family member's business, had previous work experience in a managerial capacity, and
worked in a business providing similar goods and services.
Race and ethnicity are important determinants of small business outcomes (see
Fairlie and Robb 2003 for a more detailed analysis). Asian-owned businesses generally
have better outcomes than white-owned businesses. 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.
Similar to previous studies, we find that business outcomes are positively
associated with the education level of the business owner.14 For example, businesses
with college-educated owners have a 0.055 lower probability of closure, a 0.113 higher
probability of having large profits, a 0.060 higher probability of having employees, and
have approximately 25 percent higher sales on average than businesses with owners who
did not graduate from high school.15 Female-owned businesses are less successful and
are smaller on average than are male-owned businesses. Firms located in urban areas are
14 For example, using the 1982 CBO, Bates (1990b) finds that small business failures generally decrease with the education level of the owner. Similarly, Robb (2000) found increases in e d u c a t i o n t o b e p o s i t i v e l y a s s o c i a t e d w i t h b u s i n e s s s u r v i v a l . 1 5 The implied effects on the probability of closure, large profits, and employment are
approximated by multiplying the coefficient estimate from the logit model by p(1 p) , where p is the mean of the dependent variable.