Personal Experience: A Most Vicious and Limited Circle!?
13
What can we learn from a comparison of the separate specifications? Model
A can give an indication of how firms with a higher probability of closure are
characterized. Results of the single failure risk specification, model B, further
reveal which firms fail to a higher degree, that is more often suffer fatal financial
distress, independent of whether they be legally forced to do so or not. The sepa-
ration of bankruptcy also enables us to identify differences in the effects if only
an objective type of failure is considered.
Table 3: Overview of the estimation specifications
|
Specification Observations treated as closures |
Observations treated as censored |
|
Model A All firms which were closed. |
All firms active at the end of the year |
|
Model B All firms which failed, i.e. those which were either voluntarily closed due to financial distress or went bankrupt |
All firms active at the end of the year Firms voluntarily closed without being |
|
Model C Firms which went bankrupt |
All firms active at the end of the year All firms voluntarily closed |
Estimation results
Table 4 shows the estimation results for the Cox regressions. It displays the
estimated coefficients, robust standard errors, and the 95 percent confidence in-
tervals that result from maximizing the Cox log-likelihood function. The latter
are displayed due to the large number of observations introduced in the regres-
sions, which cause increased test statistic values leading to many highly signifi-
cant results. Readers should have the chance to come to their own conclusions
about the meaningfulness of the claimed effects (Cohen 1994).
Model A refers to the single risk specification in which all exit types are
pooled. Firms in which at least one of the entrepreneurs involved made a restart
after business failure, independent of the failure type, are more likely to be
closed again than novice firms. The closure probability of restarts after closing a
firm in financial distress is on average 37 percent higher than that of novice
firms.5 Restarts after bankruptcy are faced with a probability of closure, which is
raised by an average of 45 percent. They thus behave differently from other re-
starts because restarts after sale of shareholdings and restarts after closing a
sound firm voluntarily do not differ from novice entrepreneurs’ firms in their
probability of closure. Similarly, having entrepreneurial experience in the restart
industry does not significantly affect the likelihood of closure, but having multi-
ple entrepreneurial experiences does, raising the probability of closure.
Model B refers to the single risk specification in which only the risks of firm
failure are pooled. The results are very similar to the previous single risk specifi-
cation, although closure is now is synonymous with failure. Restarts after sale of
shareholdings do not differ significantly in their probability of failure