal higher gender equality in business intentions rather than behaviour pointing to barriers for women entrepreneurship entry.However, networking with entrepreneurs largely increases the probability of women setting up their businesses.Our investigation suggests that the social network became a source of important information that increases the positive effect of self-assessed start-up skills on entrepreneurial entry and supports women's determination but also mitigates their excessive optimism.The positive effect of networking seems to be universal but more effective in more gender-equal countries.Robustness and causality checks support these results.
There are three main contributions of this article.First, it provides new evidence on the understudied relationship between the social capital of women and their entrepreneurial entry in EMDEs.Our results can be easily generalised as they stem from data on 53 EMDEs, representing 84.9% of the population in these countries, in the period 2011-2018.Second, the use of a large dataset made it possible to extend the discourse on the role of social capital beyond its influence solely on the initiation of entrepreneurial activities (as observed in other studies) and examine its impact on business aspirations as well.By considering this broader perspective, our study offers a more comprehensive understanding of the subject matter.Third, we devoted considerable attention to exploring and addressing issues of self-selection bias and reverse causality.
The rest of the article is structured as follows.Section 2 provides a review of the literature on social capital and its role in entrepreneurial entry, especially in the case of women.Section 3 introduces the data and empirical methodology used in this article.Section 4 presents and discusses the results.It focuses on the role of social capital on entrepreneurial entry (Section 4.1), business aspiration (Section 4.2), and addressing the issue of endogeneity (Section 4.3).Section 5 concludes.

Empirical literature review
The empirical studies investigating the relationship between social networks and firm formation (e.g.Ali & Yousuf, 2019;Amo, 2013;Elfring & Hulsink, 2003;Greve & Salaff, 2003;Ruiz-Palomino & Mart ınez-Cañas, 2021;Seet, Jones, Oppelaar, & de Zubielqui, 2018) show the significant positive correlation.Ruiz-Palomino and Mart ınez-Cañas (2021) elucidated how entrepreneurial social networks could help bridge the entrepreneurial intention-behaviour gap and showed that access to family-based entrepreneurial social networks had a greater impact in this regard than friends-based entrepreneurial social networks.The study used original data on 616 university students in Spain.Other significant positive findings were found in the study by Åmo (2013), who tested the link between personal educational level, personal knowledge of other entrepreneurs, and the propensity to engage in business start-up activities using the survey conducted amongst inhabitants in Norway.The significant link between social networks and activities that led to starting a business was also recorded by Seet et al. (2018).The study involved thematic analysis of semi-structured interviews with participants in an Australian start-up accelerator.The authors investigated the impact of three components of human capital, 'know-what', 'know-how', and 'know-who', on the start-up activities and found that 'know-who' was most significant for participant learning.Another case study focused on the rural area in Pakistan (Ali & Yousuf, 2019).The results showed that social capital has a significant positive impact on entrepreneurial intentions by forming perceived desirability, perceived self-efficacy, and perceived social norms toward entrepreneurship.Elfring and Hulsink (2003) analysed the impact of social networking on three entrepreneurial processes in new company development, i.e. discovery of opportunities, securing resources, and obtaining legitimacy.The authors further distinguished between strong and weak social ties.The generally approved value of strong ties was confirmed in the process of discovering opportunities and securing resources while the importance of weak ties was found when legitimacy was concerned.However, the results are based only on three cases of high-technology firms in the Netherlands.Greve and Salaff (2003) studied the social network activities of entrepreneurs through three phases of establishing a firm as well.Using data from four developed countries (Italy, Norway, Sweden, and the USA) the authors found out that entrepreneurs use networking more during the planning than other phases of entrepreneurial entry activities.
Studies focused on gender differences in the relationship between social networks and the motivation to start a business point to the problem of traditional social norms that discourage women from being entrepreneurs and recommend government policies and programs aimed at creating an environment more conducive to female entrepreneurs (e.g.Chen et al. 2021;Klyver & Grant, 2010;Kwong, Thompson, Jones-Evans, & Brooksbank, 2009).The first study explored Ghanaian female entrepreneurs' motivation for establishing their businesses and the level of support they receive from their personal social networks.The second study, using Global Entrepreneurship Monitor data from 35 countries pooled across three years (2002)(2003)(2004) and multinomial logistic regression, indicated that individuals who personally know an entrepreneur are more likely to get involved in entrepreneurial activity.However, women, compared to their male counterparts, are less likely to personally know an entrepreneur.Based on that, the authors suggest that one of the reasons why women are less likely to become entrepreneurs is that they lack social networks or role models in their social networks.A study by Chen et al. (2021) revealed the same results when examining the role of human capital and social networks for opportunity-driven entrepreneurship using data from the Global Entrepreneurship Monitor in the year 2016.Kwong et al. (2009) focused on women in the United Kingdom.
Networking and women's start-ups in EMDEs 1073 Using data from Global Entrepreneurship Monitor and the binary logistic method (similar to our article), the study revealed the significant positive relationship between social networks and the probability of being involved in nascent entrepreneurial activities in the female population.
Other studies related to gender differences in social networking and entrepreneurship examined specific issues such as the impact of sticky floor perception (Shabsough, Semerci, & Ergeneli, 2021), different types of social networks (Koellinger, Minniti, & Schade, 2013;Neumeyer et al., 2019), or the opposite direction of the relationship between networking and entrepreneurship (Sharafizad & Coetzer, 2017).The study by Shabsough et al. (2021) examined the role of social networking in the relationship between sticky floor perception and the entrepreneurial intention of 294 female workers.The results of this study indicated that social networks bring a stronger positive relationship with entrepreneurial intention among women reporting a high sticky floor perception.Women experiencing sticky floor feel trapped in lowwage positions which may lead them to seize the opportunity and establish their businesses and become entrepreneurs.Neumeyer et al. (2019) used social network data from two municipal entrepreneurial ecosystems in Florida, USA, to show that male entrepreneurs had higher comparative scores in aggressive-and managed-growth venture networks, while women entrepreneurs surpassed their male counterparts in lifestyle and survival venture networks.Different types of male and female social networks were analysed, as well, in the study by Koellinger et al. (2013), and explained a substantial part of the gender gap in entrepreneurial activity.Although most articles assume the impact of social networks on the intention to start a business, Sharafizad and Coetzer (2017) discussed the opposite direction of the relationship, i.e. the impact of different start-up motivations and phase of the business (prestart-up, start-up, and established business) on womens networking behaviours.
On the one hand, this overview confirms the importance of social capital for business activities and its relevance in explaining the gender gap in entrepreneurship.On the other hand, it points to the lack of evidence on these issues from low-and middle-income countries.The literature especially misses a broader picture based on data from many countries, the mechanism between social capital and entrepreneurial entry, and the causality of this relationship, which represents issues discussed in the following paragraphs.

Data
Individual data from the Adult Population Survey 2011-2018 (APS) of the Global Entrepreneurship Monitor (GEM) are employed in this article (GEM, 2022).The APS represents an international survey based on a comprehensive questionnaire focusing on entrepreneurial aspirations, activities, attitudes, and personal characteristics of the adult population, which is administered by phone or face-to-face interviews.Data from each country are double-checked (e.g. for out-of-range values, patterns of missing data, skip logic errors, higher than usual incomplete or refusal rates), coded, and weighted to create a harmonized data set, which ensures representativeness and consistency across participating countries.It should be noted that we applied the country sample weights (reflecting gender and age distribution at a minimum, but often reflecting also factors such as region, education level, and urban/rural stratification), as well as equal weights for all countries to identify behavioural patterns prevailing in most countries, not only in populous ones.
The APS was replenished with data on the Global Gender Gap index published by the World Economic Forum, which 'measures gender-based gaps in access to resources and opportunities in countries' (World Economic Forum, 2019, p. 45).It comprises four areas of life: economic participation and opportunity, educational attainment, health and survival, and political empowerment.Considering the ongoing discussion on measuring gender differences (e.g.Beneria & Permanyer, 2010;Stoet & Geary, 2019) and data availability, we chose the Global Gender Gap index as an approximation of social norms related to the status of women in different countries.
GDP per capita, provided by the World Bank (2022b), represents the second variable that replenishes GEM data.Both these variables are used as an approximation of countries' conditions in models examining the role of gender equality on the relationship between social networks and entrepreneurial entry.The definition of variables and descriptive statistics for samples used in this article can be found in Appendix 1 (all appendices are available in Supplementary Materials).
In our article, we pay attention to women from emerging market and developing economies (EMDEs) aged 18-64 years with no entrepreneurial experience in the last 3 years.It sheds more light on factors influencing the formation of new female entrepreneurs in lowand middle-income countries, which may represent an important factor for both reducing gender inequality and supporting economic growth in these countries.Our sample consists of 437,444 individuals (216,306 females) from 53 1 low-and middle-income countries (see Figure 1), which represented 71.1% of the world population and 84.9% of the population in EMDEs in 2018.
It should be noted that employing a large sample is connected with the potential to detect smaller and more complex effects, but also with quickly increasing statistical significance (decreasing p-values) of the results with a growing sample size.Under these conditions, relying only on p-value may support hypotheses of little or no practical significance.Therefore, special attention will be paid to presenting effect sizes, and also providing information on the coefficient, confidence intervals, and p-value depending on the sample size for the variable of interest, as suggested by Lin, Lucas, & Shmueli (2013).

Empirical model
The empirical model employed in this article stems from the theory of planned behaviour (Ajzen, 1991(Ajzen, , 2011) ) applied to entrepreneurship entry.This theory is based on the assumption that the intentions behind a given behaviour form the foundation of any planned action.Ajzen outlined three independent determinants of intention: attitudes, subjective norms, and perceived behavioural control.Attitude is defined as the extent to which an individual holds a favourable or unfavourable evaluation of the behaviour.Subjective norms pertain to the perceived societal pressure regarding List of countries: Algeria, Angola, Argentina, Bangladesh, Belize, Bolivia, Botswana, Brazil, Bulgaria, Burkina Faso, Cameroon, Colombia, Costa Rica, Croatia, Ecuador, Egypt, El Salvador, Ethiopia, Georgia, Ghana, Guatemala, Hungary, Chile, China, India, Indonesia, Iran, Jamaica, Jordan, Kazakhstan, Macedonia, Madagascar, Malaysia, Mexico, Morocco, Namibia, Nigeria, Pakistan, Panama, Peru, Philippines, Romania, Russia, South Africa, Suriname, Thailand, Tunisia, Turkey, Uganda, Uruguay, Venezuela, Vietnam, Zambia.Source: Authors, made with Natural Earth (https://www.naturalearthdata.com.).
Networking and women's start-ups in EMDEs 1075 whether to engage in or abstain from the behaviour under consideration (Ajzen, 1991).Perceived behavioural control is defined as the individual's perception of how easy or challenging it is to carry out the behaviour.This perception is influenced not only by past experiences but also by anticipated obstacles and various factors that might hinder the execution of the behaviour.Ajzen (1991) also posited that perceived behavioural control should directly predict behaviour as the increased feelings of control can enhance an individual's willingness to invest extra effort in order to successfully execute a specific behaviour.The theory of planned behaviour was applied to predict entry into self-employment in the study by Kolvereid and Isaksen (2006).The authors used longitudinal data on new business start-ups at the universities in Norway.Based on the theory and subsequent empirical study, we created our model of planned behaviour applied to entrepreneurship entry which determines three independent factors of intention to start a business: (a) the attitudes towards the entrepreneurial entry representing the subjective evaluation of its costs and benefits, (b) the subjective norms representing the social pressure associated with the entrepreneurial entry (these two factors correspond to the theory of reasoned action; see Madden, Ellen, & Ajzen, 1992) and (c) perceived behavioural control over the entrepreneurial entry (this factor corresponds to Bandura's concept of self-efficiency; see Bandura, 1986).The intention, together with control over the action, then determines the real behaviour (see Figure 2).
In our model, attitudes (AT) towards the entrepreneurial entry (STARTUP) are mainly influenced by self-assessed start-up skills, i.e., perceived possession of 'the knowledge, skill, and experience required to start a new business', education, the presence of entrepreneurs in an individual's social network, approximated by 'knowing someone personally who started a business in the past 2 years', and psychological barriers and expectations, since they affect the perception of its costs and benefits.Subjective norms (SN) are approximated by several variables that describe the status of entrepreneurs in society, which represent expectations from and social pressure on the behaviour of individuals.Behaviour control (BC) is approximated by household income as an important condition for overcoming the initial costs of any business activity, as well as retired and student status as they may decrease the perceived control upon the entrepreneurial entry.Moreover, the model also contains control variables (C) to capture the influence of other differences between individuals, countries, and years; see Equation (1), where i represents an index for an individual observation, U is the cumulative standard normal distribution function, and a, b, c, d and k are vectors of regression coefficients.
The dependent variable STARTUP captures either the intention to start up a business, i.e. 'the expectation to start up, including any type of self-employment, in the next 3 years' (futsup in the GEM dataset), or a nascent business, i.e. 'being actively involved in the start-up effort as an owner or part-owner, but the business has not generated wages in the last 3 months' (suboanw in the GEM dataset).It should be noted that due to imperfect consistency between an intention to entrepreneurship entry and real behaviour, caused by changes in the initial conditions in time or limited control over the action, we will focus nearly exclusively on starting up a new business, while intentions will be discussed only briefly.
As the presence of entrepreneurs in the social network represents a variable of our main interest in our decision-making model, we should briefly discuss its theoretical basis as a factor affecting attitudes towards entrepreneurial entry.Bourdieu (1986) considers the size of a network of connections with other individuals that can be effectively organized, as a form of capital.Although Coleman and Putnam defined 'social capital' more broadly, they saw personal networks as a source of valuable information (Coleman, 1988) or other benefits accessible through coordination and mutual cooperation with others (Putnam, 1995).Information and other benefits provided by the social network have the potential to significantly decrease costs and increase benefits of entrepreneurship, and thus positively affect attitudes towards entrepreneurial entry.In discussing the strength of ties within the network, which can be defined as a combination of factors such as the amount of time, emotional intensity, intimacy, and reciprocal services that characterize the tie (Granovetter, 1973), our operationalization of the social network variable as 'knowing someone personally who … ' suggests that our study primarily focuses on weak ties, as opposed to, for example, the China Family Panel Studies (CFPS), which considers entrepreneurs within the family.According to Granovetter's (1973) model, weak ties create more and shorter paths than strong ones, which makes networking more effective and often serves as bridges between groups.This can further strengthen the expected relationship between the network and entrepreneurial entry.A slightly different mechanism of the discussed relationship was described by Minniti (2004) who modelled the positive effect of social network externalities on the adoption of entrepreneurship.She claimed that watching and learning from others helps potential entrepreneurs gather knowledge and skills, which in turn reduces ambiguity.This leads to a situation where the benefits of adopting entrepreneurship grow as more people in the local environment engage in entrepreneurial activities.In other words, the more entrepreneurial activity there is in an area, the higher the positive network externalities.
The binomial probit model, described in Equation ( 1), allows us to estimate the impact of the presence of entrepreneurs in women's social networks (ENET) on the probability of starting their businesses in EMDEs.However, revealing the effect of ENET on the probability of entrepreneurial entry represents only the first step.Re-estimation of Equation ( 1) with the corresponding interaction terms allows us to investigate the mechanisms of how the ENET works, i.e. whether the success of other fresh entrepreneurs just decreases the psychological barriers of entrepreneurial entry or they support it also with valuable information.It should be noted that the average marginal effects of interaction terms presented in this article are calculated using a method of Norton, Wang, and Ai (2004), as they showed that the 'magnitude of the interaction effects in nonlinear models does not equal their marginal effects' (Ai & Norton, 2003, p. 123).
Starting one's own business is not, however, the only output explored in this article.The characteristics of the business are at least of the same, if not greater, importance.Therefore, we focus on answering the question of whether a social network containing entrepreneurs (ENET) has a significant effect on business aspirations.It requires the estimation of several ordered probit models explaining business aspirations (ASPIR), approximated by the expected future company size, export performance, and innovation level (all defined as ordinal variables), by ENET and other explanatory variables described above; see Equation (2), where s is cutpoints, n is the number of categories of the dependent variable and m ¼ 1 to (n − 1).
Networking and women's start-ups in EMDEs 1077 It can be noted that the application of probit models in this article was driven by two factors: First, they provide a better fit to data than logit models (based on the comparison of probit and logit estimations of fully specified Models 3, 7, and 9 by adjusted McFadden R 2 , Akaike information criteria, and Bayesian information criteria).Second, the assumption of normal distribution in the case of probit models makes them more suitable for applying the instrumental variable (IV) approach to deal with the potential endogeneity of the variable of interest.
As the Global Entrepreneurship Monitor provides pooled data, i.e. time series of cross-section data where observations do not necessarily refer to the same unit in time, probit models provide us with information on the correlation between one's social network containing entrepreneurs (ENET) and entrepreneurial output.The issues of endogeneity, selection bias, and causality are addressed by using other methods, i.e.IV approach, propensity-score matching (Heinrich, Maffioli, & V azquez, 2010;Li, 2013), and cross-lagged ML-SEM panel data model (Leszczensky & Wolbring, 2019), and discussed in a separate section of this article and in the Supplementary Materials.

Entrepreneurial entry
4.1.1.Gender disparities in start-up participation are linked to self-assessed start-up skills, social networks and other personal characteristics.In low-and middle-income countries, there is a 32.2% probability that an individual intends to start up a business in the following 3 years and a 9.4% probability of being already involved in start-up efforts (predicted probability based on Models 1 and 5 in Table 1).However, these probabilities are significantly lower for women than men.The gender ratio of the probabilities of intending to start up a business is 85.9% (the probability of having start-up intention is 34.7% for men and 29.8% for women) and it further decreases to 74.8% when the involvement in start-up activities is considered (the probability of being already involved in start-up activities is 10.7% for men and 8.0% for women).The results suggest that although the gross gender gap is statistically significant for both entrepreneurial intentions and behaviour, it is more pronounced in the case of real behaviour.Hierarchical regression analysis, which confirmed the relevance of the theory of planned behaviour for predicting entrepreneurial intentions and activity, revealed that the gender gap in an entrepreneurial entry is related mainly to the gender differences in personal characteristics determining the attitudes toward entrepreneurship (the inclusion of personal characteristics into model reduced the regression coefficient of the gender gap by about 45%; compare Models 1-2 and 5-6), while social norms towards entrepreneurship and perceived behavioural control had no significant effect on the gender gap.
The gender disparities raise a question as to whether the presence of an entrepreneur in one's social network (ENET), i.e., the variable of our interest approximated by 'knowing someone personally who started a business in the past 2 years,' contributes to the increase or decrease of gender differences in entrepreneurial entry.Two possible factors have to be considered: the gender distribution of social contacts with entrepreneurs and their different effects on the entrepreneurial intentions and behaviour of men and women.The first factor contributes to increasing gender disparities, as only 38.9% of women know someone who started a business in the past 2 years, while 46.8% of men have this kind of contact (it corresponds to the findings of Koellinger et al., 2013, for developed countries).The second factor mitigates the gender differences in intentions to start up a business, as the effect of entrepreneurs in the social network (ENET) on the probability of start-up intentions is 0.9 p.p. higher for women (Model 4), while there is no gender difference in the effect of ENET on involvement in start-up activities at the 0.05 significance level (Model 8).Networking and women's start-ups in EMDEs 1079 (2)   1) clearly shows that entrepreneurs in women's social networks increase the probability of their entrepreneurial entry by 3.3 p.p. with a 95% confidence interval ranging from 3.0 to 3.6 p.p.It shows that there is a 6.5% predicted probability of being involved in start-up activities for women who know no entrepreneurs, but 9.9% for women who have entrepreneurs in their social network (ENET).This makes ENET the third most important predictor of women's entrepreneurship entry in Model 9.The standardized coefficients of the factors considered by this model were higher only for self-assessed start-up skills (0.4547), making them the most important predictor of women's entrepreneurship entry, and the perception of good opportunities for starting a new business in the next six months (0.1865) compared to ENET (0.1782).As the results of Model 9 are of the main interest in this article, we should answer the question of how much the statistical significance and regression coefficient of the ENET variable are influenced by the sample size.For this purpose, we followed the methodology described by Lin et al. (2013) and re-estimated Model 9 for samples ranging from 500 to 216,306 observations (samples differ by 500 observations).The results revealed that even the sample with 500 observations (including only 43 nascent entrepreneurs) found the positive effect of the ENET variable on entrepreneurial entry statistically significant at the 0.01 level; see Supplementary Appendix 2. It revealed that the relationship between entrepreneurs in women's social networks and their entrepreneurial entry is statistically significant even for relatively small samples.
4.1.3.Social networks provide valuable information.The results confirmed the positive effect of 'knowing someone personally who started a business in the past 2 years' on entrepreneurial entry, for which there are two possible explanations.First, the entrepreneur can represent 'an example of good practice' showing that it is possible to successfully start a business.In this case, we can expect that knowing an entrepreneur may mitigate one's fear of failure and support entrepreneurial entry when conditions are considered favourable.Second, the entrepreneur may represent a source of information that is 'important in providing a basis for action', as noted by Coleman (1988, p. 104).In that case, we can expect the same effects as in the previous case but extended mainly by the increase of self-assessed start-up skills or the reinforcement of their positive impact on entrepreneurial entry.Therefore, we re-estimated Model 9 with corresponding interactions between social networks, self-assessed start-up skills, fear of failure, and perceived opportunity for starting a business to understand the nature of this relationship.It showed that knowing some entrepreneurs increases the positive effect of self-assessed start-up skills on entrepreneurial entry (average marginal effect of the interaction term reaches the level of 0.012; see Model AP1 in Supplementary Appendix 3).Moreover, having an entrepreneur in one's social network increases the proportion of women claiming that they possess the knowledge and skills required to start their own business, from 39.9% for those who know no entrepreneur to 65.9% for those who know some.Both findings support our hypothesis regarding the social network as a valuable source of information.On the other hand, the presence of an entrepreneur in one's social network can strengthen the negative effect of fear of failure on involvement in start-up activities (average marginal effect of the interaction term reaches the level of −0.006; see Model AP2 in Supplementary Appendix 3).It suggests that more information on running a business does not necessarily mitigate worries, but excessive optimism which is discussed later.We also discovered that the impact of having good opportunities for starting one's own business is greater when a woman has connections with entrepreneurs (average marginal effect of the interaction term reaches the level of 0.008; see Model AP3 in Supplementary Appendix 3).In this case, better information may support the confidence of women about the suitability of current conditions for starting their own business.These results support the hypothesis of social networks as a source of valuable information, not only an example of business success.
Networking and women's start-ups in EMDEs 1081 4.1.4.Gender equality increases the effect of social networks on women's start-up effort.As we investigate women's entrepreneurial entry in EMDEs, we should also examine gender-related social norms in these countries.Therefore, Model 9 was re-estimated with the Global Gender Gap index and relative GDP per capita, as an approximation of national conditions, instead of the dummy variables for countries.As expected, the results show that one standard deviation increase in the Global Gender Gap index increases the probability of intentions to start up a business by 1.0 p.p. (p < .001;not reported here) and active involvement in start-up activities by 1.7 p.p. (Model AP4 in Supplementary Appendix 3), which suggest that gender inequality does not affect intentions as seriously as their realisation.The predicted probability of starting up their own business in the case of women ranges from 4.2% to 12.8% depending on the value of the Global Gender Gap index reaching values from 0.55 to 0.79 in our sample.These results confirm the high importance of gender equality in the country for women's entrepreneurial activity.Further investigation revealed that contact with entrepreneurs may be more beneficial in countries with more equal attitudes towards women.The interaction between the Global Gender Gap Index and ENET was found to be positive (with an average marginal value of the interaction reaching 0.021), but statistically insignificant at the 0.05 level (p ¼ .08;see Model AP5 in Supplementary Appendix 3) when assuming a linear relationship between the interaction and women's entrepreneurial activity.However, Models 10-13 in Table 2 show that the effect of social networks on the probability of women's entrepreneurial entry increases with gender equality, but at a decreasing rate (2.3 p.p. and 3.9 p.p. in the countries with the lowest and highest gender equality, respectively).This means that in countries with high gender inequalities, even a small step toward greater gender equality can lead to a substantial increase in the impact of social networks on entrepreneurial entry.Conversely, in countries with higher gender equality, further improvements in gender balance may not have as pronounced an effect on the influence of social connections in entering the business world.
4.2.Entrepreneurial aspirations 4.2.1.Information from social networks helps to be more realistic.As was shown above, knowing someone who has started a business in the past two years significantly increases the probability of women starting their businesses.The results suggest that it is not only connected with seeing an example of business success but also with a flow of information on entrepreneurship.
It raises the question of whether information gained from entrepreneurs also affects the business aspirations of nascent entrepreneurs.
To provide an answer to this question, we estimated the ordered probit model explaining innovativeness, export orientation, and intended size of the future business by the same predictors as in Model 9. Innovativeness was approximated by three questions: (1) 'Will all, some or none of your potential customers consider this product or service new and unfamiliar?', (2) 'Right now, are there many, few, or no other businesses offering the same products or services to your potential customers?',and (3) 'How long have the technologies or procedures used for this product or service been available?'. Regardless of using these questions as separate dependent variables or the innovativeness factor defined by explanatory factor analysis, no statistically significant relationship with social contacts with entrepreneurs was found (not reported here).We also got the same result for export orientation (not reported here) approximated by the question 'What percentage of your annual sales revenues will usually come from customers living outside your country?'.On the other hand, the expected size of the business, approximated by the question 'Not counting owners, how many people will be working for this business five years from now?', was found to be influenced by social contacts with entrepreneurs.Model 14 in Table 3 revealed that women with entrepreneurs in their social network expect a lower number of employees (no employees or up to 10 employees) more often than those without these contacts and related information.
These results suggest that contacts with entrepreneurs have no influence on business characteristics described by a business plan, i.e., product and its technological level, or export orientation, but have a significant effect on expectations about the future development of the company.It seems that information from social networks helps women to be more realistic, as was already concluded in the previous section.This conclusion supports the theoretically expected causality from social capital to start-up behaviour, as it is more likely that betterinformed individuals lower their business expectations than individuals with more ambitious business plans systematically avoid contact with businessmen.However, the issue of causality is discussed further in the following section.

Endogeneity and robustness check
In this section, we apply various estimation methods to address the issue of endogeneity and causality, as well as to check the robustness of the results.
Originally, we intended to use the instrumental variable approach to check for reverse causality and unobserved heterogeneity between the social network and entrepreneurial entry of women, but the GEM dataset does not provide suitable instruments for social networks (all Networking and women's start-ups in EMDEs 1083 tested instruments appeared to be weak).Then we address the potential problem of the nonrandom presence of entrepreneurs in the social networks of individuals, as women intending to start their businesses may intentionally establish these social ties.We employ the propensity score matching method (PSM) to eliminate potential bias caused by the described self-selection (used also by Brixiov a, Kangoye, & Yogo, 2020 for the estimation of the effect of access to finance on employment in SMEs in Africa).The average effect of treatment for all women ranges from 3.5 to 3.7 p.p., which supports the robustness of our results presented above.Detailed description of the method and results is presented in Supplementary Appendix 4. We further checked causality by using a dynamic cross-lagged ML-SEM model with fixed effects, as Leszczensky and Wolbring (2019) suggest.The results confirmed our previous intuitive conclusion on the causal impact of social networks on the probability of starting own business.Detailed description of the method and results is presented in Supplementary Appendix 5.

Conclusions
This article focuses on the role of social networks as a catalyst for women's entrepreneurial entry in emerging markets and developing economies (EMDEs), as these networks may represent a source of information that is 'important in providing a basis for action' (Coleman, 1988, p. 104).Furthermore, networking does not require specific policy measures, and women can 'help themselves' with negligible costs.At the same time, support from local or central government can largely increase the efficiency of social capital accumulation and multiply its effect.
Individual data from the Adult Population Survey 2011-2018 of the Global Entrepreneurship Monitor (GEM) for 53 low-and middle-income countries, representing 84.9 per cent of the EMDEs population in 2018, were employed in this article.We paid attention to individuals, especially women aged 18-64 years with no entrepreneurial experience in the previous 3 years, to provide evidence on the role of networking for the formation of new female entrepreneurs.Our sample consisted of 437,444 individuals, of which 216,306 were females.These data were used for the estimation of probit models based on the theory of planned behaviour.
Our results confirmed that women in low-and middle-income countries have a lower probability than men of both having business intentions (34.7% for men and 29.8% for women) and being involved in start-up activities (10.7% for men and 8.0% for women) in the basic model controlling only for gender, country, and year of the survey.Although the gender gap was statistically significant in both cases, it was more pronounced in the case of involvement in start-up activities.Further inclusion of determinants of attitudes toward entrepreneurship (such as age, education, self-assessed start-up skills, social network, fear of failure, and perceived opportunity for starting a business) in the model led to a significant gender gap decrease by about 45%.The model was then enriched by the variables representing determinants of social norms and perceived behavioural control that had almost no effect on the gender gap.So the results indicate that the gender gap in an entrepreneurial entry is related mainly to the gender differences in personal characteristics determining attitudes toward entrepreneurship.
Further analysis focused on women revealed that having an entrepreneur in their social network increases the probability of women's entrepreneurial entry by 3.3 p.p. (the predicted probability of being involved in start-up activities was 6.5% for women with no entrepreneur in their social network and 9.9% for those with some entrepreneur in their social network).This makes social networks, subsequent to self-assessed start-up skills and the perception of good business opportunities, an important determinant of entrepreneurial entry.It can be noted that the effect of networking on women's entrepreneurial entry was found to be stronger in more gender-equal countries (measured by the Global Gender Gap index).The robustness of the relationship between networking with entrepreneurs and involvement in start-up activities was confirmed by re-estimation of the model with a different model specification and sample sizes.Moreover, the application of the propensity score matching method, using different matching algorithms, quantified a similar effect of social networks with entrepreneurs on the probability of starting their own business (the average treatment effect) ranging from 3.5 to 3.7 p.p.
Examination of the interactions between 'knowing someone personally who started a business in the past 2 years' and other determinants of business entry suggested that the entrepreneur does not represent only 'an example of good practice', showing that it is possible to successfully start a business, but is a source of important information.It was found that knowing an entrepreneur straightens the positive effect of self-assessed start-up skills and perceived good business opportunity, but also the negative effect of fear of failure on entrepreneurial entry.It suggests that information from social networks, compared to a situation of no information flow, helps women be more confident about their start-up skills and be more aware of business risks.The 'mitigation of excessive optimism', as an effect of better access to information, is also supported by evidence on entrepreneurial aspirations, as women with an entrepreneur in their social network have a significantly lower expectation of the number of employees in their business in five years, although it has no influence on their business plan (innovativeness and technical level of their production and export orientation).The negative relationship between networking and the expected number of employees supports the theoretically expected causality, as it is more likely that better-informed individuals reduce their excessive business expectations than individuals with more ambitious business plans systematically avoid contact with businessmen.We confirmed the expected causal effect of social networks on an entrepreneurial entry also by employing the cross-lagged ML-SEM model recommended by Leszczensky and Wolbring (2019) on the aggregated GEM data for each country and year investigated.
This article provides evidence of the positive effect of networking with entrepreneurs, as a source of important information, on women's entrepreneurial entry in EMDEs.Although women themselves can use the findings presented in this article to intentionally build their social networks and overcome the entrepreneurial gender gap, the accumulation of social capital can be speeded up by suitable policy measures.For example, the local or even central government could support a co-working environment or communication platforms with entrepreneurs as mentors, who would help women to discuss their business intentions and put them into action.
Although we have provided robust evidence on the importance of the presence of an entrepreneur in one's social network for women's entrepreneurial entry, the limitation of GEM data prevents us from complementing our findings with information about how the effects of networking change with varying numbers of entrepreneurs in the network and different intensity of ties with them (Granovetter, 1973).Equally interesting would be to examine the impact of social networks on performance of nascent and established entrepreneurs.These aspects of networking and its role for entrepreneurial entry require further research.

Figure 2 .
Figure 2. Theory of planned behaviour as applied to entrepreneurship entry.Source:Ajzen (1991), adapted by authors.

Table 1 .
Probability of intentions to and involvement in entrepreneurial entry (average marginal effects) .1.2.Self-assessed start-up skills, perception of good opportunities and social networks are key predictors of women's entrepreneurial pursuits.Re-estimation of Model 7 only for women (see Model 9 in Table 4

Table 2 .
Effect of an entrepreneurial network (ENET) on women's involvement in start-up efforts in countries with different levels of gender equality (average marginal effects) Standard errors in parentheses.Note 2: ÃÃÃ p < .01;ÃÃ p < .05;Ã p < .1.Note 3: Gender equality is approximated by the average Global Gender Gap index during the period 2011-2018.The first quartile is thus represented by 25% of countries in our sample (i.e. 13 countries) with the lowest average level of the Global Gender Gap index, 4 th quartile by 25% of the most genderequal countries.