Students' perceptions of work in public accounting and employment preferences

Click to increase image sizeClick to decrease image size

Notes

1. In general, recruiters for national public accounting firms in the Northeast United States use an average GPA of 3.0 or above for interviewing students. Smaller public accounting firms interview students with GPA of less than 3.0.

2. An alternative logistic regression model was developed which, in the initial stepwise procedures, included only the multi-item factors in Table 5 as independent variables. Single item variables in Table 2, X6–X26, were excluded in this analysis. The results of this stepwise procedure are similar to those reported in Table 6 with the notable exception of overtime pay (X6). Since X6 was not included in the initial stepwise procedure in this alternative approach, it does not appear in the final model.

3. Traditionally, a probability of 0.50 is used as the cut-off point in making predictions about the dependent variable in binary logistic models (Stock and Watson, 2003; Wooldridge, 2003). The advantage of this rule is that it is easy to understand. The disadvantage is that it may not reflect the true quality of the prediction. In some cases, prior knowledge about the optimal probability value can be used to determine the cut-off point. In the case of the current study, such information was not available. Other approaches in choosing a cut-off point include, (1) using sample proportions (2) examining the probability distributions of the two levels of the dependent variable using the techniques of Palepu (1986). In the current study, both approaches yield a cut-off point of approximately 0.37. Regarding sample proportions, 46 of 123 students or 36.5% prefer public accounting. Regarding probability distributions, a plot of the distributions for the two employment preferences (public accounting v. other) reveals an intersection point of 0.37, which according to Palepu (1986), is an appropriate cut-off. Using 0.37 as the cut-off probability produces results similar to those reported using the 0.50 cut-off. With 0.37 as the cut-off point, for the original sample (n = 123), the model correctly predicted 67.5 % of the students. For the additional samples, the model correctly predicted 67.6% (n = 34) and 68% (n = 50).

4. To check the validity of the parameter estimates and standard errors in Table 6, a bootstrapping technique was used. The bootstrapping procedure involved re-sampling with replacement and had 1000 iterations. The parameter estimates and standard errors from the bootstrapping are very similar to those reported in Table 6. Probit estimation of regression coefficients also yields results similar to those in Table 6. Using probit, all variables in Table 6 are significant at the 10% level. The pseudo r-square is 0.2767.

5. In the logistic regression model where βs are the parameters, exp(β_i)-1 measures the change in the odds for a one unit change in the relevant variable (keeping all other variables constant).