Gender differences in business performance: evidence from the Characteristics of Business Owners survey

The 1992 CBO survey was conducted by the U.S. Bureau of the Census to provide economic, demographic, and sociological data on business owners and their business activities (see Bates 1990; U.S. Census Bureau 1997; Headd 1999; Robb 2000 for more details on the CBO). It includes oversamples of black-, Hispanic-, other minority- (which is primarily Asian), and female-owned businesses. The survey was sent to more than 75,000 firms and 115,000 owners who filed an Internal Revenue Service (IRS) form 1040 Schedule C (individual proprietorship or self-employed person), 1065 (partnership), or 1120S (subchapter S corporation). Only firms with US$500 or more in sales were included. The universe from which the CBO sample was drawn represents nearly 90% of all businesses in the USA (U.S. Census Bureau 1996). Response rates for the firm and owners surveys were approximately 60%. All estimates reported in this article use sample weights that adjust for survey non-response (Headd 1999).²

The CBO is unique in that it contains detailed information on both the characteristics of business owners and the characteristics of their businesses. For example, owner characteristics include education, detailed work experience, hours worked in the business, marital status, age, weeks and hours worked, personal income, and how the business was acquired. Business characteristics include closure, profits, sales, employment, industry, startup capital, types of customers, health plans, and exports. Most business characteristics refer to 1992, with the main exception being closure, which is measured over the period 1992–1996. Additional advantages of the CBO over other nationally representative data sets for this analysis are the availability of measures of financing at startup and the large oversample of female-owned businesses. Finally, the CBO allows one to explore the causes of gender differences in several business outcomes, such as closure rates, sales, profits, and employment size, instead of focusing solely on self-employment earnings.

The sample used for our analysis includes firms that meet a minimum weeks and hours restriction. Specifically, at least one owner must report working for the business at least 12 weeks in 1992 and at least 10 h/week. These restrictions exclude 22.1% of firms in the original sample. The weeks and hours restrictions are imposed to rule out very small-scale business activities, such as casual or side-businesses owned by wage/salary workers; they also allow us to check the sensitivity of our main results.

Table 1 reports estimates of closure rates between 1992 and 1996, and 1992 profits, employment size, and sales for female- and male-owned firms from the CBO. The magnitude of the differences in business outcomes is striking. For example, only 17.3% of female-owned firms have annual profits of US$10,000 or more, compared with 36.4% of male-owned firms. Female-owned firms also have lower survival rates than male-owned firms. The average probability of a business closure between 1992 and 1996 is 24.4% for female-owned firms, compared with 21.6% for male-owned firms.

Female-owned firms are substantially smaller on average than are male-owned firms. The mean of log sales among female-owned firms was 9.57 in 1992, compared with 10.36 for firms owned by men. Female-owned firms are also less likely to have employees than firms owned by men. Seventeen percent of female-owned firms hire employees, compared with 23.7% of male-owned firms. Average employment is also much smaller among female-owned firms than among male-owned firms.

Firms owned by men are also much more likely to have larger sales than firms owned by women. Figure 1 presents CBO data showing the distribution of firms by receipts size, indicating that firms owned by men have higher levels of receipts than those owned by women. Women-owned firms are much more apt than male-own firms to have receipts of less than US$5,000.

In summary, estimates from the CBO indicate that female-owned businesses are more likely to close, less likely to have profits of at least US$10,000, and less likely to hire employees than businesses owned by men. Female firms also have mean annual sales that are roughly 80% of male levels.

Previous studies of female/male disparities in business performance indicate similar results. For example, Srinivasan et al. (1994) use data from the National Federation of Independent Business (NFIB), finding that women-owned firms have a higher probability of closure and a lower probability of growth than male-owned firms. Using 1992 Survey of Minority-Owned Business Enterprises (SMOBE) and CBO data, Robb (2000) finds that women own just over one-quarter of businesses with employees and generate less than 20% of employer firm receipts. Woman-owned firms are also found to have lower survival rates than male-owned firms. Using earlier CBO data, Boden and Nucci (2000) find that businesses owned by women are less likely to survive than businesses owned by men in both years.³ Using the longitudinal Kauffman Firm Survey data on new firms, Robb (2008) finds firms owned by women have lower sales, profits, employment, and survival rates than those owned by men.

In this study, we focus on the factors measurable with CBO microdata, such as human capital, business human capital, and financial capital. The standard economic model predicts that these factors are important inputs in a firm’s production process. The models we estimate are relatively parsimonious specifications that focus on the more exogenous owner and firm characteristics that predict business success. Once the owner and firm characteristics that are associated with business success are identified, we can use the rich CBO data to estimate how gender differences in these factors contribute to female/male differences in business outcomes.

The CBO data contain information on four major business outcomes—closure, profits, employment, and sales. Although none of these measures alone represents a perfect, universally agreed upon measure of business success, taken together they provide a fairly comprehensive picture of what it means to be successful in business. Logit and linear regression models are estimated for the probability of a business closure from 1992 to 1996, the probability that the firm has profits of at least US$10,000 per year, the probability of having employees, and log sales.⁴ Table 2 reports estimates of marginal effects for the logit regressions and coefficients for the ordinary least squares (OLS) regression. Because of concerns regarding potential endogeneity, we follow the approach taken in many previous studies of self-employment, reporting estimates from separate sets of regression models that exclude and include startup capital and industry controls.⁵ We discuss the results without startup capital and industry controls first.

Race and ethnicity are important determinants of small business outcomes.⁶ In the regressions, white is the excluded race category, and the included dummy variables are black, Hispanic, Native-American, and Asian. Thus, the interpretation of the coefficient on each variable is the remaining difference between whites and that minority group in the business outcome. For example, the coefficient on the black-owned business variable in Specification 3 implies that black-owned firms are 9.51 percentage points less likely to hire an employee than are white-owned firms, even after controlling for differences in other variables included in the regressions. After controlling for numerous owner and business characteristics, black-owned businesses continue to lag behind white-owned businesses. In all specifications except the closure probability equation, the coefficient estimate on the black-owned business dummy variable is large, positive, and statistically significant. In the closure probability equation, the coefficient estimate is positive, but statistically insignificant. The results are more mixed for Latino-owned firms. They have a lower probability of having large profits, but have a higher probability of hiring employees than white-owned firms. The coefficient estimates in the other two specifications are statistically insignificant. Asian- and Native American-owned businesses generally have better outcomes than white-owned businesses after controlling for the included variables. However, in the next set of regressions, which include startup capital and industry controls, the positive Asian coefficients essentially disappear. The black coefficients also become noticeably smaller after the inclusion of these additional variables.

Female-owned businesses continue to have lower measures of business outcomes than male-owned businesses after controlling for the included owner and business characteristics. The finding of relatively large and statistically significant coefficients on all of the female dummy variables indicates that the included controls for education, family background, work experience, and other owner and firm characteristics cannot entirely explain gender differences in business outcomes. By comparing these to the original gender differences in the business outcomes reported in Table 1, we can get a sense of how much of the gender disparities in business outcomes are explained by gender differences in all of the included owner and business characteristics. However, the current estimates do not reveal the relative importance of each of the owner and business characteristics in explaining gender differences in business outcomes. Before addressing this question, we will continue the general discussion of identifying the determinants of business outcomes.

Similar to previous studies, we find that small business outcomes are positively associated with the education level of the business owner (Bates 1997; Astebro and Bernhardt 2003; Headd 2003; Robb 2008). Estimates from the CBO indicate that owner’s education improves all four of the available business outcomes. For example, compared with businesses with owners that have dropped out of high school, businesses with college-educated owners are 5.5 percentage points less likely to close, 11.3 percentage points more likely to have profits of US$10,000 or more, and 6.1 percentage points more likely to have employees, and they have approximately 25% higher sales. Owners who have completed graduate school are 10.4 percentage points more likely to hire employees and have sales that are roughly 37% higher than those of businesses owned by college graduates. In fact, we generally see better business outcomes with each higher level of education.

Also, firms located in urban areas are more likely to close and are less likely to have employees, but are more likely to have large profits and have higher sales than firms located in non-urban areas. Previous work experience has mixed effects across outcome measures, although we find some evidence that suggests individuals with 20 or more years or very few years of prior work experience have worse outcomes, on average.

Having a family business background is important for small business outcomes (see Fairlie and Robb 2007a for more details). The main effect, however, appears to be through the informal learning or apprenticeship type training that occurs in working in a family business and not from simply having a self-employed family member. The coefficient estimates on the dummy variable indicating whether the owner had a family member who owned a business are small and statistically insignificant in all of the specifications except for the closure probability equation. In contrast, working at this family member’s business has a large positive and statistically significant effect in all specifications. The probability of a business closure is 0.042 lower, the probability of large profits is 0.032 higher, the probability of employment is 0.055 higher, and sales are roughly 40% higher if the business owner had worked for one of his/her self-employed family members prior to starting the business.⁷ The effects on the closure, profit, and employment probabilities represent 15.3–26.6% of the sample mean for the dependent variables.

Inherited businesses are more successful and larger than non-inherited businesses. The coefficients are large, positive (negative in the closure equation), and statistically significant in all specifications. However, because inheritances only make up 1.6% of all small businesses, their importance in determining broad business outcomes is slight, and their role in the differences in business outcomes by gender is probably minor.

The CBO also provides detailed information on other forms of acquiring general and specific business human capital. Available questions include information on prior work experience in a managerial capacity and prior work experience in a business whose goods and services were similar to those provided by the owner’s business. Management experience prior to starting or acquiring a business generally improves business outcomes but does not have a consistent effect across specifications. In contrast, prior work experience in a similar business, which provides specific business human capital, is an important determinant of business success. In all specifications, the coefficient estimates are large (negative in the closure equation), positive, and statistically significant.

We estimate a second set of small business outcome regressions that include dummy variables for different levels of startup capital and major industry categories. These estimates are reported in Table 3. As expected, business outcomes are positively associated with the amount of capital used to start the business. The coefficients on the startup capital dummies are large, positive (negative for the closure probability), and statistically significant in all specifications. In almost every specification, outcomes improve with each higher level of startup capital. The relationship between startup capital and business success is also strong for each type of business outcome. Perhaps the most interesting finding is the relationship between startup capital and closure. Firms with US$100,000 or more in startup capital are 23.0 percentage points less likely to close than are firms with less than US$5,000 in startup capital, and are 9.9 percentage points less likely to close than are firms with US$25,000–$99,999 in startup capital. These results hold even after controlling for detailed owner and firm characteristics, including business human capital and the industry of the firm. Owners who have less access to startup capital appear to start less successful businesses, which is consistent with the findings of previous studies (Bates 1997; Robb 2000; Headd 2003).

Industry is also linked to business success, as many of the dummy variables for industries are large in magnitude and statistically significant (retail trade is the left-out category). The estimates vary across specifications, however, making it difficult to summarize the association between industries and business outcomes.⁸

The regression analysis identifies several owner and firm characteristics that are strongly associated with business outcomes. The next question is whether female-owned businesses and male-owned businesses differ in these characteristics. Large differences between female and male firms in the key determinants of business success will contribute to differences in business outcomes. The exact contributions are estimated and discussed in the next section.

To explore differences between female- and male-owned businesses, we first examine the owner’s education level, which is found to be an important determinant of business outcomes. Female business owners are not clearly more or less educated than male business owners. As illustrated in Fig. 1, a lower percentage of female business owners are high school dropouts than male business owners (8.3% compared with 11.6%), but a lower percentage of female business owners have graduate degrees than male business owners (11.4% compared with 14.8%). In the middle of the distribution, female owners are more likely to have some college and college degrees than male owners. Overall, it is difficult to know whether female or male owners have an educational advantage in terms of business outcomes.⁹ The decomposition discussed below will provide direct evidence on this question.

Estimates from the CBO indicate that female and male primary business owners have different family business backgrounds. Table 4 reports the percentage of owners that had a family member who was a business owner and the percentage of owners that worked for that family member. As expected, female and male business owners do not differ substantially in the percentage reporting that they had a family member who owned a business prior to starting their business. However, female business owners are less likely to have worked in the family business than male business owners. Estimates from the CBO indicate that conditional on having a self-employed family member, only 38.3% of female business owners had worked for that person’s business, whereas 46.2% of male business owners who had a self-employed family member worked for that person’s business.¹⁰ The result is that female business owners overall were less likely than male business owners to work for a family member’s business. The unconditional rate of working for family member’s business was 19.4% for women and 24.0% for men.¹¹ Given the findings that work experience in a family business may provide important opportunities for acquiring general and specific business human capital (Lentz and Laband 1990; Fairlie and Robb 2007a), these gender differences will contribute to gender differences in business outcomes.

Inheritance was an infrequent source of business ownership, with only 1.4% of female business owners and 1.7% of male business owners citing this as the source of their businesses. As expected, the low levels of business inheritances suggest that it does not contribute substantially to gender differences in business outcomes.

Related to the family business background of the owner, being married is associated with business success. Spouses may provide financial assistance, paid or unpaid labor for the business, health insurance coverage, and other types of assistance useful for running a business. Estimates from the CBO indicate that 76.4% of female business owners are married, compared with 81.7% of male business owners (see Appendix).

For other types of business human capital, estimates from CBO microdata indicate that female and male business owners have roughly similar management experience. As indicated in Table 4, 52.3% of female business owners and 56.6% of male business owners have previous work experience in a managerial capacity prior to owning their current business. This type of experience provides an opportunity to gain professional and management experience useful in running future business ventures.

Although managerial experience in general is roughly similar, female business owners have much less prior work experience working in business(es) similar to their own. Forty-two percent of female business owners previously worked in a business that provided similar goods or services as the businesses they currently own; the rate for male business owners was 53.8%. As noted above, the effects of this type of prior work experience are large, and thus may explain part of the gender gap in business performance.

Although not reported, the regression models also included a measure of the number of years of work experience prior to starting the business. We find similar distributions of years of prior work experience by gender. The effects across outcome measures for this variable are also mixed, suggesting that it cannot contribute substantially to gender difference in outcomes.

Estimates from the CBO indicate that women started their businesses with less capital than men. Figure 2 indicates that for each level of startup capital above US$5,000, there are a lower percentage of female-owned businesses than male-owned businesses. The difference, however, at the highest startup capital level ($100,000 and more) is relatively small. For women, 4.1% started with more than US$100,000 in capital compared with 5.5% of male-owned businesses. Slightly more than two-thirds of female-owned firms were started with less than US$5,000 compared with 56.7% of male-owned firms.

Table 5 shows the distribution of firms by industry for female and male-owned firms. Female firms are much less frequently found in construction than male firms. The difference is large—only 3.3% of female firms are in construction, compared with 16.3% of male firms. On the other hand, female-owned businesses are more likely to be found in retail trade, personal services, and professional services than male-owned businesses. The decompositions in the next section will shed light on whether these industry differences contribute to differences in business outcomes.

Estimates from the CBO indicate that female business owners differ from male owners for many characteristics, such as prior work experience and industry. The estimates reported in Tables 2 and 3 also indicate that many of these variables are important determinants of small business outcomes. Taken together, these results suggest that gender differences in prior work experience and startup capital contribute to why female-owned businesses have worse outcomes on average than male-owned businesses. The impact of each factor, however, is difficult to summarize. In particular, we wish to identify the separate contributions from gender differences in the distributions of all of the variables or subsets of variables included in the regressions.

To explore these issues further, we employ the familiar technique of decomposing inter-group differences in a dependent variable into those due to different observable characteristics across groups and those due to different “prices” of characteristics of groups (see Blinder 1973; Oaxaca 1973).¹² The standard Blinder–Oaxaca decomposition is used with the marginal effects from the logit specifications for closure, profits and employment, and the coefficients for the linear log sales specification.¹³ Similar to most recent studies applying the decomposition technique, we focus on estimating the first component of the decomposition that captures contributions from differences in observable characteristics or “endowments.” We do not report estimates for the second or “unexplained” component of the decomposition because it partly captures contributions from group differences in unmeasurable characteristics and it is sensitive to the choice of left-out categories, making the results difficult to interpret (see Jones 1983; Cain 1986 for more discussion). Another issue that arises in calculating the decomposition is the choice of coefficients or weights for the first component of the decomposition. The first component can be calculated using either the white or minority coefficients, often providing different estimates, which is the familiar index problem with the Blinder–Oaxaca decomposition technique. An alternative method is to weight the first term of the decomposition expression using coefficient estimates from a pooled sample of the two groups (see Oaxaca and Ransom 1994 for example). We follow this approach to calculate the decompositions by using coefficient estimates from a logit regression that includes a sample of both men and women, as reported in Tables 2 and 3. As noted above, the coefficient estimates do not differ substantially by gender.

The contribution from gender differences in the characteristics can thus be written as:where \( \bar{X}^{j} \) are means of firm characteristics of gender j, \( \hat{\beta }^{*} \)is a vector of pooled coefficient estimates, and j = F or M for women and men, respectively. Equation 1 provides an estimate of the contribution to the gender gap of gender differences in the entire set of independent variables. Separate calculations are made to identify the contribution of group differences in specific variables to the gap.

Table 6 reports estimates from this procedure for decomposing the large female/male gaps in small business outcomes discussed above. The separate contributions from gender differences in each set of independent variables are reported. Gender differences in the racial ownership of the firm are relatively small and do not contribute substantially to the gaps in small business outcomes.

Gender differences in education levels explain part of the gap in business outcomes, but the effects are not consistently large. Educational differences explain 11.0% of the female/male gap in closure rates, but only 0.6% of the gap in profits. The higher concentration of female business owners in the middle of the educational distribution, relative to male business owners, appears to have a modest negative effect on business performance.

The lower percentage of female business owners who are married explains a small share of the business outcome gaps. Female business owners are less likely to be married than are male owners, and marriage is associated with business success. These results are difficult to interpret, however, given potentially different lifestyle choices that are interrelated with marriage. Regional and urbanicity differences are small between female and male firms, resulting in essentially zero contribution estimates. Gender differences in the amount of prior work experience and management experience are small, and thus do not have large effects or mixed effects across specifications.

As reported in Table 4, female business owners have the same likelihood of having a self-employed family member as male business owners. The result is that gender differences in this factor do not contribute to female/male disparities in survival, profits, employment, and sales. Likewise, gender differences in business inheritances also do not contribute to differences in business performance. In contrast, we find larger differences between female and male owners in whether they worked in a family business. Thus, the explanatory power of gender differences in prior work experience in a family member’s business is stronger (although not large). Gender differences in this variable explain 0.8–7.0% of the female/male gaps in small business outcomes. Apparently, the lack of work experience in family businesses among future female business owners, perhaps by restricting their acquisition of general and specific business human capital, limits the successfulness of their businesses relative to men’s businesses.

Providing some additional evidence on this point, gender differences in prior work experience in a business providing similar goods and services consistently explain part of the gaps in outcomes. Although the coefficient estimates in the small business outcome regressions were generally similar in magnitude to coefficient estimates on the family business work experience variable, the contributions from gender differences are larger. The gender disparity in the percentage of owners that worked in a business with similar goods and services is larger than the disparity in the percentage of owners that worked in a family business. These differences in prior work experience provide the largest contribution of any reported factor in the table, explaining from 5.9% to 17.3% of the gaps in business outcomes. The lack of prior work experience in a similar business among female owners may limit their acquisition of general and specific business human capital that is important for running successful businesses.

Differences between male- and female-owned businesses in terms of closure rates, profits, employment, and sales may be related to barriers to success for female-owned businesses. For example, Brush et al. (2004) note that female entrepreneurs have access to different business and investment social networks than male entrepreneurs.¹⁵ Differences in business outcomes, however, may also be related to gender differences in the goals and types of businesses and preferences for level of work activity. Previous research indicates that women who are married to self-employed men are more likely to be self-employed or enter self-employment and that the choice of self-employment is partly driven by the desire for flexible schedules and other family-related reasons for women more than men (Devine 1994; Boden 1996, 1999; Carr 1996; Bruce 1999; Robb 2000; Lohmann 2001; Lombard 2001). Female owners may have different goals for business growth and tolerances for taking risks associated with business growth (Cliff 1998).

Estimates from the CBO indicate that female-owned businesses have worse average outcomes than male-owned businesses. Female firms are 12.9% more likely to close, 52.6% less likely to have profits of at least US$10,000, and 31.1% less likely to hire employees than male firms. They also have mean annual sales that are roughly 80% lower than the mean sales of male-owned firms. Even conditioning on hours, we find that women-owned businesses have much lower levels of sales than businesses owned by men. Female business owners are less likely to have very low levels of education than male business owners, but they are also less likely to have graduate degrees.

Female business owners are also less likely to have prior work experience in a family business and prior work experience in a business providing similar goods and services. Because of these differences in prior work experience, female business owners may have had fewer opportunities to acquire the specific and general business human capital that is important for running successful businesses. Female businesses are also found to have relatively low levels of startup capital. Estimates from the CBO indicate that 13.3% of female-owned businesses started with more than $25,000 in capital, compared with 17.7% of male-owned firms. Finally, female businesses locate in different industries than male businesses. Female businesses are more likely to be in retail trade, personal services and professional services, and less likely to be in construction.

We use a decomposition technique to measure the contribution of gender differences in firm and owner characteristics to differences in business outcomes between female- and male-owned businesses. The decomposition estimates indicate that female-owned businesses are less successful than male-owned businesses because they use less startup capital, have less prior work experience in a similar business, and less prior work experience in a family business. Gender differences in industry distributions, however, are not a major explanation for female/male gaps in business outcomes.

As reported above, evidence from the USA and several other countries suggests that women are less likely than men to report having a desire for self-employment, although the difference is not large (Kourilsky and Walstad 1998; Blanchflower et al. 2001). In the end, unobservable factors, such as different preferences, discrimination, and risk aversion, may be responsible for low levels of female entrepreneurship and lower returns (Bird and Brush 2002; Carter et al. 2003). From a policy perspective, however, these are difficult to address. Policies that increase human capital and access to financial capital, such as entrepreneurial training and loan assistance programs, are easier to implement and expand.