4.2 Descriptive statistics and test of differences
In addition to testing the model, the 8 variables in the model were tested for differences. Table 2 provides the descriptive
statistics (mean and standard deviation, or percentage of the sample) for each variable. To test for differences, success,
mediocre, and failed performance were used as the independent variable and each of the 8 variables in the model were
used as the dependent variables. Chi-square was run for the three variables with dummy values. The one-way ANOVA
was run to compare mean differences between the successful, mediocre, and failed firms for the other five interval level
variables. The results of the test of differences between the successful, mediocre, and failed businesses supports the
model. In all but three of the variables (education, partners, and parents), the mean or proportion percentage differences
were significant, as seen in Table 2. The successful firms had a higher proportion using the Internet, they started with
more working capital, they kept updated and accurate financial and accounting information, they developed more detailed
plans, and they pursued marketing efforts.
Place Table 2 about here
Although not significantly different, the successful business owners have a higher level of education. The lack of
significant difference may be due to the fact that there is no straight correlation between entrepreneurship and education.
Examining the descriptive statistics, when starting business, the sampled entrepreneurs had an average of 2.9 years of
college. Most respondents have undertaken entrepreneur and management activities before their new venture. Those who
worked at the employee level previously did so for an average of 8.5 years before starting their own business at the age of
34. Education variability among entrepreneurs is high. Some start a business with just an elementary school education,
whereas others have completed graduate studies. Because there are exceptions with low levels of education, does not
mean that education is not important. Further research is required on this point.
4.3 Ordered probit regression model test results
Ordered probit regression model test results are presented in Table 3. As shown, the model is significant and all the
parameter estimates beta coefficients, except for education, are significant, and 5 of the 8 variables are significant at
the .01 level. Thus, these variables are used to estimate their marginal effects. The ordered probit regression result testing
the model -2 log likelihood (LL) was -805.15 and the Chi-square was 66.65, with the model significance level (p = .000).
Place Table 3 about here
The classification results show that for a typical firm, which adopts mean values for all the X vector, the
expected probability of success is 34%, the odds of showing a mediocre performance are 28%, and those of pursuing an
unsuccessful venture are 38%. The model is also useful at predicting the probability of success of any firm. For example,
if one takes the median values for the X vector, instead of the mean values, the estimated probabilities are respectively
8
Convention