4 Firms location is given by the location of the respective incubator ie by

4 firms location is given by the location of the

This preview shows page 12 - 14 out of 30 pages.

4. Firms’ location is given by the location of the respective incubator, i.e. by cities. In a next step, Creditreform provided for each of the five locations data on all firms in their database that have been founded up to the end of 2006 and matches one of the main industry-groups defined above. Using the same database for treatment observations and control observations is an important aspect (Caliendo and Kopeinig 2005 ; Heckman et al. 1999 ), though often neglected. Over all five locations, 43,467 potential control firms were identified that did not receive public support by an incubator. ‘‘Appendix 3 ’’ shows that the ‘On-incubator’ sample differs significantly in important characteristics from the group of potential controls. Thus, incubator firms are a selective group of firms and incubator support is not arbitrary. 7 ‘Hightech-Manufacturing’ (NACE Rev. 2 codes 20–37), ‘Wholesale trade and retail trade’ (51, 52), ‘Construction’ (45), ‘Computer’ (including hard- and software, 72), ‘Research and development’ (73), ‘Consulting and business-related services (BRS)’ (including engineering consultants, 74), ‘Education’ (80) and ‘Recreation/sports/culture/others’ (including also non-knowledge based services like, for example, call- center and facility management 90–93). A control group study of incubators 313 123
Image of page 12
To avoid biased results due to these significant differences, we apply propensity score matching. For every observation from the ‘On-incubator’ sample and the group of potential controls, the vector of exogenous variables is condensed into one single measure: the so called propensity score. In the present application, the propensity score reflects the like- lihood that a firm i has received support by an incubator, conditional on a set of individual characteristics x i : Pr ( S i = 1 | X = x i ), with x i given by the variables defined above. Pro- pensity scores are estimated using probit models. Relevant exogenous variables (industry affiliation, legal form, start-up period) are regressed on a binary dependent variable indi- cating as to whether a firm were incubated or not. Location was not included as exogenous variable. Since we are interested in a differentiated analysis, we performed estimation of propensity scores for each incubator location separately. This procedure is particularly recommended if heterogeneous effects for sub-populations are expected (Caliendo and Kopeinig 2005 ). The respective estimation results are not further discussed here, but are provided in the ‘‘Appendices 1 and 2 ’’. According to Caliendo and Kopeinig ( 2005 ) all variables are fixed over time, measured at start-up. There are three general assumptions in the context of matching. First, the conditional independence assumption assumes that all relevant exogenous variables affecting both treatment and outcomes (survival) are observed. Second, the stable unit treatment value assumption demands that treatment of one particular firm (incubation) does not affect the outcome of other firms (survival).
Image of page 13
Image of page 14

You've reached the end of your free preview.

Want to read all 30 pages?

  • Summer '17
  • The Land, Incubation period, Survival analysis, Propensity score, Incubation, Business incubator, Michael Schwartz

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture