WPS8065 Policy Research Working Paper 8065 Crossovers: Female Entrepreneurs Who Enter Male Sectors Evidence from Ethiopia Salman Alibhai Niklas Buehren Sreelakshmi Papineni Rachael Pierotti Africa Region Gender Cross Cutting Solution Area Finance and Markets Global Practice Group May 2017 Policy Research Working Paper 8065 Abstract Occupational sector selection is an important determinant higher profits and having more employees. The descrip- of returns for female entrepreneurs. If sectors that are tradi- tive results show that crossovers do not necessarily have tionally male owned could provide an opportunity to earn more education or greater skills than noncrossovers. Rather, higher returns, then what factors could encourage women women’s relationships and networks, especially those pro- to cross over into these sectors or prevent them from doing vided through male relatives, and being opportunity-driven so? To examine this question, this paper uses data from entrepreneurs appear to influence the likelihood of enter- Ethiopia to compare the firm performance and characteris- ing a more-profitable, male-dominated sector. The study tics of women in male-dominated sectors (crossovers) with explores the implications and challenges of encouraging women who are in female-concentrated sectors (noncross- female entrepreneurs to enter male-dominated sectors, in overs). The findings show that female-owned enterprises an effort to provide new insight into how the earning gap in male-dominated sectors perform better on average than between male and female entrepreneurs can be closed. those in female-concentrated sectors, with firms achieving This paper is a jont product of the Africa Region, the Gender Cross Cutting Solution Area, and the Finance and Markets Global Practice Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at spapineni@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Crossovers: Female Entrepreneurs Who Enter Male Sectors Evidence from Ethiopia By Salman Alibhai, Niklas Buehren, Sreelakshmi Papineni, and Rachael Pierotti∗ Keywords: Gender, Entrepreneurship, Firms, Sector Choice and Ethiopia. JEL codes: J16 L25 L26 O12 ∗ This paper is a product of the World Bank Africa Gender Innovation Lab (AFRGIL). Alibhai (email: aalibhai@worldbank.org); Buehren (email: nbuehren@worldbank.org); Papineni (email: spap- ineni@worldbank.org); Pierotti (email: rpierotti@worldbank.org). This study builds on the research framework of “Breaking the Metal Ceiling: Female entrepreneurs who succeed in male-dominated sectors in Uganda” (2011), and we thank the authors Francisco Campos, Markus Goldstein, Laura McGorman, Ana Maria Munoz Boudet, and Obert Pimhidza for their support during the conceptualization stage and for the survey instrument that was used to collect much of the data used in this paper. This study was integrated into the Women Entrepreneurship Development Project (WEDP), an ongoing lending operation of the Finance & Markets Global Practice, and we thank the Team Leader Francesco Strobbe for his support. We thank Seblewangel Ayalew Woreta for superb field assistance. We are grateful to the World Bank Group’s Umbrella Facility for Gender Equality and the Government of Canada for financial support. 1 2 I. Introduction Sub-Saharan Africa has the highest rate of female entrepreneurship in the world, with more women starting businesses in Africa than anywhere else (World Bank, 2012). Although women now make up approximately half of the entrepreneurs in Sub-Saharan Africa, there is still a large gender gap in business performance (see for example, Klapper and Parker 2011 or Bardasi et al. 2011 for a review of the literature). The latest data from the World Bank Enterprise Surveys (2015) indicates that male-owned businesses in Ethiopia are 2.7 times larger than female- owned enterprises in terms of number of employees and have average monthly sales that are 4 times higher.1 The existing evidence that explains the gender gap in business performance highlights a number of potential factors, including capital and time constraints. Campos and Gassier (2015) emphasize that the systematically different choices that men and women make, which are in turn driven by gender-specific con- straints, can explain gender differences in business performance. One key differ- ence they highlight is the sectors in which men and women choose to operate their enterprise. The sector of operation has been identified as an important determi- nant of returns for female entrepreneurs in many developing countries. De Mel, McKenzie, and Woodruff (2008) find that women are more likely to be present in sectors characterized by low returns, and Singh, Reynolds, and Muhammad (2001) find a high concentration of female-owned firms in low-income informal sectors. These findings raise the question of whether there are certain sectors that women could choose where potential returns are higher. If so, are there any defining characteristics of the women who are more likely to be operating in these sectors that could help direct programming and targeting? In this paper, we use enterprise data collected from Ethiopia to explore what could be driving sector selection in the context of growth-oriented female en- trepreneurs. Specifically, we study female-owned enterprises that operate in sec- tors dominated by men and show that, on average, they are able to make higher returns than female-owned enterprises operating in female-concentrated sectors. Our findings are consistent with the results of Campos and colleagues (2011) in Uganda, who found that women-owned firms in male-dominated sectors, on av- erage, were more profitable than firms in female-concentrated sectors. In this paper, we use data from Ethiopia to corroborate the results obtained in Uganda by testing the hypothesis of higher returns for women in male-dominated sec- tors and then seek to uncover factors correlated with crossing over into these higher-return male-dominated sectors. Comparing the average outcomes of crossovers and noncrossovers suggests that entering and operating in a male-dominated sector leads to higher profits, indi- cating that encouraging female entrepreneurs to enter sectors that are dominated 1 A “female-owned” business was defined as one in which the top manager was a woman. In the 2015 Ethiopian enterprise survey, 9 percent of the 848 businesses indicated that their top manager was a woman. CROSSOVERS 3 by men could be an avenue for programming. Nevertheless, it is important to evaluate the higher-profit potential in the male-dominated sectors along with an appreciation of the heterogeneity of the female entrepreneurs and the business challenges that come with operating in sectors dominated by men. We supple- ment the analysis in Ethiopia by asking whether crossing over is enough. We analyze this by examining whether there are any defining characteristics of the top profit earners who have crossed over into male-dominated sectors. In Ethiopia, it is not surprising that women entrepreneurs tend to be less successful than men because women typically start from a more-difficult situ- ation—without easy access to finance, land, training, education, or effective busi- ness networks (Triodos Facet 2011). Female entrepreneurs face higher barriers than men when accessing finance, especially in providing collateral for loans, be- cause most assets that lenders accept are registered to men, and there are cultural barriers to women using collateral. The share of women in Ethiopia without ed- ucation is almost twice that of men, which in turn limits female entrepreneurs’ ability to expand their businesses (Ethiopia DHS 2011). A better understand- ing of the characteristics of women who have successfully circumvented some of the challenges and entered higher-return male-dominated sectors will help us to suggest policy interventions that could address the constraints women face in an effort to encourage the productivity of women-owned businesses and closing the gender gap. The World Bank report “Unleashing the Potential of Ethiopian Women” postulates that reducing gender inequalities in education and the la- bor market might increase annual gross domestic product growth in Ethiopia by approximately 1.9 percentage points.2 The remainder of this paper is organized as follows. Section II describes our main data source. Section III describes how we defined which sectors can be considered male dominated. Section IV presents the empirical results where we compare the business performance and challenges of crossover firms with those of noncrossover firms and then explore a number of correlates associated with crossing over to male-dominated sectors. Section V provides the conclusion. II. Data In this paper, we use data collected at baseline for the Women’s Entrepreneur- ship Development Project’s (WEDP) impact evaluation. WEDP3 is the World Bank’s International Development Association–funded program that offers loans and entrepreneurship training to growth-oriented female entrepreneurs4 in Ethiopia. 2 Unleashing the Potential of Ethiopian Women, World Bank 2008. 3 Launched in 2012, WEDP aims to increase the earnings and employment of micro and small en- terprises fully or partly owned by female entrepreneurs in six selected cities. The WEDP objective is achieved by tailoring financial instruments to the needs of the participants and ensuring availability of finance and developing the entrepreneurial and technical skills of the target group. Women interested in participating in WEDP and fulfilling the criteria for project beneficiaries (not full time in school and being growth oriented) receive a WEDP membership card that entitles them to access WEDP services. 4 The WEDP targets a specific group—(growth-oriented female entrepreneurs—defined as female en- trepreneurs with the ambition and potential to expand their micro-enterprises, innovate, and generate 4 The baseline data collection was conducted in October and November 2014 in six Ethiopian cities: Addis Ababa, Adama, Bahir Dar, Dire Dawa, Hawassa, and Mekelle. Baseline survey data were collected from a sample of 2,369 female entrepreneurs who had registered for WEDP at the time of the survey, and an additional survey module was administered to a subset of 800 female entrepreneurs that included more questions about sector choice. The sampling for the additional survey mod- ule was stratified according to sector of business reported in the WEDP registra- tion database, and we oversampled entrepreneurs who we thought might operate in a male-dominated sector to ensure that our sample had a representation of women-owned businesses in male-dominated sectors. We obtained full baseline and additional sector data on 790 female entrepreneurs who form the sample for this paper. The WEDP baseline questionnaire contained a set of questions on household demographic characteristics, socioeconomic status, business sales, profits, costs, employees, entrepreneurial profile (e.g., age, place of birth, education level), and questions designed to elicit an entrepreneur’s business knowledge and level of financial literacy. The additional survey module asked detailed questions about family background, entrepreneurial capability (digit span forward recall, Raven’s Test, entrepreneurial psychology), role models, and access to social networks and retrospective questions about motivations for becoming an entrepreneur. The survey module additionally included some open-ended questions about how the women became exposed to their sector of operation, including, “Please elaborate on the story that led you to have a business in this sector.” In addition to the quantitative surveys, 90 qualitative interviews were conducted with WEDP-registered women. The qualitative research was designed to capture narratives from WEDP entrepreneurs about their attempts to obtain loans from formal financial institutions. For contextual information, all qualitative interviews started with questions about the history of the woman’s business. The qualitative sample included 17 women who were operating businesses in male-dominated sectors, and we use their business histories in this paper to provide illustrations of the ways in which husbands may facilitate women entering male-dominated sectors. III. Which Sectors Are Male Dominated? To proceed with the crossovers analysis, we first need to categorize which sectors are male dominated. In the literature, male-dominated sectors are often defined as those in which men own more than 50 percent of the firms. In the Uganda study, Campos and colleagues (2011) defined male-dominated sectors as those in which men owned more than 75 percent of enterprises.5 In this study, we took paid employment. 5 Although male-dominated sectors are often defined as those in which more than 50 percent of the firms are owned by men (overrepresentation), in this study, we focused on enterprises in which male CROSSOVERS 5 motivation from the Uganda study, but we drew on a question administered in the WEDP baseline survey that asked the entrepreneur, “Are most enterprises in your business sector owned by men or women?” If more than 75 percent of responses from the entire baseline sample (2,369 female entrepreneurs) were that men owned most enterprises in their business sector, we defined that sector as male dominated. We used the entire core baseline sample of 2,369 entrepreneurs to determine this definition of a male-dominated sector and defined a woman who had a business in any of these sectors as a “crossover.” Using this definition of a male-dominated sector, we established 164 crossover businesses, with the remaining 626 businesses in the sample considered noncrossover. Figure 1. Sectors classified as male dominated As a robustness check, we examined the sector label names that have been classified as male dominated based on this definition. We believe that the sectors ownership was more than 75 percent, because the overrepresentation could have been a reflection of the specific sample, which is often not representative of the entire population of firms in the country. 6 match the ones that we would classify as male dominated if we were to select based on their labels alone, with the possible exceptions of software development and business services. We are confident that the sectors identified as male dominated in this paper are ones that could be generalized as so for Ethiopia in the absence of a full census of firms. We use this definition of crossovers throughout this paper to explore differences in outcomes between crossover and noncrossover firms. IV. Results A. Firm Performance We compare mean differences between business outcomes for crossover and noncrossover female businesses to test whether female enterprise owners in male- dominated sectors have higher returns than female owners in female-concentrated sectors in Ethiopia. We begin by investigating the differences in profits, revenues, costs, and num- ber of workers between firms owned by crossovers and noncrossovers.6 Many of the women who cross over into male-dominated sectors make higher profits than women in female-concentrated sectors, as was found in the Ugandan sample of firms. In Table 1, we present the means of a number of business performance outcomes, as well as whether the female owner had business partners and the age of the business for the full sample of enterprises (column 1), separately for crossovers (column 2) and noncrossovers (column 3), and the differences in means between crossovers and noncrossovers (column 4). The asterisks in column 4 signify a significant difference in the particular outcome variable between crossover and noncrossover enterprises. The results in Table 1 show that crossovers perform better than noncrossovers. Crossover firms have, on average, double the profits of noncrossover firms. This profit difference is statistically significant and was found to be robust with alter- native definitions of crossovers. We found no difference in revenues for crossovers and non-crossovers when revenues are defined over the last month and for a typical month. Crossover businesses also have statistically significantly higher operating costs than non-crossover businesses. One possible explanation for the revenue, cost, and profit results is that a large proportion of the noncrossover businesses are in the retail sector (50 percent), which is typically a sector associated with higher revenues but lower profit margins. On average, the crossover firms are 6 We look for differences in average outcomes for crossovers and non-crossovers and estimate for business i: (1) yi = βCROSSOV ERi + δXi + i where yi is the outcome of interest and CROSSOV ER is an indicator variable that equals 1 if the firm is a crossover (operates in a male-dominated sector). All regressions cluster standard errors within subcity groups (there are 20 subcities in the sample). We report regression estimates controlling for the age of the business and city dummy variables (Xi ). CROSSOVERS 7 Table 1—Business Performance and Characteristics significantly larger than noncrossover firms as measured according to the number of employees, with crossover firms employing an average of 4.3 workers, whereas noncrossover firms employ an average of 2. Crossover firms are also significantly more likely to report operating with busi- ness partners than noncrossover firms. A business partner might provide the women with support to enter and operate in a male-dominated sector, which we will explore further later in this paper. The average age of crossover and non- crossover firms are not statistically significantly different. When we included a control for age of business in the average monthly profits regression, we still found that profits of crossover firms were significantly higher than those of noncrossover firms (at the 1 percent level of significance), so the explanation for the profit cannot simply be that older crossover firms had longer to expand their businesses to reach a higher level of operating profit. Overall, the business performance results in Table 1 show that female-owned en- terprises in male-dominated sectors perform better than those in female-concentrated sectors. It is important to evaluate the higher growth potential in the male- dominated sectors alongside the possible business challenges that women could face in sectors dominated by men. 8 B. Challenges We compare a set of discrimination outcomes and operational challenges that women operating in male-dominated sectors face with those in female-concentrated sectors in Table 2. Table 2—Business Challenges Table 2 shows the main challenges that crossovers and noncrossovers face and that crossovers report more operational challenges and seem to have more dif- ficulty building networks within their sector than noncrossovers. A crossover is significantly more likely to face the challenge that clients prefer to do business with male business owners and is more likely to face problems with male employees. Harassment outcomes are similar for crossover and noncrossover entrepreneurs, with as many as 11 percent of the women reporting being sexually harassed within the past 12 months and 22 percent experiencing some form of other abuse in the past 12 months, suggesting that female entrepreneurs in Ethiopia face harass- ment problems when attempting to operate businesses in general, although this is not necessarily a feature of the sectors within which crossovers operate. In terms of networks, crossovers are significantly more likely to face difficulty in CROSSOVERS 9 building networks in their sector of operation and do not seem to differentially benefit from networks of women in the same industry, with crossovers reporting that they are more likely to feel despised by other women business owners. These challenges suggest that any policy recommendation encouraging women to enter male-dominated sectors should be complemented with regulatory policies to help alleviate the gender-specific challenges that women may face operating in these sectors. C. Exploring the factors correlated with crossing over We looked for potential differences between crossovers and noncrossovers in household demographic characteristics, access to finance, education and skills, preference for business, and role models to better understand the factors associ- ated with women crossing over to male-dominated sectors. We interpreted any differences simply as correlations between the outcome of interest and a woman’s likelihood of being a crossover but cannot always be sure of the direction of causality. 1. Demographic Characteristics The demographic characteristics of women who cross over into male-dominated sectors may differ from those of women who operate in female-concentrated sec- tors. In Table 3, column 4, we present the mean differences in household demographic characteristics between crossovers and noncrossovers. Crossovers were more likely to be married than noncrossovers. Religious background did not seem to play a role in the probability of crossing over in the Ethiopian context, although the majority of households in the sample were Orthodox Christians. The spouse of a crossover was more likely to work in a business himself than noncrossover spouses. These outcomes suggest that a woman’s decision or ability to cross over could be related to the influence of the productive activities of her husband. For example, having a husband working in a business himself could facilitate the woman’s entry into a male-dominated sector. We are cautious in how we interpret these differences because we cannot distinguish whether being married increases a woman’s likelihood of being a crossover or whether the causality is reversed, although it appears that crossovers are more likely to live in homes where there is some spousal support than noncrossovers. We investigate further the role of the husband in helping women cross over and provide some illustrations from the qualitative data to assess how they could be influential in a later section of this paper. 2. Financial Access and Capital Requirements Fafchamps and colleagues (2014) reported that women who reported lower prof- its said that they chose the sector because of low capital requirements. We first 10 Table 3—Household Demographics analyzed whether the sectors defined as male dominated actually require more starting capital and then assessed whether the wealth status and financial ac- cess of women who cross over into male-dominated sectors differ from those who remain in female-concentrated sectors. To estimate the capital requirements of male-dominated sectors, we comple- mented the survey data collected with administrative data collected at the time of WEDP program registration. We used data on capital at business start and sector of operation from the registration dataset. In the full WEDP registration data, as of July 2016, we found approximately 10 percent (1,150 firms) who are crossovers out of 12,549 firms for which we have start-up capital information. We regressed crossover status on capital required to start the business and found that firms in male-dominated sectors, on average, required approximately three times the amount of capital to start a business than firms in the female-concentrated sec- tors. Because the majority of crossover firms in the WEDP registration database are in the small transport services and construction sectors, the higher starting capital requirement is unsurprising. Because capital requirements are higher in crossover firms, we assessed whether there is a correlation between having better wealth or access to finance and cross- ing over into male-dominated sectors. CROSSOVERS 11 Table 4—Household Wealth and Financial Access Table 4 shows differences in household wealth and access to finance between crossovers and noncrossovers. The survey did not directly collect information on household income levels, but we created an index of household assets to use as a proxy for wealth. We found that crossovers were significantly more likely to live in households with greater wealth, obtaining a significantly higher mean score on the household asset index. Crossovers also reported having access to more money in an emergency and were more likely to report contributing money regularly to household expenditures. We do not claim a direction of causality because we could not determine whether the financial flexibility increased the woman’s likelihood of crossing over or being a crossover in the first place allowed the woman greater financial freedom through higher profits gained from owning a crossover business. We found no difference between crossovers and noncrossovers in the likelihood of saving money from their business in the bank or having borrowed money from any source for their business in the past 12 months. 3. Education, Skills and Experience We compared a number of educational, entrepreneurial ability, and psychologi- cal index outcomes of women who crossed over into male-dominated sectors with those who remained in female-concentrated sectors. Table 5 indicates that there are no differences in the level of education of crossovers and noncrossovers. Crossovers performed slightly better than non- crossovers on the forward digit span recall test (a proxy for short-term memory). We found no differences for the Raven’s Test score (used to measure abstract logical thinking). Whether the women previously received training did not differ between crossovers and noncrossovers, although crossovers scored slightly better 12 than noncrossovers on questions measuring business knowledge. Table 5—Education, Skills and Experience We also tested whether crossover status was correlated with the entrepreneurial propensities of crossovers by assessing differences in a number of noncognitive skills (results not presented). Khwaja and Klinger (ongoing) designed a psycho- metric test that analyzes ethics and character, intelligence, attitudes and beliefs, and business skills to gauge innate entrepreneurial spirit, and de Mel and col- leagues (2008) used measures of cognitive ability and personality tests to predict entrepreneurial ability. We compared some of these characteristics in crossovers and noncrossovers and found that measures of entrepreneurial propensity did not appear to be related to the probability of crossing over. Measures of self-efficacy, striving for achievement, impatience, impulsiveness, passion for work, tenacity, locus of control, work centrality, and organization are all similar in crossovers and noncrossovers. Overall, the finding that crossovers and noncrossovers do not differ significantly on various measures of cognitive and noncognitive skills suggests that there is not a requirement to have greater skills and entrepreneurial ability to enter and operate in male-dominated sectors. CROSSOVERS 13 4. Opportunity versus Necessity Entrepreneurship The entrepreneurship literature shows that entrepreneurs who choose to start a business because they are able to identify a good business opportunity and act on it (opportunity entrepreneurs) are different from those who are forced to become business owners because of lack of other employment alternatives (neces- sity entrepreneurs) (Calderon et al. 2015). We asked business owners to explain the primary reason for starting their business to establish whether their business was opened as a preferred choice or out of necessity as a means of survival. We asked business owners to explain their primary reason for starting their business and coded the responses and grouped them to distinguish whether the reason was a preferred business choice (preference) or simply a need for money for the household (necessity). Table 6—Preference for Business Table 6 shows that crossovers are significantly more likely to report starting their business as a preference and less likely to do so out of necessity. The main reason given for preference for starting the business was “saw a market opportu- nity when starting this business,” suggesting that the decision to enter a male- dominated sector may be linked to the availability of information and market entry opportunities. An alternative measure of preference for business (Revealed preference for business in Table 6) was whether the business owner would accept a monthly salary of anywhere between 0 and 15,000 Ethiopian birr to move into wage employment. If they reported that they would take a wage job in this salary range, they were defined as preferring wage work to being in business; 38 percent of the sample indicated that they would prefer to move into wage work based on this definition. Women who demonstrated a preference for business rather than 14 wage work were more likely to be crossovers. The direction of causality is unclear for the revealed preference variable because it may be argued that the wage a per- son would accept to leave their business is correlated with the profits they make in that business. Crossovers were also significantly less likely than noncrossovers to have started the business themselves, which adds weight to the suggestion that crossovers had easier market entry opportunity. 5. Role Models and Early Influences In the Uganda study, Campos and colleagues (2011) highlight the “importance of male role models and ‘gate openers,’ and the active engagement of the role models in engaging a female entrepreneur in the activities of the crossover sector and in breaking traditional norms.” We assessed whether having a role model influenced the sector of operation in Ethiopia. Table 7—Early Influences Table 7 shows the early influences in the women’s lives to assess whether there is any predictability in crossing over based on childhood influences; no conclusive evidence of this was found in Ethiopia. Women were asked explicitly in the additional survey module whether they had a male or female role model while growing up. To further analyze potential early influences, we coded the outcomes to highlight the occupational choice of the parents and grandparents. The father having a government wage job was positively correlated with being a crossover, CROSSOVERS 15 and the mother or father having worked on a farm was negatively correlated with being a crossover. For the other occupational choices of parents and grandparents (not reported), we found no differences between crossovers and noncrossovers. 6. Husband’s Influence Despite the finding that a woman’s early influences in life do not seem to pre- dict the sector of operation that she enters, one repeated observation was the mention of the spouse as being an influential person in the decision to cross over. Crossovers were more likely to be married, have a business partner, and report that their husband was a business owner himself than noncrossovers. In Ethiopia, crossovers are not significantly more likely to report that they had a male role model but were more likely to mention that their husband had the idea to start the business. A pattern that emerged from the Ethiopian data and that the qual- itative interviews corroborated is that the husband was often helpful in providing a market entry opportunity into the crossover sector. To further investigate the influence of the husband, we first restricted the analy- sis to the sample of 511 married women and asked whether the husband was aware of the existence of the business and, if so, specifically about the ways that the husband helped with the business. Fifteen percent (77 women) of married women said that their husbands did not help with the business in any way, and 6 percent of the married sample explicitly stated that they inherited the business directly from their husband. We define a dummy variable husband helped equal to 1 if the married owner said her husband had the idea to start the business, contributed money, gave advice, or helped register the business with the authorities. Table 8, in the sample of married women, shows the influence of the husband according to crossover status. When women were asked about who had the idea to open the business, 42 percent of married crossovers said that their husband did. This is statistically significantly higher than for noncrossovers, 18 percent of whom reported that her husband had the idea to start the business. Correspond- ingly, crossovers were less likely to report that opening the business was their own idea than noncrossovers. Although 83 percent of married women overall report that their husbands helped in the business in some way, crossovers were signifi- cantly more likely than noncrossovers to have received help. The main channel of influence reported in the data was that the husband contributed money. These findings suggest that the husband could have a significant role in in- fluencing the wife’s choice of business sector and facilitating her entry into a male-dominated sector. Table 1 showed that crossover firms were more likely to explicitly report that the type of business was a “business partnership.” With respect to joint ownership, 43% of crossover firms named at least one more owner in the survey, compared with 15% of noncrossover firms. We find evidence that, crossover or not, female entrepreneurs in the sample were actively involved in their businesses. Eighty-four percent of women in the sample reported that they were involved in the day-to-day production or service 16 Table 8—Husband’s Influence delivery of the business and identified themselves as the owner and manager of the business, with primary decision-making power. Crossovers were more likely than noncrossovers to identify themselves as just administrators in the business, with 26 percent of crossovers reporting that they only administered, compared with 5 percent of noncrossovers, with crossover firms using employees and husbands for production and service delivery. The business histories from the qualitative research provide some illustrations of the ways in which husbands may facilitate women entering and operating in male-dominated sectors. In the crossover businesses, husbands provided various combinations of the idea for the business, start-up capital or assets, trade-specific skills, market connections, and co-management. The story from one of the crossover women in Bahir Dar, whose business sup- plies construction materials, provides a good illustration of a husband playing an important supporting role. The female entrepreneur had learned about the con- struction industry while working with her brother in the family business before she was married. Her husband owned a general mechanics shop and provided his wife with start-up capital for her business in the form of a loan. Because he operated a business in a related industry, he was also able to help her with connections to input suppliers. As the entrepreneur described her experience, the capital and encouragement from her husband were instrumental to her operating this business, but she maintained ownership. Other entrepreneurs operating crossover businesses described their husbands as playing a more-integral role in the start-up or operation of the business. A rela- tively successful WEDP entrepreneur in Hawassa talked about managing multiple businesses, some jointly with her husband and some on her own. Her business provided heavy truck rentals, and her husband jointly financed it. She explained that she used profits from one of her retail shops combined with her husband’s CROSSOVERS 17 profits from one of his businesses to build a house. After the house construction was complete, they decided jointly to use the house blueprint as collateral for a loan to purchase the heavy trucks. In other crossover firms, the husband contributed his technical skill, motivating his wife’s trade choice. An entrepreneur in Addis Ababa opened a leather man- ufacturing business after receiving machinery from a business that her husband had operated and then closed. Her husband was a trained leatherworker and pro- vided skilled labor to the business. In another example, a female entrepreneur in Bahir Dar was running an injera supply business when she married her husband, who was a gold and silversmith. They decided to jointly open a jewelry produc- tion shop for which he provides the skilled labor and she provides management. Finally, some of the crossover entrepreneurs whose husbands are the most central to business operations are those who are providing small transport services or renting vehicles that their husbands drive for or maintain on a regular basis. Overall, the extent of the husband’s involvement in the WEDP entrepreneurs’ businesses varied from household to household. Some types of support, such as start-up capital, would be important to crossover and noncrossover businesses alike. In other cases, as some of these examples illustrate, it is likely that the husband’s involvement was integral to the decision to cross over into the male- dominated trade. D. Summary of Factors On average, being a crossover business owner earns a woman higher profits and allows her to have more employees. So, what does it take to cross over into a male- dominated sector? Table 9 summarizes our findings of what predicts crossing over into a male-dominated sector. Explanatory factors are sequentially added to the regression, grouped into the following themes: cognitive skills (column 1), com- mitment to business (column 2), connections (column 3), and noncognitive skills (column 4).7 In each of the following regressions, we include only the explanatory variables for which we believe there is an unlikely possibility of reverse causality. Column 5 in Table 9 includes all four themes in one regression to predict the likelihood of crossing over into a male-dominated sector. The evidence from Ethiopia suggests that entering a crossover sector has a lot to do with circumstance, with starting a business venture because of a market opportunity and having a husband who is in business himself strongly predicting 7 We estimate the predictors of crossing over with a regression of the form for business owner i: (2) CROSSOV ERi = αCogSkillsi + βCommitmenti + δConnectionsi + γN onCogSkillsi + i where CROSSOV ERi is the outcome of interest, an indicator variable that equals 1 if the firm is a crossover. All regressions cluster standard errors within subcity groups (there are 20 subcities in the sample). The explanatory variables are grouped into themes: cognitive skills (education, digit span, Raven’s), commitment to business (preference for business, started because of a market oppor- tunity, business knowledge), connections (married, spouse owns a business himself, male role model), and noncognitive skills (self-efficacy, impulsiveness, organization). 18 Table 9—Summary of factors correlated with crossing over crossing over into a male-dominated sector. E. Is crossing over enough? Going beyond means The results show that, on average, female-owned firms in male-dominated sec- tors have higher profits than female-owned firms in female-concentrated sectors, although when we assess the entire distribution of profits, we find that firms to- ward the top end of the profit distribution largely drive the outperformance of crossovers at the mean level. Figure 2 shows that the difference in profits is significantly positive at the higher quantiles of profitability, where we find a positive significant difference between crossovers and noncrossovers between the median and 60th percentile of the profit distribution. In the figure, we present quantiles 10, 25, 40, 50, 60, 75, and 90. This result suggests that it might not just be about sector, but also about potential within sectors, so we provide supplemental analysis, assessing whether there are any defining characteristics of the top profit earners in our sample of crossover firms. We define top profit earners as firms in the top 25 percent of the profit distribution, which is a somewhat arbitrary cut-off. The 25 percent level was chosen because it is expected to give us a more-refined picture of what it means to perform well in business and is representative enough that a few large outlier firms do not drive the results. Table 10 shows that the factors that predict crossing over into a male-dominated CROSSOVERS 19 Figure 2. Ordinary Least Squares and Quantile Regression Estimates for Average Monthly Profit Model Note: Figure 2 presents a summary of the quantile regression results for the effect of being a crossover on average monthly profits. We plotted seven distinct quantile regression estimates ranging from 0.1 to 0.9 as the solid line. These point estimates may be interpreted as the correlation between crossover status and average monthly profit. The dashed line in the figure shows the ordinary least squares estimate of the mean effect. The shaded grey area depicts a 95 percent pointwise confidence band for the quantile regression estimates. sector also predict success in a male-dominated sector. Table 10 reveals that the top 25 percent of profit performers among crossover firms showed more commit- ment to being an entrepreneur than the remaining 75 percent. (The top crossovers were more likely to start the business because they had a preference for business and had a slight advantage in business knowledge.) In addition, the top crossovers were more likely to report a husband who was in business. In terms of skills, ed- ucation did not make a difference in performing well among crossover firms, but the top crossovers scored higher on digit span recall and self-efficacy and reported being less impulsive. The top crossovers were more likely to state explicitly that they had a male role model while growing up, which mirrors the finding in Uganda (Campos et al. 2011). 20 Table 10—Top Profit Earners among Crossover firms V. Conclusion Providing support to female entrepreneurs to transition into more-profitable, male-dominated sectors is an important step toward closing the gender gap. In this paper, we give evidence that women are able to earn higher profits in male- dominated sectors than women who remain in female-concentrated sectors. We also observe that women in male-dominated sectors are able to create larger firms in terms of number of employees and capital levels. When we explored the factors associated with women crossing over and succeed- ing in these male-dominated sectors, we found that the women did not necessarily possess greater entrepreneurial ability or education, but they had better support networks and were more likely to have the assistance of a husband or a male role model. These support networks helped them enter untraditional business sectors and circumvent some of the discrimination faced when interacting with CROSSOVERS 21 male employees and clients in these sectors. Crossovers also demonstrated greater entrepreneurial commitment and were more likely to be opportunity-driven en- trepreneurs who entered businesses willingly rather than out of a lack of income- generating alternatives. Programs that encourage entry into male-dominated sectors may be particu- larly pertinent when thinking about policy efforts to encourage female-owned en- terprises to grow from microenterprises into small and medium enterprises. The higher capital requirements of the crossover businesses means that dedicated lend- ing initiatives for female entrepreneurs are critical to easing financial constraints and helping these businesses grow. Female-owned businesses in male-dominated sectors also offer an opportunity to boost the wage labor supply for women because there is evidence that the number of female employees is typically higher in firms with a female than a male top manager (World Bank Enterprise Note, 2014). The salary for employees in crossover sectors is approximately double the salary of workers in noncrossover sectors.8 Programs that train women on the skills needed to operate in male- dominated sectors could help them compete for these higher-paying, salaried jobs. Policy efforts to encourage women to enter nontraditional sectors should estab- lish which women are committed to operating a business (measured as a preference for business rather than a necessity for money or inability to find wage work) as a first step in targeting the appropriate women for these programs. McKenzie (2015) provides evidence on the use of business plan competitions as a tool for identifying high-growth entrepreneurs. Understanding which entrepreneurs are in pursuit of an opportunity could also help policy makers decipher which en- terprises should be targeted for assistance in growing and which women should instead be helped into the wage labor pool. With the finding that women with supportive husbands and male role models are more likely to perform well in a male-dominated sector, perhaps programs should encourage men to introduce their wives to their own business networks, pass on key technical skills, and help them obtain start-up capital. A better understanding of how husbands support their wives in business may help inform policy to replicate the support or advice structures that they provide. Facilitating access to networks and providing training on how to overcome discrimination and improve negotiation skills could give women a collective voice and ameliorate some of the challenges women face when operating in a male- dominated sector. Moving forward, we encourage experimentation to determine the most-effective approaches. 8 Average salary per employee was calculated by the total cost paid for salaries for employees in the last 30days divided by the total number of workers reported. 22 REFERENCES [1] Amin, Mohammad. 2014. The Critical Importance of Data Collection Efforts in Developing Countries: The Case of Gender. World Bank Group Enterprise Note No. 31. Washington, DC: World Bank. [2] Bardasi, Elena, Shwetlena Sabarwal, and Katherine Terrell. 2011. How Do Female Entrepreneurs Perform? Evidence from three Developing Regions. Small Business Economics, 37, (4), 417-441 [3] Calderon, Gabriela, Leonardo Iacovone, and Laura Juarez. 2015. “Oppor- tunity vs Necessity: Understanding the Heterogeneity of Female Micro- Entrepreneurs”, Mimeo. World Bank. [4] Campos, and Gassier. 2015. Gender and Entrepreneurship in Sub-Saharan Africa. World Bank White Paper. Washington, DC: World Bank. [5] Campos, Francisco, Markus Goldstein, Laura McGorman, Ana Maria Munoz Boudet, and Obert Pimhidzai. 2011. Breaking the Metal Ceiling: Female Entrepreneurs Who Succeed in Male-Dominated Sectors in Uganda. World Bank Policy Brief No. 93894. Washington, DC: World Bank. [6] De Mel, Suresh, David McKenzie, and Christopher Woodruff. 2008. “Re- turns to Capital in Microenterprises: Evidence from a Field Experiment.” Quarterly Journal of Economics 123 (4): 1329-1372. [7] De Mel, Suresh, David McKenzie, and Christopher Woodruff. 2009. “Are Women More Credit Constrained? Experimental Evidence on Gender and Microenterprise Returns”. American Economic Journal: Applied Economics, 1-32. [8] Fafchamps, Marcel, David McKenzie, Simon Quinn, and Christopher Woodruff. 2014. “Microenterprise Growth and the Flypaper Effect: Evi- dence from a Randomized Experiment in Ghana.” Journal of Development Economics, 106, 211–226. [9] ILO (International Labour Organization). 2012. Global Employment Trends for Women 2012. Geneva, Switzerland: ILO. [10] Khwaja, Asim Ijaz, and Bailey Klinger. Forthcoming. Psychometric Analysis of Entrepreneurs. Cambridge, MA: Entrepreneurial Finance Lab, Harvard University. [11] Klapper, Leora F., and Simon C. Parker. 2011. Gender and the Business Environment for New Firm Creation. World Bank Research Observer, World Bank Group, vol. 26(2), pages 237-257 CROSSOVERS 23 [12] McKenzie, David. 2015. “Identifying and Spurring High-Growth En- trepreneurship. Experimental Evidence from a Business Plan Competition”. Policy Research working paper; no. WPS 7391; Impact Evaluation series. Washington, DC: World Bank Group. [13] McKenzie, David, and Christopher Woodruff. 2008. “Experimental Evidence on Returns to Capital and Access to Finance in Mexico.” The World Bank Economic Review 22 (3): 457-482. [14] Mead, and Liedholm. 1998. “The dynamics of micro and small enterprises in developing countries.” World Development 26(1):61–74. [15] Singh, Reynolds, and Muhammad. 2001. A Gender-Based Performance Anal- ysis of Micro and Small Enterprises in Java, Indonesia. Journal of Small Business Management 39(2): 174 - 182 [16] World Bank. 2009. “Unleashing the Potential of Ethiopian Women: Trends and Options for Economic Empowerment.” Washington, DC: World Bank. [17] World Bank. 2012. “World Development Report 2012: Gender Equality and Development.” Washington, DC: World Bank. [18] World Bank. 2015. Enterprise Survey Ethiopia. Washington, DC: World Bank.