Modelling the Effects of Public Support to Small Firms in the UK - paradise Gained?
Selection and Assistance Effects on Business Growth
The second stage of the econometric analysis is the identification of the ‘selection’
and ‘assistance’ effects on business growth conditioned (or allowing for) the impact
of other potential drivers of business growth. As previously the choice of explanatory
variables was conditioned by previous studies (e.g. Roper and Hewitt-Dundas, 2001)
and surveys of the determinants of business growth (e.g. Storey, 1994; Barkham et al.,
1996). The modelling approach adopted was again a ‘general-to-simple’ procedure
with variables being dropped successively to derive more parsimonious explanations.
In terms of the main parameters of interest - i.e. the selection and assistance effects
the results differ between the productivity growth and sales growth equations. Table 4
presents the results of the regression models for productivity growth.1 In terms of the
overarching priorities and PSA targets of the DTI this is the key measure of the
impact study.
It proved difficult to identify very robust equations for growth in productivity,
turnover and employment. This is reflected in low F statistics for the equations (see
Table 4) and limits the strength of any implications which can be drawn from the
models. In terms of the key 'selection' and 'assistance' effects, however, the
productivity growth models suggest that:
• Assistance Effect - BL assistance in 1996 had a positive and statistically
significant effect on productivity growth over the subsequent four- year
period.
• Selection Effect -a negative and marginally significant selection effect is
identified in regard of productivity growth.
Two factors suggest caution in terms of the positive productivity effect of BL
assistance. First, the overall weakness of the productivity equations suggests missing
variables which may be distorting the result. Second, the large size of the apparent
BL productivity effect (102% to 108% over 4 years) may suggest that the assistance
dummy variable is also picking up the effect of other unrelated influences on business
performance. What is clear from the model, however, is that BL assistance is
effectively being targeted at firms which without support would have had lower than
expected productivity growth. One possible explanation is that firms which perceived
that they had some form of productivity problem were more likely to seek out BL
assistance as a means of ‘catching-up’ with their competitors. The positive aspect of
this is that the assistance provided by BL does seem to have accelerated productivity
growth in these firms. In terms of sales growth the results are less satisfying although
BL assistance was still seen to be having a small positive effect.
In methodological terms these results are also of some interest. In particular, the
productivity equations highlight the value of allowing separately for the ‘assistance’
and ‘selection’ effects. The negative selection effect would otherwise have biased
1 Equations were also estimated for turnover and employment growth as part of the full analysis for the
SBS and are not reported here due to space constraints. However, it should be noted that they are less
robust than the productivity equations.
Stephen Roper and Mark Hart
17