Modelling the Effects of Public Support to Small Firms in the UK - paradise Gained?
database was not a representative cross-section of the business community as a whole
being, not surprisingly given the focus of the BL initiative, more concentrated in the
production sector (e.g., two-fifths compared to 10%) and more likely to employ more
people. However, the real issue for the analysis is whether the groups (assisted
businesses and their matched comparators) are similar in structure so that robust
comparisons can be made.
Comparisons between Assisted and Comparator groups:
Evidence was examined within the Impact Indicators database, as updated and re-
checked, whether the Assisted and Comparator groups remained sufficiently similar
to each other. Comparison on prior growth, age, sector, and structure including chi-
squared-tests showed no major differences between the characteristics of the
remaining assisted and non-assisted comparator businesses to render any further
survey and subsequent econometric work inappropriate.
Comparison with Wider 'Assisted Business' Population
Unfortunately no suitable database exists to facilitate this comparison, although the
Small Business Service plans to introduce a national client database which would
prove invaluable to such a comparison.
Overall, the creation and analysis of the of the Impact Indicators database by
SBRC/NIERC/Prism Research demonstrated that there is validity in continuing with
the study on the grounds that there are:
• sufficient assisted and associated comparator non-assisted businesses in the
database to facilitate the construction of a survey sampling frame for the next
stage of the study (survey and econometric modelling);
• no major differences between the characteristics of the remaining assisted and
comparator non-assisted businesses to render any further survey and
subsequent econometric work inappropriate.
THE IMPACT INDICATORS SURVEY
In order to obtain the additional set of 'conditioning' variables necessary for the
econometric modelling of 'selection' and 'assistance' effects a total of 366 businesses
were interviewed either face-to-face or by telephone between October and December
2000. This total included 165 assisted and 201 non-assisted comparator businesses.
However, this total includes 33 businesses which did not match the original selection
criterion of each assisted business having at least 3 valid comparators in the database.
It was, therefore, decided to undertake the econometric analysis with the 333
companies (137 Actions and 196 Controls) which conform to the original selection
criterion. The 62.2 per cent response rate among valid assisted firms (i.e., 137 out of a
total of 220 businesses) is significantly above that usually expected in this type of
interview survey.
Stephen Roper and Mark Hart
10