family member are not simply capturing the effects of management experience or specific
business human capital on small business outcomes. Instead, prior work experience in a
family member's business has an independent effect on small business outcomes, which
may in part be due to the acquisition of less specific, general business human capital.
A concern with the estimates reported in Table 2 is the amount of missing data for
some of the independent variables in the CBO. Approximately 10 percent of the
observations for each of the specifications reported in Table 3 are excluded because of
missing values for one or more of the independent variables. Although these levels of
missing data are not extremely high, we examine the sensitivity of our results to two
alternative methods of correcting for missing data. First, we estimate regressions in
which dummy variables are included for missing values of specific independent
variables.18 For example, if the education level of the business owner is missing then the
four education level dummy variables would be equal to zero and a special missing
education dummy variable would be equal to one. Thus, the missing observation for
owner's education would not contribute to the coefficient estimates on the main education
level dummies, but would contribute to coefficient estimates on other variables. This
technique is becoming increasingly common in the literature because it is easy to
implement and allows for an increase in the efficiency of some coefficient estimates.
Although not reported, we find estimates that are similar to the ones reported in Table 2
for all four specifications.
18 Race, gender, region, and urban are from administrative record data and have no missing values.