Modelling the Effects of Public Support to Small Firms in the UK - paradise Gained?
The modelling approach adopted is a ‘general to simple’ modelling approach where
all variables of interest are included in the models and then subsequently eliminated to
obtain a more parsimonious representation of the underlying data. The main criteria
for elimination of variables is statistical significance although in some cases variables
were eliminated or combined because of correlation or the similarity of the estimated
coefficients. In the final equations in Table 3 variables were generally left in the
equations if they were of particular interest or had a t-statistic in excess of one.
Five main groups of variables were included in the estimation:
(a) Firm characteristics including the legal status of the company, size, prior
growth (intended to pick up any orientation towards backing winners) and the
age of the firm.
(b) Informational sources by which the firm might have made contact with BL.
These included other companies known to have received support from BL, BL
mailshots etc, magazine articles etc.
(c) Owner-manager characteristics including age and qualifications and whether
the owner-manager of the business was the founder of the business.
(d) A set of industry dummies designed to control for any sectoral targeting in BL
support or differences in the propensity of firms in different sectors to make
use of BL services.
(e) A set of dummy variables designed to control for any geographical differences
in the intensity of BL support.
The factors included in groups (a) and (c) in particular were chosen following earlier
similar studies (e.g. Roper and Hewitt-Dundas, 2001). Factors included in group (b)
reflect the sources of information on BL highlighted in PACEC (1998). One issue
that arises with the measurement of factors such as the characteristics of the owner-
manager, information sources used by the firm etc., is that these reflect the situation at
the time of the interview with firms in 2000. In the probit analysis we are essentially
assuming that these variables were similar in 1996 and are therefore valid
conditioning factors for assistance by BL. The less attractive alternative, due mainly
to reasons of accurate recall on behalf of the owner-manager, would have been to ask
firms about their activities in 1996.
The final estimated equations, the results of the general to simple modelling
procedure, are in Table 3. The first two equations contain the same factors with the
difference being the inclusion of turnover growth or employment growth to represent
the previous growth performance of the business. The third equation in the table
represents a further simplification of the model including only a limited range of the
more statistically significant variables. It is this version of the probit model which is
used in the sample selection modelling reported in the next section.
Stephen Roper and Mark Hart
14
More intriguing information
1. Protocol for Past BP: a randomised controlled trial of different blood pressure targets for people with a history of stroke of transient ischaemic attack (TIA) in primary care2. The name is absent
3. The name is absent
4. Urban Green Space Policies: Performance and Success Conditions in European Cities
5. ANTI-COMPETITIVE FINANCIAL CONTRACTING: THE DESIGN OF FINANCIAL CLAIMS.
6. Improving behaviour classification consistency: a technique from biological taxonomy
7. The name is absent
8. The name is absent
9. Quelles politiques de développement durable au Mali et à Madagascar ?
10. Altruism and fairness in a public pension system