Small Business Economics

, 37:417

How do female entrepreneurs perform? Evidence from three developing regions

Authors

    • PRMGE World Bank
  • Shwetlena Sabarwal
    • AFTED World Bank
  • Katherine Terrell
    • University of Michigan
Article

DOI: 10.1007/s11187-011-9374-z

Cite this article as:
Bardasi, E., Sabarwal, S. & Terrell, K. Small Bus Econ (2011) 37: 417. doi:10.1007/s11187-011-9374-z

Abstract

Using the World Bank Enterprise Survey data, we analyze performance gaps between male- and female-owned companies in three regions—Eastern Europe and Central Asia (ECA), Latin America (LA), and Sub-Saharan Africa (SSA). Among our findings are significant gender gaps between male- and female-owned companies in terms of firm size, but much smaller gaps in terms of firm efficiency and growth (except in LA). Part of the reason women run smaller firms is that they tend to concentrate in sectors in which firms are smaller and less efficient (in ECA and SSA). By contrast, we find no evidence of gender discrimination in access to formal finance in any of the three regions, although in ECA women are less likely than men to seek formal finance. Finally, while female entrepreneurs receive smaller loans than their male counterparts, the returns from each dollar they receive is no lower in terms of overall sales revenue.

Keywords

EntrepreneurshipGenderFinanceLatin AmericaEastern Europe and Central AsiaSub-Saharan Africa

JEL Classifications

D24J16L25L26M21O16O54

1 Introduction

The performance gap between female and male entrepreneurs is a focal point of the increasing literature on female entrepreneurship. This is because, if inefficiencies exist that create such a gender differential, it would imply that countries are not fully utilizing their human and physical capital and that the repercussions on their (and their country’s) growth potential are negative. It then becomes important to assess the causes of this as well as possible policy responses.

The relationship between gender and entrepreneurial performance is intriguing, partly because of the opposing perspectives on the subject. The ‘constraint-driven gap’ perspective argues that there are substantial gender-specific barriers to entrepreneurship that constrain the performance of female entrepreneurs. These barriers relate to difficulties that women might face in obtaining credit, in cultivating business networks, in dealing with government and other officials, etc. Many of these obstacles might stem from existing cultural norms that restrict the mobility of women or seclude them in a male-dominated arena. By contrast, while the ‘preference-driven gap’ perspective does not deny that differences in business performance may exist between male- and female-owned enterprises, it also argues that they are driven by fundamental differences in the motivations and approaches that male and female entrepreneurs have towards their businesses.

Based on these perspectives, it is possible to hypothesize about the reasons for gender gaps in performance. On the one hand, personal or environmental constraints that place women at a disadvantage could translate into female under-performance in entrepreneurship. Similarly, if women opt for smaller (and potentially sub-optimally smaller) enterprises—for example, a negative gap in performance could be explained by a desire to better accommodate the demands of the family on their time or by their ‘excessively low’ propensity to risk.

On the other hand, it is also possible (precisely because of the greater constraints they face) that those women who opt to become entrepreneurs might be more skillful than their male counterparts, implying that women who become entrepreneurs might represent a ‘very select’ sub-group in terms of their innate abilities, motivation, and creativity—more so than most male entrepreneurs. This selection effect would account for the higher levels of performance among female, as opposed to male, entrepreneurs, which can totally or partially compensate for the negative impact of constraints or preferences.

Empirically, there has been little rigorous research on the subject, particularly in developing countries. A large proportion of entrepreneurship research in economics has tended to focus exclusively on male entrepreneurs (Brush 1992), thereby completely ignoring the relevant phenomenon of female business-ownership. The studies that have investigated female entrepreneurship have mostly focused on developed countries and used small surveys that are usually not representative of the country. However, in recent years there has been a promising change, and research into female entrepreneurship has accelerated and deepened (as documented in Minniti 2009; Klapper and Parker 2010). In this paper, we provide one of the first comprehensive analyses of entrepreneurial performance by gender in three regions of the world, namely, Eastern Europe and Central Asia (ECA), Latin America (LA), and Sub-Saharan Africa (SSA). We do so using comparable firm-level data from formal (registered) enterprises. By analyzing these three regions separately, we allow for gender differences in firm behavior to vary across these regions. We calculate the size of the gaps using various measures of performance (sales revenue, value added, total factor productivity, sales and employment growth) and we test several explanations for these gaps: (a) age of the firm; (b) sector concentration, (c) demand vs. supply constraints in accessing formal credit; and (d) gender differences in returns from formal credit.

In line with findings regarding developed countries, our results indicate that, in this large part of the developing world, female-owned enterprises are substantially and significantly smaller than those owned by men. One reason for this is that women are more likely than men to operate in industries where firms are smaller and less efficient. This is true in ECA and SSA, but not in LA. Contrary to our expectations, we did not find evidence of gender-based discrimination in access to formal finance (‘supply constraint’) in any of these three regions—quite a surprising result. After correcting for selection into entrepreneurship, women entrepreneurs are as likely as men to obtain a loan from formal financial institutions. Finally, we tested whether female-owned firms are smaller because women use bank financing less effectively than men. Women’s returns from loans are indeed smaller than those of men, but once we control for the size of the loan, we find that there is no gender-based difference in terms of the impact of loans on overall sales.

The paper proceeds as follows: Sect. 2 contains a review of the literature and Sect. 3 a description of the data. Section 4 provides estimates of various measures of performance gaps and in Sect. 5 we explore four broad explanations for these gaps: gender-based differences in firms’ age, industrial concentration, access to credit, and use of credit; Sect. 6 concludes the paper.

2 Existing research

Studies investigating whether the gender of the entrepreneur affects the performance of the enterprise yield mixed results. Some studies provide evidence of female under-performance (Brush 1992; Rosa et al. 1996), while others do not find gender-based differentials in entrepreneurial performance (Du Rietz and Henrekson 2000; Bardasi et al. 2007). Recent reviews of this literature suggest that most studies do find evidence of performance gaps between female and male entrepreneurs (for example, see Klapper and Parker 2010).

In general, evidence indicates that female- and male-owned enterprises differ in size. Recent evidence from the United States suggests that, on average, male-owned businesses are twice as large as female-owned businesses, in terms of both sales and assets (Coleman 2007). Results also reveal that, on average, firms owned by women generate only 78% of the profits of comparable male-owned businesses (Robb and Wolken 2002). Also, women have been found to generate less sales turnover relative to men, even in the same industrial sector (Loscocco and Robinson 1991; Chaganti and Parasuraman 1996).

Male- and female-owned businesses have also been compared in terms of their survival probabilities. Evidence indicates that, in Dutch businesses, the survival rate of male entrepreneurs’ businesses is greater than that of their female counterparts (Bosma et al. 2004). Similarly, Lohmann and Luber (2004) show that in Germany only 42% of self-employed women remain self-employed after 5 years, while the corresponding rate for male entrepreneurs is 63%.

However, the hypothesis of relative female under-performance is not universally corroborated in entrepreneurship literature. Watson (2002) shows that in Australia women business owners earn similar rates of return on equity and assets to male business owners, but have less start-up capital, which explains their lower incomes and profits, as compared to men. Using World Bank Enterprise Surveys (2002–2006), Bardasi et al. (2007) find that, in Africa, female-owned businesses are at least as productive as those of male entrepreneurs, when measured by value added per worker and total factor productivity. Similarly, Kepler and Shane (2007) show that there are no significant gender differences in terms of the performance outcomes of nascent entrepreneurs. Other studies show that female-owned enterprises do not under-perform in terms of employment creation (Fischer et al. 1993; Chaganti and Parasuraman 1996) or survival rates (Kalleberg and Leicht 1991; Bruderl and Preisendorfer 1998).

The empirical literature on explanations for gender-based gaps in entrepreneurial performance can be organized under the two main headings above, namely, constraint-driven gaps and preference-driven gaps (see also, Klapper and Parker 2010).

2.1 Constraint-driven gaps

Barriers to female entrepreneurship can arise from existing cultural and institutional structures, both formal and informal. Coate and Tennyson (1992) have noted that it is possible for labor market discrimination to spillover into markets that are relevant to self-employment. This discrimination would become further exacerbated if entrepreneurial ability is perceived to be signaled by earlier investments in human capital (Cressy 1996). Mayoux (1995) documents some of the most common obstacles faced by women entrepreneurs, which include barriers in access to bank credit, problems dealing with male officials because of norms of female seclusion in some parts of the world, and difficulties in accessing information, because it is seen to be channeled predominantly through male networks.

It has been hypothesized that observed gender differences in entrepreneurial performance may stem from discrimination against female entrepreneurs in accessing finance. While in Ethiopia access to finance is an acute problem for both men and women, Bardasi and Getahun (2009) find that the performance of female entrepreneurs is particularly negatively affected by access to and the costs of finance. Several studies suggest that raising capital is more difficult for women than for men (Brush 1992; Carter and Cannon 1992; Carter 2000). Using data from Europe and Central Asia, Muravyev et al. (2009) find that female-managed firms have a 5.4% lower probability of securing a bank loan than male-managed firms. This study also evaluates the existence of financial constraints by looking at interest rates and finds that female-managed firms pay an average of 0.6% higher interest rates than those of their male counterparts. Both these factors indicate discrimination against female entrepreneurs and the authors suggest that the propensity of such discrimination is found to be higher in the least financially developed countries in the region. This is corroborated by Aidis et al. (2007) who, using original survey data from Lithuania and Ukraine, show that access to funds is a more relevant barrier for female business owners than it is for their male counterparts. In contrast, several studies (Cavalluzzo and Cavalluzzo 1998; Blanchflower et al. 2003; Storey 2004; and Cavalluzzo and Wolken 2005) find no statistically significant gender bias in terms of access to finance.

Alternatively, significant differences in male and female access to finance may be accounted for by differences in other characteristics affecting their credit worthiness, including human capital factors, personal wealth, etc. For instance, Watkins and Watkins (1984) claim that, in Britain, women entrepreneurs are more likely to start a business without having a demonstrable record of achievement, vocational training and experience, than their male counterparts. Brush (1992) argues that men are more likely than women to have an education and experience, emphasizing technical and managerial elements which might impact their entrepreneurial performance. Women may have more difficulties in securing a loan than males, because they tend to start smaller businesses, concentrate in the service sector and work part-time, all of which may not encourage banks to lend to women (Verheul and Thurik 2001).

2.2 Preference-driven gaps

It has been argued that the reasons for becoming an entrepreneur differ by gender (Delmar and Davidsson 2000; Boden 1999; Shane et al. 1991). The desire to effectively combine work and family responsibilities often motivates women to start their own businesses as the option offers flexible work arrangements. It has been illustrated that women, especially women with young children, cite flexibility of work schedule and other family-related reasons for becoming self-employed, which is not the case of men (Boden 1999). Using Contingent Work Survey data from the United States, Boden (1999) shows that having young children positively and significantly affects the probability of women becoming self-employed, while no such effect is apparent in the case of males. Data from the Unites States shows that self-employed women either have very long or very short working weeks, implying that female, self-employed workers show greater levels of dispersion in working hours than other groups (Devine 1994). Similarly, Lombard (2001), using CPS (Current Population Survey) data for the Unites States, observes a positive relationship between women’s demand for flexibility (in terms of variability in working hours) and participation in self-employment, with this relationship being strongest among women with small children. This would imply that when wage employment provides flexible work arrangements and family-related support mechanisms, levels of entrepreneurship among women will decline. In their analysis of GEM data for 29 countries, Kovalainen et al. (2002) observe a negative relationship between the statutory maternity leave in days and the rate at which women start their own businesses. Meanwhile, Verheul et al. (2004), also using GEM data, found that the importance of the family is positively linked to entrepreneurship for both men and women.

Verheul et al. (2004) also find that satisfaction with life (answer to the question, “How satisfied are you with life?”) is positively and significantly linked to entrepreneurship only in the case of women. Using original survey data from Lithuania and Ukraine, Aidis et al. (2007) show that, although ‘independence’ is cited as an important motivation for starting one’s own business, in the cases of both women and men, women are more likely to cite necessity and other ‘push’ factors (such as the need to supplement household income) as important reasons for this choice. Men, however, are more likely than their female counterparts to cite ‘pull’ factors (availability of resources, opportunity to increase income, etc.) as primary motivations for starting their own businesses. In line with the points made above, evidence from Italy shows that men are more likely to enter into self-employment after being laid off or if they seek career advancement, while women are more likely to become self-employed after a period of inactivity or unemployment (Rosti and Chelli 2005).

Women may prefer to take fewer risks than men, yet the propensity to take risk has been considered an important predictor of entrepreneurial success (Schumpeter 1939; Evans and Leighton 1989; Earle and Sakova 2000), and some papers show that women tend to experience higher rates of risk aversion than men (Jianakoplos and Bernasek 1998; Barber and Odean 2001; Dohmen et al. 2006). These differences could have important implications for business performance, if higher risk aversion leads women to restrict investment in their business ventures. In contrast, Masters and Meier (1988) found that female entrepreneurs are more similar than different to male entrepreneurs in their risk-taking propensity.

Preference gaps can also arise in industry selection. When comparing the performance of male and female entrepreneurs at the macro-level, it becomes imperative to take into account their relative sectoral concentrations. It has been suggested that female entrepreneurs are disproportionately concentrated in the small scale sector, which might explain existing gender gaps in entrepreneurial performance, at least in part. Mayoux (1995) claims that, “Women are overwhelmingly clustered in a narrow range of low investment, low profit activities for the local market.” Women are also seen to be heavily concentrated in the service sector, both as entrepreneurs (Bates 1995), and as employees (Verheul et al. 2004), while certain industrial sectors, like construction, remain heavily dominated by men (Bates 1995). Also, women have been revealed to be less likely than men to conduct business in high-technology sectors (Loscocco and Robinson 1991; Anna et al. 2000). In addition, it has been suggested that the differences between female and male entrepreneurs’ choice of sector and product/service (Fischer et al. 1993; Brush 1992; Chaganti and Parasuraman 1996) could be linked to gender gaps in opportunities.

Hundley (2001) claims that women’s choices, with respect to the industrial sector, can be important in explaining gender differences in entrepreneurial performance. He shows that, in the Unites States, the higher propensity of women to work in low value added sectors explains about 9–14% of the gender-based self-employment earning differential. This was largely due to the concentration of women working in the personal services sector and their under-representation in the more lucrative professional services and construction industries.

3 Data

In this paper we use data from three developing regions, namely, Eastern Europe and Central Asia (27 countries), Latin America (13 countries), and Sub-Saharan Africa (22 countries).1 Data for Eastern Europe and Central Asia (ECA) come from the 2005 Business Environment and Enterprise Performance Survey (BEEPS) data, produced by the World Bank and the European Bank for Reconstruction and Development (EBRD). Data for Latin America (LA) and Sub-Saharan Africa (SSA) come from the 2006 and 2007 World Bank Enterprise Surveys. By adhering to similar sampling techniques and questionnaires, these data sources yield comparable enterprise-level data.

The samples are constructed using a process of stratified random sampling from a national registry of firms, which implies that only registered firms (i.e., formal firms) are included in the sample. Further, the sampling methodology for the survey generates samples that are representative of the manufacturing and service sectors as a whole. The sample of firms in each country is stratified by size, sector and location, using simple random sampling or random stratified sampling. In the case of large economies, firms are stratified at the two digit industry level. In the case of small economies, there may not be enough firms to stratify at the two digit level; in that case, a sample of firms is randomly selected from the manufacturing and retail sectors as well as from the rest of the economy. In each country, the sectoral composition of the sample, in terms of ‘manufacturing’ versus ‘services,’ is determined by their relative contribution to GDP. Firms that operate in sectors subject to government price regulation and prudential supervision, such as banking, electric power, rail transport, water and waste water, are not surveyed. (See www.enterprisesurveys.org for further details on the number of surveys and their design.)

The data enable us to identify the sex of the principal owner of privately held shareholding companies, partnerships, and sole proprietorships. Hence in this paper we define male versus female entrepreneurs as “male versus female sole or principal owner of privately held shareholding companies, partnerships and sole proprietorships.” From our perspective, other strengths of these data include the fact that the same survey instrument was administered in a number of developing countries from different regions, which allows for cross-regional comparisons. In addition, the data include a host of performance variables for each firm, as well as variables capturing institutional factors (especially in the area of finance) that may affect the relative performance of male- and female-owned businesses. The data have some weaknesses too, including (a) the small number of firms sampled in most countries; (b) the inability to identify the sex of the other owners of the firm when there is more than one; (c) the lack of information on the precise managerial involvement of the main or sole owner; (d) the lack of demographic information on the entrepreneurs; and (e) the numerous missing answers to some variables of interest (e.g., capital). In the subsequent analysis and conclusions we will propose robustness tests to assess some of the data limitations and discuss their likely implications.

In the empirical analysis we pool country level data to construct region-specific datasets. We do not pool data for different regions because we believe there are important cultural and institutional differences across these regions.2 Our analytical sample consists of 4,903 firms for ECA, 7,393 firms for LA, and 8,233 firms for SSA.3 The sample is created as follows: first, from the initial dataset only firms that are privately held companies, partnerships and sole proprietorships are retained (public and foreign enterprises are excluded). Next, firms that have missing information on the sex of the principal owner (or owners), sales, or the number of permanent employees are dropped. Finally, to control for outliers, we drop from the sample those firms that are in the top 0.1% of the regional sale distribution.4 We also drop firms that are more than 75 years old (37 firms in ECA, 142 in LA, and 12 in SSA).5 Some basic firm characteristics in the analytical sample are summarized by the region and sex of the entrepreneurs in the Appendix (Table 7).6

4 Performance gaps

Note that, in each region, female entrepreneurs are a minority, accounting for about 27% of the firms in ECA, 37% in LA and 27% in SSA. The share of entrepreneurs who are women is very similar in high-income countries—38.8% in the Unites States, 30.4% in the United Kingdom, and 32.5% in Sweden, to mention a few examples (Verheul and Thurik 2001). Quite interestingly, despite the relatively egalitarian position of men and women in Swedish society, the gap in female entrepreneurship in this country is very similar to that observed in other high income economies. While our analysis does not focus on female participation in entrepreneurship, it is important to observe that selection into entrepreneurship—quite likely gender-specific—impacts the observed gaps in performance.

We measure performance gaps between male- and female-owned firms in a number of ways: in terms of firm size (total revenue), growth (sales and employment growth), and efficiency (value added per worker and total factor productivity).

In general, female entrepreneurs fare worse than their male counterparts in terms of these measures of performance. Table 1 presents the coefficients of the dummy for female ownership, estimated using OLS regressions where the dependent variable is a measure of performance, as indicated above. The regressions are region-specific and the number of control variables increases progressively (from column 1 to column 3). Controlling for country and sector (column 3), the yearly amount of sales revenue of the average female entrepreneur is significantly less than for her male counterparts (A). The gap is especially large for ECA (−38% less with respect to male entrepreneurs) and smaller for SSA (−12%).7
Table 1

Performance gaps between male- and female-owned businesses (OLS regressions coefficients of female-owned dummy)

Variable

ECA

LA

SSA

(1)

(2)

(3)

(1)

(2)

(3)

(1)

(2)

(3)

A. Ln (sales revenue)

 Female-owned firm

−0.608***

−0.664***

−0.489***

−0.334***

−0.299***

−0.319***

−0.049

−0.268***

−0.128***

 SE

(0.053)

(0.050)

(0.049)

(0.052)

(0.040)

(0.039)

(0.048)

(0.044)

(0.042)

 Observations

4861

4861

4861

7084

7084

7084

8233

8233

8233

 R-squared

0.03

0.17

0.23

0.01

0.68

0.70

0.00

0.17

0.28

B. Sales growth

 Female-owned firm

1.486**

1.111*

1.094*

−8.439***

−9.234***

−8.319***

−1.108

0.235

0.724

 SE

(0.600)

(0.626)

(0.632)

(1.569)

(1.501)

(1.510)

(1.434)

(1.392)

(1.421)

 Observations

3627

3627

3627

5744

5744

5744

5947

5947

5947

 R-squared

0.00

0.03

0.03

0.01

0.15

0.20

0.00

0.09

0.09

C. Employment growtha

 Female-owned firm

−2.736**

−1.914

−0.898

−1.148

−1.448

−2.437**

−2.485**

−1.492

−1.518

 SE

(1.268)

(1.273)

(1.317)

(1.045)

(1.023)

(0.998)

(1.104)

(1.101)

(1.120)

 Observations

4807

4807

4807

6688

6688

6688

6716

6716

6716

 R-squared

0.00

0.02

0.03

0.00

0.08

0.11

0.00

0.03

0.04

D. Value added per worker

 Female-owned firm

−0.859**

−1.357***

−1.012***

−1.216***

−1.490***

−1.728***

0.215**

−0.088

0.064

 SE

(0.377)

(0.232)

(0.238)

(0.307)

(0.212)

(0.217)

(0.102)

(0.094)

(0.095)

 Observations

4562

4562

4562

5332

5333

5333

5334

5334

5334

 R-squared

0.00

0.54

0.55

0.00

0.39

0.40

0.00

0.19

0.24

E. Total factor productivity

 Female-owned firm

−0.014

−0.014

−0.022**

−0.001

−0.208***

−0.158***

0.015

−0.003

0.007

 SE

(0.010)

(0.010)

(0.009)

(0.040)

(0.025)

(0.023)

(0.013)

(0.012)

(0.012)

 Observations

3087

3087

3087

4357

4357

4357

5076

5077

5077

 R-squared

0.98

0.98

0.98

0.79

0.92

0.93

0.96

0.97

0.97

Notes: Coefficients of a dummy for female ownership shown in the table. Specification (1) only includes a dummy for female-owned firms; (2) includes also country fixed effects; (3) includes also country and industry fixed effects. Robust standard errors are in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%. Value added is defined as sales revenue minus cost of materials. Sales and employment growth are expressed in percentage terms.

ECA Eastern Europe and Central Asia, LA Latin America, SSA Sub-Saharan Africa

aNormalized employment growth, as explained in the text (Sect. 4)

We also examine gender gaps in firm growth over a 3-year period, both in terms of employment and sales. We measure growth as the average annual percentage change over the previous 3 years: \( G = \frac{{\left( {X_{t} - X_{t - 3} } \right)}}{{X_{t - 3} }} \), where X = sales (or number of permanent employees). We also use the ‘normalized’ growth measure proposed by Davis and Haltiwanger (1992): \( g = \frac{{X_{t} - X_{t - 3} }}{{\frac{{X_{t} + X_{t - 3} }}{2}}} \).8

Using either of these growth measures, we find that—when once we control for country and industrial sector—female-owned firms do significantly worse than their male-owned counterparts, in terms of sales growth (B), and employment growth (C) only in the case of LA. A lack of noticeable gender gaps in sales and employment growth in the two other regions (SSA and ECA) could be a reflection of the fact that female-owned enterprises may start from a lower base.

With respect to productive efficiency, we ask—are female-owned businesses less productive than those owned by males, in terms of the revenue that they generate from given inputs? We analyze this issue using two firm level measures: (1) value added (defined as sales minus intermediate goods) per worker; and (2) total factor productivity (TFP). TFP is obtained from estimating a Cobb-Douglas production function with pooled firm-level data from all countries available for a given region:
$$ \ln Y_{ijz} = \alpha^{K} \ln K_{ijz} + \alpha^{L} \ln L_{ijz} + \alpha^{M} \ln M_{ijz} + \delta F_{ijz} + I_{jz}^{'} \varphi + C_{z}^{'} \eta + \varepsilon_{ijz} , $$
(1)
where lnYijz is the log of sales revenues of firm i operating in industry j and in country z. The inputs include: K, capital stock (at replacement value); L, labor (number of permanent employees) and M, intermediate material input. F is a dummy variable equal to one for a female entrepreneur, I is a set of industry fixed effects and C is a set of country fixed effects.9 The estimated δ coefficient of F in Eq. 1 measures the gender gap in TFP.

The findings for these two measures of efficiency, presented in Table 1D and E, indicate that, when we control for country and sector, the average female-owned firm is significantly less efficient in ECA and LA but not necessarily in SSA, where there is no significant difference in either value added or TFP. The ‘male–female entrepreneurial’ gaps in efficiency are greatest for female entrepreneurs in LA.

On the whole, evidence of female underperformance in these three areas—size, growth and efficiency—is found most consistently in LA. In contrast, female-owned firms in SSA are smaller, but no less efficient or growth-oriented. Evidence from ECA suggests large gaps in firm size, but small gaps in firm efficiency and no gaps in firm growth rates.

It might be argued (see paper by Aterido and Hallward-Driemeier in this same issue) that the manager, rather than the principal owner, is the decision-maker and the one responsible for the performance of the firm—so that, in our analysis, we should consider the sex of the manager instead. We are able to test the extent to which this distinction matters using a sub-sample of enterprises, namely, those that are sole proprietorships (firms with just one owner, who is also the manager). Unfortunately, only in ECA and Nigeria is it possible to distinguish between sole proprietorship and partnership enterprises (with multiple owners and, typically, a more complex management structure, so that there is no guarantee that the principal owner, of whom we know the sex, is also the manager). When we limit the sample to sole proprietorship enterprises in ECA and Nigeria and re-estimate the performance gaps, as in Table 1, we find that the coefficient of the female ownership dummy (gender gaps in performance) for the whole sample are very similar to those estimated for the sample which only includes firms in which the owner is also the manager (compare column 1, Sole proprietorship, with column 1, Whole sample, and column 2, Sole proprietorship, with column 2, Whole sample, in Table 8 of the Appendix). Note that the share of owner-managers to total owners is very high in ECA (approximately 85%) and Nigeria (approximately 91%). Should this pattern hold in LA and the rest of the SSA region, our results would not be affected by the choice of the entrepreneur (whether the owner or the manager) simply because, in the lion’s share of firms, where the principal owner is a woman, she would also be the principal manager. While the robustness test presented in Appendix Table 8 does not allow us to determine whether it is the sex of the owner or of the manager that impacts the firm’s performance, we are confident that the gender gaps presented in this paper apply to the vast majority of firms in developing countries, and are associated with female (or male) ownership and management (by the same individual).10

The central question that arises from the preceding analysis is—why are women-owned enterprises consistently smaller than those of men in this large part of the developing world? Moreover, the difference in efficiency in ECA and LA, as measured by value added per worker and TFP, also needs to be explained. Finally, why do female entrepreneurs in LA consistently under-perform across the board, compared to their male counterparts? In efforts to address this, we therefore ask what explains the differing patterns of relative performance of female entrepreneurs across the three regions. We attempt to answer these questions within the constraints imposed by the data.

5 What explains the smaller size and lower efficiency of female-owned firms?

In this section we consider alternative explanations for the observed gender gaps in firm performance. We answer the following questions: Are gender gaps in performance driven by gender differences in: (a) the age of the firm (Sect. 5.1); (b) concentration of women entrepreneurs in specific industrial sectors (Sect. 5.2); (c) access to formal credit (Sect. 5.3); and (d) the use of credit (Sect. 5.4)? The first question may be answered in several ways—if evidence indicates that the average female-owned firm performs less well than its male-owned counterpart because it is younger, then this could be because of gender-specific mechanisms which influence whether people become entrepreneurs or leave the field. It could also be the result of men and women’s diverse preferences and/or the different ways in which the institutional context and the business climate impact the two groups. The second issue (sector selection) may relate to the ‘preference-driven gap’ perspective, although it may also stem from issues which make it difficult to penetrate the entrepreneurial world and which force women into specific industries. The last two issues mostly concern the ‘constraint-driven gap’ perspective.

5.1 Gender differences in age of firm

Differences in the age of female- and male-owned firms may, at least partially, explain the existence of a gender gap in firms’ performance. If more women are increasingly becoming entrepreneurs around the world, their firms will be younger than those of their male counterparts. They could hence have less years of experience in leadership positions, which might explain their firms’ lower levels of performance (e.g., smaller size or lower value added per worker). However, differences in the age of male- and female-owned firms may also be an outcome (rather than a source) of differences in their performance. If women’s firms do not survive at the same rate as male-owned firms, we would expect their age distribution to be more skewed toward younger firms than is the case for male-owned firms. Using cross-section data—as we do in this paper—does not allow us to distinguish between the two hypotheses; however, it is reasonable to assume that younger firms are, on average, less experienced and potentially ‘less selected’ than older firms. Two caveats are worth mentioning. First, the age of the firm does not necessarily correspond to the experience of the entrepreneur, so younger firms may in fact embed a large amount of experience and know-how. Second, the ‘selection’ of entrepreneurs, as expressed by their survival rate, is not independent of their original decision to become entrepreneurs (i.e., it depends on the entry barriers into entrepreneurship, which can be very different for men and women).

We find that, on average, female-owned enterprises are indeed younger than male-owned enterprises in ECA (by 1 year) and especially in LA (by more than 3 years), while there is no significant difference in SSA (Appendix, Table 7).

To account for potential differences in performance that can derive from differences in the age of the firm, we re-estimate the regressions shown in Table 1, controlling for the age of the firm (in addition to country and sector fixed effects). The results, presented in Table 2, indicate that differences in the age of the firm—where they exist—explain very little, or nothing at all, of the differences in performance between male- and female-owned enterprises, including in LA, where differences in the age of male versus female entrepreneurs are the most remarkable. When including age among the regressors, the gender gaps in performance (the size of the coefficient of the female ownership dummy) barely change (compare column (1) and (2) in Table 2). The explanation of gender gaps in performance therefore has to be found elsewhere.
Table 2

Performance gaps between male- and female-owned businesses, controlling for age of the firm

Variable

ECA

LA

SSA

(1)

(2)

(1)

(2)

(1)

(2)

Ln (sales revenue)

 Female-owned firm

−0.489***

−0.464***

−0.319***

−0.346***

−0.128***

−0.100**

 SE

(0.049)

(0.048)

(0.039)

(0.038)

(0.042)

(0.041)

 Observations

4861

4861

7084

7084

8233

8207

 R-squared

0.23

0.27

0.70

0.71

0.28

0.33

Sales growth

 Female-owned firm

1.094*

1.074*

−8.319***

−8.872***

0.724

0.571

 SE

(0.632)

(0.634)

(1.510)

(1.485)

(1.421)

(1.428)

 Observations

3627

3627

5744

5744

5947

5925

 R-squared

0.03

0.03

0.20

0.19

0.09

0.1

Employment growtha

 Female-owned firm

−0.898

−1.466

−2.437**

−1.994**

−1.518

−1.918*

 SE

(1.317)

(1.307)

(0.998)

(0.971)

(1.120)

(1.113)

 Observations

4807

4807

6688

6688

6716

6695

 R-squared

0.03

0.04

0.11

0.15

0.04

0.06

Value added per worker

 Female-owned firm

−1.012***

−1.011***

−1.728***

−1.879***

0.064

0.076

 SE

(0.238)

(0.238)

(0.217)

(0.218)

(0.095)

(0.095)

 Observations

4562

4562

5333

5333

5334

5321

 R-squared

0.55

0.55

0.40

0.4

0.24

0.24

Total factor productivity

 Female-owned firm

−0.022**

−0.022**

−0.158***

−0.137***

0.007

0.007

 SE

(0.009)

(0.009)

(0.023)

(0.016)

(0.012)

(0.012)

 Observations

3087

3087

4357

4342

5077

5064

 R-squared

0.98

0.98

0.93

0.97

0.97

0.97

Notes: Specification (1) corresponds to specification (3) of Table 1, that is, it includes a dummy for female-ownership and country and industry fixed effects. Specification (2) includes also age of the firm. Robust standard errors are in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%. Value added is defined as sales revenue minus cost of materials. Sales and employment growth are expressed in percentage terms.

ECA Eastern Europe and Central Asia, LA Latin America, SSA Sub-Saharan Africa

aNormalized employment growth, as explained in the text (Sect. 4)

5.2 Do female entrepreneurs concentrate in low-productivity sectors?

The literature has hypothesized that the poorer performance of female-owned businesses can be attributed to the fact that women-owned firms are disproportionally found in ‘poorly performing’ industries. The question is then whether these ‘female sectors’ are inherently comprised of smaller, less efficient firms or whether this is an endogenous outcome as women—whose firms are less efficient than those of their male counterparts, on average—are ‘crowded’ into a few sectors.11 We investigate this hypothesis by testing, first, whether female-owned firms are disproportionately concentrated in certain industries. Second, we test whether the firms in these industries tend to be smaller or less efficient than in other industrial sectors. Finally, we assess whether gender-differences in performance are observed within types of sectors (female- and male-dominated).

It is important to note that the available sectoral disaggregation in LA is slightly different than what can be constructed for ECA and SSA, because a different industrial classification was used in LA countries. Even so, we find some consistent patterns in the relative sectoral concentration of female entrepreneurs across the three regions. In Fig. 1A we represent the distribution of women entrepreneurs across industrial sectors. Female-owned firms are present in all industries. However, very few are in electronics or IT industries while a very large percentage (nearly 30% in ECA and SSA) are in the ‘wholesale and retail trade.’ In LA, ‘food preparation’ is the dominant industry for women—about 25% of all women entrepreneurs operate in this sector. The distribution of women across industries not only highlights the sectors in which they work, but may also reflect the importance of that economic activity to the economy overall. In order to assess whether women’s businesses are focused disproportionately (‘crowded’) in certain sectors, we calculate an index of concentration (Fig. 1B), defined as the percentage of all entrepreneurs in a sector who are female, divided by the percentage of all entrepreneurs in the region who are female. Values greater (lower) than one indicate that, in that sector, there are disproportionately more (less) female entrepreneurs than in the whole economy at large. In ECA and SSA (where the industrial classification system is the same) female entrepreneurs are overrepresented (i.e., the concentration index has values >1) in the following industrial sectors: garments and leather goods, hotels and restaurants, wholesale and retail trade, other services, and textiles. Though the LA region is classified differently, the pattern is similar to that of the other two regions—women are overrepresented in the retail trade and the garments sector. However, unlike in ECA and SSA, in LA women are also overrepresented in the production of machinery and equipment as well as in the food industry. The main message of Fig. 1 is that women do indeed concentrate in a few sectors, while men are distributed across a broader spectrum of industries (i.e., in the case of men, the concentration index is >1 in a larger number of sectors).
https://static-content.springer.com/image/art%3A10.1007%2Fs11187-011-9374-z/MediaObjects/11187_2011_9374_Fig1_HTML.gif
Fig. 1

Concentration of female entrepreneurs in industrial sectors, by region. A Distribution of female entrepreneurs across industries (%, values sum to 100). B Index of concentration of female entrepreneurs by industry. Note: The index of concentration is defined as the ratio between the percentage of women entrepreneurs in a specific sector and the average percentage of women entrepreneurs in the whole region

Can the relatively poor performance of female-owned firms be attributed to the lower average performance of the sectors in which women concentrate? Given the potential endogeneity issue—i.e., that women entrepreneurs’ (lower) performance is what drives the lower average performance of the sectors where they concentrate—we measure the relative performance of the various industries by the performance of male-owned firms. Hence, we first test whether the performance of male-owned firms in the female-dominated industries is lower than that of male-owned firms in other industries, which are not dominated by female-owned firms.

To distinguish between the two hypotheses of the “intrinsically” lower performance of the sector and the lower performance of women within these sectors, we group the industrial sectors in two categories—one including the female-dominated sectors (those where the index of concentration is larger than 1) and one including the male-dominated sectors (those where the index of concentration is smaller than 1)—and we analyze the relative performance of male- and female-owned businesses using the following specification:
$$ \ln Y_{ikz} = \alpha_{0} + \delta F_{ikz} + \phi FD_{kz} + \gamma F_{ikz} * FD_{kz} + C_{z}^{'} \eta + \varepsilon_{ikz} . $$
(2)
where lnYijz is the log of sales revenues (or value added) of firm i operating in sector type k (female-dominated or male-dominated) and in country z. The coefficients of interest are δ, which indicates the performance of female-owned firms overall; φ, capturing differences in performance between firms in female-dominated (FD = 1) and male-dominated (FD = 0) sectors; and γ, the additional effect associated with female-owned firms operating in a female-dominated sector. Country-specific effects are controlled for in vector C.
The results of estimating regression (Eq. 2) for firm performance, measured in terms of size (sales) and efficiency (value added per worker), are presented in Table 3. Do women concentrate in ‘poorly performing’ industrial sectors, but would otherwise do as well as male-owned firms? The answer varies depending on the region and the indicator of performance. In ECA and SSA, it is indeed the case that firms operating in sectors that are female-dominated are significantly smaller than those that operate in male-dominated sectors (−41% in ECA and −56% in SSA)—which implies that women do concentrate in sectors where firms are on average smaller. In both regions, however, the concentration story does not appear to be the whole explanation of why female-owned firms are smaller. In line with the results shown in Tables 1 and 2, even after controlling for the entrepreneurs’ choice of sector, we estimate a negative gap associated with the fact that the firm is owned by a woman. In ECA, the gender gap in sales revenue exists across industrial sectors (−24% on average), but it is larger in female-dominated industries, where female-owned firms are about 60% smaller than female-owned firms operating in male-dominated sectors and about 70% smaller than male-owned firms in male-dominated sectors. In SSA, however, female-owned firms that operate in male-dominated industries are as large as their male-owned counterparts, but firms in female-dominated sectors are smaller on average (−56%) and in those sectors there is also a gender-specific gap (−28% within this category). In other words, in SSA, female-headed firms appear to be more polarized in terms of sales. As for value added, in ECA the concentration story is much weaker—a negative, but not significant, coefficient is estimated for industrial sectors that are female-dominated. However, women entrepreneurs in sectors with a higher concentration of women do significantly worse than male and female entrepreneurs in male-dominated industries; it is this effect that drives the under-performance of women on average. In SSA, meanwhile, the gender gap in value added appears to be driven by the concentration of women in sectors where all firms (including those which are male-owned) have a lower value added, as is the case of sales revenue. LA represents an entirely different case. In LA women entrepreneurs appear to concentrate in sectors where the average firm size and average value added are higher than those of their male counterparts. However, when the results are analyzed by sector, female-owned firms are smaller and produce lower value added than male-owned enterprises—and this is especially true in female-dominated industries.12
Table 3

Performance gaps between male- and female-owned businesses, controlling for operating in female-dominated sectors

Variable

Ln (sales revenue)

Value added per worker

ECA

LA

SSA

ECA

LA

SSA

Female-owned firm

−0.273***

(0.080)

−0.107*

(0.060)

0.09

(0.068)

−0.202

(0.381)

0.09

(0.335)

0.207

(0.139)

Female-dominated sector

−0.529***

(0.051)

0.498***

(0.047)

−0.830***

(0.045)

−0.397

(0.245)

1.302***

(0.252)

−0.821***

(0.096)

Female-owned firm*female-dominated sector

−0.399***

(0.101)

−0.393***

(0.078)

−0.324***

(0.088)

−1.544***

(0.481)

−2.477***

(0.420)

−0.22

(0.187)

Observations

4861

7084

8195

4562

5333

5313

R-squared

0.20

0.69

0.23

0.54

0.39

0.21

Notes: Robust standard errors are in parentheses. * Significant at 10%; ** significant at 5%; *** significant at 1%. Value added is defined as sales revenue minus cost of materials. Female-dominated sectors are defined as those where the index of concentration is larger or equal 1 (the index of concentration is defined as the ratio between the percentage of women entrepreneurs in a specific sector and the average percentage of women entrepreneurs in the whole region). The definition of female-dominated sectors is region-specific.

ECA Eastern Europe and Central Asia, LA Latin America, SSA Sub-Saharan Africa

In sum, across all three regions, women entrepreneurs are predominantly working in similar industrial sectors. Moreover, our analysis (the results shown in Table 3) suggests that, both in terms of firm size and value added, there is some support for the concentration story (i.e., women do appear to concentrate in industrial sectors with smaller/less performing firms) for ECA and SSA—but not for LA. In ECA there is also a substantial gender gap within sectors (for sales in particular), while in SSA the concentration effect is definitely the most relevant. This evidence is powerful as it indicates that part of the puzzle surrounding female under-performance lies with their selection of the sector in which they work. There are, however, significant gender differences in performance within ‘sector types’—and these are particularly large in female-dominated industries.

As far as the concentration story goes, the interesting policy question is whether women are ‘pulled’ or ‘pushed’ into industrial sectors that tend to comprise smaller and often less productive firms. If this is a question of choice, what are the features of these sectors that make them attractive to women? Because these are ‘low performing’ sectors (in terms of size and, at times, value added), is there any policy rationale for trying to change female entrepreneurs’ preferences in terms of sectoral choice? And if this is not a choice, but women are ‘forced’ to operate in these sectors, what are the key elements that exclude them from the other sectors? Due to a lack of information on entrepreneurial preferences and motivations in our data, we cannot test many of the hypotheses that have been offered to answer these questions,13 but we are able to test, in the following section, to what extent these choices may be constrained by access to credit.

5.3 Are female entrepreneurs more constrained in accessing credit?

Not surprisingly, access to formal credit is considered to be one of the most important predictors of entrepreneurial success and survival. In recent years, some empirical research has shown that female-owned businesses tend to have less access to formal credit (Carter and Rosa 1998; Coleman 2007) than male-owned businesses. If this is true, then gender gaps in access to formal credit could make it difficult for women to penetrate industries where firms require larger investments (which therefore remain male-dominated). Moreover, gender gaps in access to credit may also provide a crucial explanation of gender-based gaps in entrepreneurial performance, even within industrial sectors.

The sex of the entrepreneur could affect both the demand for and supply of credit on the part of the banking institution. Female entrepreneurs may be less likely to apply for bank loans than male entrepreneurs (if, for example, they are more risk averse). They may also be less likely than male entrepreneurs to obtain bank loans if, for example, there are cultural barriers (in terms of property rights or perhaps discrimination) or women’s firms are supposedly considered to be less creditworthy.

Our data include a host of detailed questions on firms’ access to finance and credit which enable us to carry out rigorous tests of the hypothesis that: (a) women are less likely than men to demand a loan; and (b) women are less likely to obtain a bank loan. (See Table 9 in the Appendix for gender-disaggregated summary statistics of access to credit variables.)

In dealing with the question of access to credit, we need to address selection issues. Ultimately, we are interested in explaining the probability of obtaining formal credit and its relationship to the sex of the entrepreneur; however, there is a group of entrepreneurs that do not apply for formal credit because they do not need external financing. Further, data reveals that there is also a group of entrepreneurs that need a loan but do not apply for a number of reasons. For these two groups we do not observe the probability of obtaining a loan. In other words, the observed sample that applies for formal loans is a self-selected, non-random sub-sample of the total population of entrepreneurs. To obtain the ‘true’, unconditional relationship between the sex of the entrepreneur and the probability of obtaining formal credit, we need therefore to correct for this selection.

The equation of interest models the probability of obtaining a loan as follows:
$$ { \Pr }\left( {{\text{Loan}}_{ijz} } \right) = \theta \left( {\beta_{0} + \delta F_{ijz} + \gamma X_{ijz} + \varepsilon_{ijz} } \right) $$
(3)
where, Loan equals 1 if firm i operating in industry j and in country z obtained a loan in the last fiscal year and 0 otherwise; F is a dummy variable, indicating that one of the principal owners is a woman; and X is a vector of firm specific characteristics that could affect the likelihood of obtaining formal credit (i.e., the creditworthiness of the firm from the point of view of the bank), such as: how well the firm is run (measured by a performance variable, such as the valued added per worker), whether the firm is facing a demand constraint (capacity utilization), the financial literacy of the entrepreneur (whether or not the owner has a bank account), the stability of the firm (captured by a quadratic in its age) and the size of the firm (measured by the amount of sales 3 years before the current period). This last variable is important to control for because small firms may be less likely to obtain credit.
We model the selection process as occurring over three mutually exclusive outcomes: (a) not needing a loan (and therefore not applying for one); (b) needing a loan but not applying; (c) needing a loan and applying. Notice that group (b) corresponds to the “discouraged borrowers” (Kon and Storey 2003). To correct for this selection, we estimate a multinomial logit selection model (by maximum-likelihood) in the first stage, as outlined in Dubin and McFadden (1984), and extended in Bourguignon et al. (2007). This model is an extension of the standard Heckman (1979) two-stage selection model, which enables it to handle selection over multiple outcomes by way of a multinomial model:
$$ \Pr \left( {\left( {{\text{Need}}/{\text{Apply}}} \right)_{ijz} } \right) = \theta \left( {\widetilde{\alpha } + \widetilde{\vartheta }M_{ijz} + \widetilde{\varphi }Z_{ijz} + \widetilde{\varepsilon }_{ijz} } \right) $$
(4)
where M is the vector of explanatory variables included in Eq. 3 and Z is the vector of instruments that identify the selection equation. The model comprising Eqs. 3 and 4 assumes that \( \varepsilon \sim N \) (0,1), \( \tilde{\varepsilon }\sim N \) (0,1), and corr \( \left( {\varepsilon ,\,\tilde{\varepsilon }} \right) = \rho \). As instruments we use two variables that are likely to be correlated with the need for formal credit, but not with the probability of obtaining it. The first variable is the percentage of sales paid for before delivery; this is likely to be negatively related to the firm’s probability of seeking formal credit. The second is the percentage of working capital financed through retained earnings (a proxy for retained earnings and firm preferences for financing). The coefficient of the gender dummy in Eq. 4 indicates whether female entrepreneurs are different in their propensity to seek formal credit.
The main results are shown in Table 4. We find that, conditional on firm performance and other firm characteristics capturing the firm’s credit worthiness, and after correcting for selection, female-owned firms are as likely as their male-owned counterparts to obtain a loan in all the three regions (Table 4B). This is a strong result suggesting that, among formal enterprises, there is no gender-based discrimination in access to bank credit in a large part of the developing world.
Table 4

Difference in access to credit by gender

Variable

ECA

LA

SSA

A: Whether or not apply for loan by reason (multinomial logit; marginal effects)

 Don’t need and don’t apply

  % of sales paid for before delivery

0.000

(0.000)

0.001***

(0.000)

−0.002***

(0.000)

  % of working capital financed by retained earnings

0.003***

(0.000)

0.006***

(0.000)

−0.001***

(0.000)

  Female-owned firm

−0.030*

(0.017)

−0.053***

(0.016)

0.015**

(0.022)

Need but don’t apply

  % of sales paid for before delivery

0.000

(0.000)

−0.003***

(0.000)

0.003***

(0.000)

  % of working capital financed by retained earnings

0.002***

(0.000)

0.001***

(0.000)

0.005***

(0.000)

  Female-owned firm

0.052***

(0.017)

−0.007

(0.014)

−0.047**

(0.026)

Need and apply

  % of sales paid for before delivery

0.000

(0.000)

0.002***

(0.000)

−0.001***

(0.000)

  % of working capital financed by retained earnings

−0.005***

(0.000)

−0.006***

(0.000)

−0.003***

(0.000)

  Female-owned firm

−0.022

(0.021)

0.060***

(0.018)

0.032

(0.023)

B: Whether or not obtained loan (Logit: Obtained = 1; marginal effects)

 Female-owned firm

0.049

(0.069)

0.014

(0.040)

−0.019

(0.053)

 Value added per worker

0.001

(0.001)

−0.001

(0.001)

−0.001

(0.001)

 Sales 3 years ago

0.025***

(0.009)

0.027

(0.018)

0.037***

(0.009)

 Capacity utilization

0.001

(0.001)

0.001

(0.001)

0.001

(0.001)

 Bank account

−0.035

(0.036)

−0.246

(0.226)

−0.03

(0.054)

 Age of the firm

0.001

(0.002)

−0.001

(0.003)

−0.006**

(0.003)

 Age square

0.001

(0.001)

0.000

(0.001)

0.000**

(0.001)

 Observations

3,353

4,147

2,113

 Sigma square

0.443

(1.335)

0.613**

(0.261)

0.228

(0.725)

 rho1 (Don’t need and don’t apply)

−1.481

(1.693)

−3.120

(2.488)

0.954*

(0.566)

 rho2 (Need but don’t apply)

−0.387

(1.444)

−0.072

(1.970)

−0.136

(0.766)

 rho3 (Need and apply)

−0.923

(0.755)

−1.638

(2.204)

−0.833***

(0.335)

Notes: Bootstrapped standard errors (100 replications) in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%

ECA Eastern Europe and Central Asia, LA Latin America, SSA Sub-Saharan Africa

The multinomial logit for selection correction reveals some interesting findings. In ECA, female-owned firms are significantly more likely than their male-owned counterparts to need a loan but not apply for it. In other words, the hypothesis of demand constraints for formal financing among female entrepreneurs finds support in the ECA region. In contrast, in SSA, female entrepreneurs are significantly less likely than male entrepreneurs to need a loan, but not apply for it. (In LA, the coefficient is negative, but not statistically significant.) For both LA and SSA this finding seems to contradict the common perception that female entrepreneurs are less likely to access formal financing because they are more risk averse, or less financially literate. In fact, in LA, a female entrepreneur is more likely than a male to apply for a loan if they need it.

The most important empirical question that emerges from this analysis is—why are female-owned firms constrained in seeking formal finance in ECA, and not in other regions? This result is particularly surprising in the light of the relatively higher human capital attainment among women in ECA, compared to their counterparts in LA and SSA. One explanation could be the high collateral rates for women in ECA. This explanation would be coherent with the theoretical model developed by Kon and Storey (2003), who show that, in the presence of a collateral, firms (‘good’ firms) that are aware that the bank requires a collateral, but do not have the minimum amount required, are also discouraged from applying for loans. The average value of collateral (as a percentage of the loan) for ECA female entrepreneurs is 166% and is the highest rate of all regions, the corresponding numbers being 152% in LA and 147% in SSA (see Appendix Table 9). Although the average collateral requirements are also high for men (160%) in ECA, a significantly larger proportion of women (with respect to men) claim that they did not apply due to strict collateral requirements (7.6% of all female entrepreneurs who did not apply for credit vs. 5.7% of men). More women than men in ECA also cite high interest rates as the primary reason for not applying, though the average interest rates charged to men and women in ECA are not significantly different. Men in ECA, however, are much more likely than women not to apply for a loan because they did not need it and not because of perceived high costs. These factors seem to suggest, albeit somewhat indirectly, that women in ECA might perceive the cost of applying for loans to be (too) high. Such a gap might explain why female entrepreneurs are significantly more likely than their male counterparts to need a loan but not apply for it.14

However, it should also be noted that in LA the cost of credit is higher for women than for men. First, the cost of collateral for women in LA at 151.5% is much higher than that of their male counterparts, which stands at 118%. Second, although we are missing data on interest rates in LA, about 16% of female entrepreneurs that did not apply for a loan claim high interest rates to be the main reason, compared to only 10% of men. Despite these differences, female entrepreneurs are more likely than their male counterparts to apply for a loan if they need it, as shown in Table 4.

Comparing simple averages does not answer the question of whether women pay more for financing than men do and whether this differential differs across regions; and it does not explain the difference in demand for finance by women in ECA and the other two regions. In order to better address this question, we run regressions on the interest rate and collateral (as a percentage of loan value) holding constant factors that would affect a bank’s economic decision on these two financing decisions, including the characteristics of the firm in Eq. 4 which measure its performance (such as value added, firm’s size in terms of sales, firm’s age and the characteristics of the loan, such as duration and amount, where possible).

The results of the regressions on the interest rate and collateral, presented in Table 5, indicate that, among those who do obtain a loan in ECA and SSA, there is no significant difference in the rate of interest paid by female and male entrepreneurs, after controlling for the firm’s and loan characteristics. However, the cost of the loan in terms of the share of collateral is significantly higher for women than for men in ECA, while no gender differences are found in LA or SSA. Hence, this seems to explain the (rational) response on the part of female entrepreneurs in ECA not to apply for a loan.15
Table 5

Conditions for bank loans by gender

Variable

Rate of interesta

Collateral as a percentage of loan value

ECA

SSA

ECA

LA

SSA

Female-owned firm

−0.154

(0.267)

0.146

(0.331)

0.074*

(0.038)

−0.04

(0.028)

0.022

(0.056)

Value added per worker

−0.004

(0.007)

0.001

(0.005)

0.001

(0.001)

0.001

(0.001)

−0.002***

(0.001)

Sales 3 years ago

−0.386***

(0.084)

−0.357***

(0.106)

−0.01

(0.012)

0.006

(0.008)

0.025

(0.018)

Bank account

0.096

(0.477)

−0.263

(0.760)

−0.166**

(0.072)

0.290***

(0.080)

−0.079

(0.132)

Duration of loan

−0.011***

(0.004)

−0.001

(0.005)

−0.001

(0.001)

0.002***

(0.001)

−0.001

(0.001)

Age

−0.005

(0.023)

0.026

(0.030)

0.008**

(0.003)

0.017***

(0.003)

−0.006

(0.005)

Age square

0.001

(0.001)

0.001

(0.001)

−0.001*

(0.001)

−0.001***

(0.001)

0.001

(0.001)

Ln (loan amount)

na

−0.003

(0.095)

na

−0.035***

(0.010)

0.035**

(0.016)

Loan required collateral

0.770**

(0.347)

1.053**

(0.517)

   

Rate of Interest

  

0.002

(0.003)

na

0.012**

(0.005)

Observations

1489

546

1250

1312

486

R-squared

0.64

0.58

0.19

0.43

0.38

Notes: Robust regressions with country, industry and loan year fixed effects. Standard errors in parentheses. * Significant at 10%; ** significant at 5%; *** significant at 1%.

ECA Eastern Europe and Central Asia, LA Latin America, SSA Sub-Saharan Africa

aRate of interest not available for LA

5.4 Are female entrepreneurs using formal credit as efficiently as male entrepreneurs?

Finally, we ask whether women operate smaller firms because they do not use loans as efficiently as their male counterparts. We examine whether there are any systematic differences in the impact of formal finance on firm size by sex of the entrepreneur, among those who have obtained financing. We regress log sales revenues on different measures of access to bank credit (sequentially), including (1) a binary variable identifying firms that have any part of their working capital and/or new investment financed through banks; (2) a binary variable indicating whether the firm has had a bank loan at any time in the past; (3) a binary variable indicating whether the firm has a loan that was approved/received in the 3 years before the last fiscal year; (4) the average share of working capital financed through borrowing from commercial banks; and (5) the size and duration of the loan. For each of these variables an interaction term between the variable and the dummy for female ownership is included in the regression to capture gender differences in terms of the impact of using formal credit. To avoid the problem of reverse causality, given that firms with larger sales might be more likely to obtain financing from banks, we control for lagged sales (sales 3 years ago). Hence, the coefficients of bank financing indicate the effect on sales today conditional on sales 3 years ago and, as such, can be interpreted as the impact of bank financing on the growth of sales revenue in the 3-year period. As in other regressions, we control for industry and country fixed effects.

The results, shown in Table 6, generally indicate that bank financing in any of these measures is associated with higher sales revenue. However, we find that results for female-owned firms vary by region. For ECA, there are no gender gaps in terms of the impact of formal bank financing on overall firm sales for any of the measures we use. In contrast, in LA and SSA there is some evidence of gender gaps in the impact of bank credit on firm sales. These gaps are evident in the case of firms that have any financing and have had a loan at any time in the past; the gap also increases with the share of working capital financed from a bank. In LA, it appears that receipt of any financing or a bank loan in the last year does not have any positive impact on the size of women-owned firms. Since there is no gender gap in human capital in ECA, while this gap is substantial in SSA, the differences in the impact of formal bank financing on sales across regions may be (partly) due to gender gaps in education and training. However, when we examine the return from a dollar of a loan (by including loan size in the regression and controlling for the duration of the loan), it appears that women in both LA and SSA do at least as well as men. Hence, a plausible explanation is that bank financing produces a lower return for women because women receive smaller loans compared to men. Controlling for loan size, the returns (from a dollar loan) are not significantly different for men and women. Notice, however, that we did not control for selection in this last regression for lack of valid instruments. The results, therefore, refer to the sample of firms that received a loan.
Table 6

Use of bank financing

Dependent variable: Ln (sales)

ECA

LA

SSA

Any financing

0.006

(0.005)

−0.018

(0.016)

0.003

(0.011)

  Female-owned firm*any financing

−0.002

(0.010)

−0.102***

(0.026)

−0.025

(0.022)

  Observations

3,659

5,906

5,947

  R-squared

0.99

0.98

0.97

Have a loan (any year)

0.012**

(0.006)

0.149***

(0.015)

0.103***

(0.018)

  Female-owned firm*have loan (any year)

0.003

(0.010)

−0.206***

(0.023)

−0.064**

(0.029)

  Observations

3659

5904

4866

  R-squared

0.99

0.98

0.97

Loan last 3 years

0.019**

(0.008)

0.038*

(0.023)

0.038

(0.025)

  Female-owned firm*loan last 3 years

0.012

(0.015)

0.004

(0.035)

0.061

(0.042)

  Observations

3,659

5,942

5,947

  R-squared

0.99

0.98

0.97

% of working capital financed from bank

0.001

(0.001)

0.001***

(0.001)

0.002***

(0.001)

  Female-owned firm*% working capital

0.001

(0.001)

−0.002***

(0.001)

−0.002**

(0.001)

  Observations

3,609

5,942

5,690

  R-squared

0.99

0.98

0.97

Loan size

na

0.034***

(0.007)

0.028***

(0.010)

  Female-owned firm*loan size

 

0.010*

(0.006)

0.008

(0.013)

  Loan duration

 

0.000

(0.001)

0.001

(0.001)

  Observations

 

2767

878

  R-squared

 

0.98

0.97

Notes: Regressions also include dummies for female-owned firm, ln(sales) 3 years earlier frim, country and industry fixed effects; these coefficients were suppressed to conserve space. Standard errors are in parentheses. * Significant at 10%; ** 5%; *** 1%

ECA Eastern Europe and Central Asia, LA Latin America, SSA Sub-Saharan Africa, na not available

6 Conclusions

One the one hand, it is often argued that women face gender-specific barriers as entrepreneurs and that these barriers lead to gender gaps in firm performance. On the other hand, it has also been contended that female entrepreneurs have different motivations and preferences than their male counterparts and it is these differences that drive observed gaps in entrepreneurial performance. So far, not much empirical evidence on the subject exists for developing countries, despite the established importance of entrepreneurship in mainstream economic development literature. In this paper, we address this significant research gap by providing a comprehensive analysis of entrepreneurial performance by gender in three regions of the world, namely, Eastern Europe and Central Asia (ECA), Latin America (LA), and Sub-Saharan Africa (SSA), using comparable firm-level data from formal enterprises (the World Bank Enterprise Surveys). By analyzing these three regions separately, we allow for gender differences in firm behavior to vary across these regions. We contribute to the literature by measuring the size of the gaps in various measures of performance (sales revenue, productivity, value added per worker, sales growth, employment growth). We explore whether these gaps differ by industry and the age of the firm. Then we test several explanations for these gaps: (a) gender differences in the age of firms; (b) industrial sector concentration; (c) demand vs. supply constraints to formal bank credit; and (d) gender differences in the returns to formal bank credit.

Our first finding is that, on average, female-owned enterprises are significantly smaller (in terms of overall sales) than those of their male-owned counterparts in each region. This is consistent with the literature on female entrepreneurship in developed countries. The findings on the other performance measures are mixed—gender gaps in firm productivity (measured with value added and total factor productivity) are observed in both ECA and LA, but not Sub-Saharan Africa; and gender gaps in firm growth (measured as growth in sales and in employment) are found only in the LA region.

We then assess whether the differences in performance between male and female-owned enterprises are (partially) driven by differences in the age of their firms, the hypothesis being that women’s enterprises may be younger (either because of women’s recent entry into the playing field or because of women’s lower survival rate), and therefore embed lower levels of know-how and experience. We find that in ECA and LA female-owned firms are younger than male-owned firms on average, although the differences are small. And in fact, controlling for firm age does not alter the initial estimates and does not explain much of the gender differences in performance in ECA and SSA.

We test whether the performance gaps are driven by the fact that women tend to operate in specific sectors, different from those where men operate. We find that in all regions women do indeed concentrate in a smaller number of sectors than men—female entrepreneurs are generally more likely than men to work in sectors like garment manufacturing, retail and wholesale trade, and hotels and restaurants. In ECA and SSA, male-owned firms operating in these female-dominated industrial sectors are also smaller, which partially supports the concentration hypothesis—that is, the hypothesis that women tend to concentrate in sectors with smaller firms, which is why they have poorer outcomes. However, this hypothesis does not seem to hold in LA, where firms operating in female-dominated sectors are actually larger and better performing. A common finding, though, is that gender gaps in total sales and value added are larger within female-dominated sectors.

So, are firms smaller because women are constrained in accessing credit on either the demand or supply side? On the demand side, we find that women in ECA are less likely to apply for credit than men, whereas in Latin America and SSA they are as likely as men, if not more so, to apply for formal credit. On the supply side, we find that (conditional on a firm’s creditworthiness characteristics) banks are as likely to lend to a female-owned firm as to a male-owned firm in all three regions. Hence, we find some evidence of credit constraints on the demand-side (in ECA), but not on the supply-side. When we examine the relative costs of a loan for male and female entrepreneurs, we find that the collateral costs are higher for female-owned firms in ECA, which may explain women’s reluctance to apply for loans.

Finally, women’s returns from a loan, measured by the rate at which they increase their sales revenue, are the same as men’s in ECA, but lower in LA and SSA. We conclude that they are lower in these two latter regions for probably two reasons (a) in SSA, the size of the loan must be smaller for women, since there is no significant gender difference in the rate of return to a dollar, and (b) in LA, the average return is actually higher for a female entrepreneur than for a male.

There is still a lot we don’t know about why the observable differences between female and male entrepreneurs exist in developing countries and how best to harness female entrepreneurship as an engine for economic growth. Specifically, we need to learn more about the preferences and goals of female and male entrepreneurs, how they differ, and what this implies in terms of policy prescriptions. Our analysis has focused on a broad area of the developing world, and while we inevitably found regional differences in gender gaps, more striking are the similarities among women entrepreneurs. In all regions, women run smaller enterprises. In all regions, the concentration of women and men in specific industrial sectors also follows very similar patterns—and this penalizes women entrepreneurs. The fact that we do not find strong evidence of credit constraints in any region forces the analyst to look more closely at other elements, defining an agenda for future research—male and female entrepreneurs can more or less deliberately select different types of sectors or firms or growth patterns because, for example, their risk aversion differs, or they have different preferences for (or opportunities in) alternative types of activities (for example, paid employment), or because their activity as entrepreneurs is subjected to constraints that little have to do with business (household care needs, restrictions in geographical mobility, cultural expectations, etc.) and place a greater burden on women than men. These elements may also be crucial in explaining the selection of individuals into entrepreneurship, a process that is very hard, if not impossible to analyze, because of fundamental data limitations. Moreover, it may be the case that some of the elements that did not turn out to be especially relevant in explaining gender gaps in performance among entrepreneurs, such as access to finance, play a substantial role in the process of entry into entrepreneurship.

As for regional differences—we have documented those, but cannot unfortunately explain them with our data. We suspect that—because of substantial regional differences in business climate, credit markets, labor markets, individual education levels, and gender gaps in a number of dimensions—the men and women who are entrepreneurs are quite different across regions in ways we cannot observe. (We are also aware that this can be true for individual countries, but had to find a compromise between accuracy and feasibility of the analysis.) Unfortunately, enterprise survey data contain very limited individual- and household-level information on the entrepreneur. Moreover, cross-sectional data represents a challenge when it comes to establishing causality. Improving data to investigate the profile, attitudes, and role of the entrepreneur—and not just the characteristics of her or his enterprise—is another powerful implication of our analysis. Richer data will allow for a more complete analysis of all the options as well as the constraints (including at the personal, household, and social level) that impact men and women in different ways and explain the different levels of performance of their businesses.

Footnotes
1

The 27 ECA countries include 16 from Central and Eastern Europe (Albania, Bosnia and Herzegovina, Bulgaria, Croatia, Czech Republic, Estonia, Former Yugoslavia, the former Yugoslav Republic of Macedonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovak Republic, Slovenia, and Turkey) and 11 from the Commonwealth of Independent States (Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan, Kyrgyzstan, Moldova, Russia, Tajikistan, Ukraine and Uzbekistan). The 13 countries in LA are: Argentina, Bolivia, Chile, Colombia, Ecuador, Guatemala, Honduras, Mexico, Nicaragua, Panama, Paraguay, Peru and Uruguay. The 22 SSA countries are: Angola, Botswana, Burundi, Burkina Faso, Cameroon, Cape Verde, Gambia, Ghana, Guinea, Guinea Bissau, DRC, Ethiopia, Kenya, Mali, Namibia, Nigeria, Rwanda, Senegal, Swaziland, Tanzania, Uganda, and Zambia.

 
2

We acknowledge that there are important differences across countries within each region and that ECA, in particular, is a rather artificial construct. However, these regions are those formally recognized by the World Bank Group and data, analytical work, and policy recommendations are routinely collected, produced, and disseminated for each region as a whole, as well as for individual countries (including, for example, in the Doing Business report).

 
3

There is a question as to whether the sample should be weighted to be representative of the relative sizes of these economies or populations. Since the distribution of firms across countries in the pooled analytical sample corresponds quite closely to the countries’ relative size within the region in terms of GDP and population, we do not re-weight our dataset.

 
4

In the case of Sub-Saharan Africa, two observations with extremely high values for fixed assets are also dropped.

 
5

We decided to discard the oldest, ‘historical’ firms because their performance may be disproportionately affected by brand recognition rather than by the ability of the (current) entrepreneur. This selection, moreover, increases the samples’ comparability across regions.

 
6

We are not too concerned about some of the sample selections made, which purely affect the interpretation of the coefficients—for example, our results are only valid for domestic firms that are privately held shareholding companies, partnerships and sole proprietorships, and are younger than 75 years of age. In Sect. 4, we discuss the implications of not being able to determine the owner’s level of involvement in managerial decisions, and we propose a robustness test to explore this issue. In the conclusions we discuss the implications of pooling countries at the regional level. More of concern are the missing data on sex and capital in the relevant sample, whose presence can bias the results in ways that are hard to gauge, a priori. Out of the relevant sample (firms with privately held shareholding companies, sole proprietorship and partnership), those firms with missing data on the gender of the entrepreneur are 1.2% in LA, 1.5% in SSA, and 17.6% in ECA. Firms with missing data on capital are generally a larger portion of the sample, 33.9% in ECA, 37% in LA, and 5% in SSA. Missing data are of concern if the process that generates them is correlated with the dependent variables of interest. We run descriptive and basic regressions (not including capital among the explanatory variables) for the broader sample, and we run the same analysis for the restricted sample with full information on capital. The results remain qualitatively unchanged.

 
7

Notice that the percentage change in a dependent variable that has been log-transformed due to a dummy switching from 0 to 1 is 100*[exp(β) − 1].

 
8

The latter formulation allows for the inclusion of new firms (those for which Xt−3 = 0); moreover, it “compresses” the very large growth rates that tend to be associated with small firms (because they typically start from a very low amount of sales). Notice that g varies between −2 and +2 and it is monotonically related to G. The two growth rate measures are linked by the identity G = 2g/(2 − g) and are approximately equal for small growth rates. When using g instead of G, the results are essentially the same in terms of significance, but the gap is smaller in LA, −0.067 for sales growth and −0.025 for employment growth, both significant at the 1% confidence level.

 
9

Equation 1 can also be interpreted as a first order approximation of more complicated revenue (production) functions. Notice that the production function can be alternatively specified with the dependent variable expressed as total revenue (as in 1) or as value added (i.e. ln(Vi) = ln(Yi − Mi)), with labor and capital (and not intermediate material) included on the right-hand side.

 
10

Whether it is the sex of the owner or of the manager that matters for firm’s performance remains an open question that can only be answered with more detailed data on the ownership and management structure as well as on the sex of owners and managers. Moreover, the sample should include enough variation in ownership (and management) types to be able to disentangle the two effects.

 
11

This argument is reminiscent of the “occupational crowding” literature, arguing that women being crowded into a relatively small number of occupations results in them earning lower wages than men.

 
12

In the case of LA, industrial sectors, such as garments, food, and textiles drive this result.

 
13

Some researchers suggest that women entrepreneurs choose sectors with either of the following characteristics: (1) where it is easier to combine work with household responsibilities; (2) where women can utilize skills they have mastered as part of their socialization process; (3) which require a small initial investment; (4) where women can easily get credit from suppliers, and (5) sectors for which there is a readily tested, and large market.

 
14

Some authors have suggested that entrepreneurs from ethnic minorities and other demographic groups are discouraged from applying for a loan because they fear that their application will be turned down because of poor credit history, prejudice, etc. (Cavalluzzo et al. 2002). Our results for ECA are also coherent with this explanation.

 
15

Robustness tests have been carried out running richer specifications of the regressions presented in Table 5, including also variables for the degree of competition in the industrial sector and capacity utilization for ECA and manager experience for SSA; since the results are virtually the same, the more parsimonious specification is presented in Table 5.

 

Acknowledgments

We acknowledge the support of the PREM Gender group at the World Bank for funding this research. We would like to thank Andrew Morrison, John Jackson, and Jan Svejnar for discussions that significantly improved the paper. Thanks are also due to two anonymous referees and the participants to the Workshop ‘Female Entrepreneurship: Constraints and Opportunities,’ which was organized by the World Bank in Washington D.C. on June 2 and 3, 2009, and to the ‘5th IZA/World Bank Conference: Employment and Development,’ May 3–5, 2010, Cape Town, South Africa, for useful comments and suggestions.

Copyright information

© Springer Science+Business Media, LLC. 2011