Modelling the Effects of Public Support to Small Firms in the UK - paradise Gained?
In practice attrition within the Impact Indicators database is presented in Table 1.
Table 1: Attrition Effects within the Impact Indicators Database
Assisted firms:
|
Initial sample__________________________________________________________ |
892 | |
|
• Not found on IDBR/or present on the IDBR but outside the definition of |
- 497 |
395 |
|
• Firms that should have been excluded when checked__________________ |
-17^^ |
378^^ |
|
• Closed |
-55^^ |
323~ |
|
• Too few controls____________________________________________________ |
-76^ |
247~ |
|
Remained in the analysis____________________________________________ |
247 |
Matched Comparator Firms:
|
Initial sample__________________________________________________________ |
9408 | |
|
• Not found on respective BL CMS (excluded as might be known to BL |
-1520 | |
|
• Linked to assisted firms that did not meet definition of significantly |
-5193 |
2695 |
|
• Closed__________________________________________________________ |
-224 |
2471 |
|
• Too many visits during study period by their Business Link____________ |
-980 |
1491 |
|
• Loss of matched assisted firm________________________________________ |
-223 |
1258 |
|
• Matched assisted firm having too few matched comparators____________ |
-117 |
1141 |
|
Remained in the analysis____________________________________________ |
1141 |
Source: Impact Indicators Database/SBRC/NIERC/Prism Consolidated Database
The main implication of the level of attrition is the size of sample that is required even
when using administrative data that places no burden on the firms involved in the
study. The level of attrition can also to some extent be related to some degree of
human error especially in terms of applying the definition of significant assistance. It
might be concluded that any future study to this would need a national Client
Management System (CMS) for Business Links, which is populated, minimally, with
all businesses on the IDBR to be generally practicable.
Validity Checks
It was felt there were minimally three types of validity checks that needed to be
investigated before the econometric analysis could be undertaken. These were:
Comparisons with the Wider Business Population:
The sample was compared with the entire IDBR database to test whether the sample
of assisted businesses was significantly different from the wider business population.
Comparisons were restricted to industrial sector and employment size. The BL client
Stephen Roper and Mark Hart