dresses, if the attempt to contact the firm was unsuccessful. Applying this approach, also
already closed firms could have been surveyed, which is a distinctive feature of this
study.
With 3,702 complete interviews out of the 12,000 sample observations, the response
rate was about 31 percent, which is high compared to other German studies conducted
by telephone. 1,008 cases of all observation refer to businesses that were already closed
at the time of the survey.
In 237 cases out of the about one thousand observations the interviewees stated that a
former entrepreneur of the closed firm restarted, i.e. established a new business in the
meanwhile. Running the analyses with this information, however, would approve two
limitations. Firstly, a number of interviewees said that they actually do not know about
the venture activities of the closed firms’ former entrepreneurs. Secondly, a long time
passed since the survey was conducted resulting in outdated information about the ven-
ture activities gathered in the study. Either of both issues advises to look for additional
information about the venture activities of the relevant individuals. To do so, the entre-
preneurs who participated in a closed venture were traced in the data deliveries of
Creditreform following subsequently to the state of the ZEW Foundation Panel which
forms the parent population of the survey. With this approach it was possible to identify
225 additional restarts up top the year 2007 which wouldn’t have been considered refer-
ring solely to the survey data.
Selection of variables
The hypotheses are tested within a multivariate framework. The occurrence of an entre-
preneurial restart is the dependent variable in a probit estimation. The dependent vari-
able RESTART equals one if an entrepreneur of a closed business’ entrepreneurial team
ventures anew and zero otherwise. Two different specifications are fitted. In the first
specification the second set of main covariates, i.e. the variables indicating which group
of stakeholders incurs financial losses, is omitted and then introduced in the second
specification. This enables to identify possible effect which may arise from an omitted
variable bias if neglecting these variables.
As shown in Table 1 the introduced independent variables can be subdivided into main
covariates and other control variables. Two primary sets of explanatory variables are in-
troduced to account for different closure reasons and for the stakeholder groups which
have suffered financial losses from closure. Other ancillary independent variables control
for specific firm characteristics; these include governance, legal status, size, location and
industry affiliation