We’ve launched two new evidence reviews today. Both cover aspects of innovation policy, one looking at the impact of grants, loans and subsidies to support R&D, the other looking at tax credits. Perhaps more than any other area we’ve looked at these reviews only scratch the surface in terms of considering the effects of a very complex set of interlinked policies. That said, I do think the reviews highlight a number of important issues.
First, these specific approaches are popular and account for a considerable amount of public expenditure, so it’s right to ask whether they achieve their stated objectives. Looking first at the direct effect on R&D we see positive effects for 10 out of 17 evaluations of tax credits. Overall this is, arguably, encouraging. However, raising R&D is a specified programme objective in these programmes. So it is of some concern that success rates are not even higher than this. One possible explanation is that in the schemes where evaluations don’t find effects, tax credits are not large enough to affect firm behaviour very much or at all (and any responses may not be large enough to be picked up by the evaluations). An alternative is that these studies aren’t methodologically strong enough to pick up the effects. Interestingly, in this case, weaker methodologies may understate the effects (in strong contrast to many other areas). Indeed, the higher quality evaluations consistently show more positive effects. A third possibility is that tax credits may increase the price of R&D inputs rather than increase the quantity of R&D.
The findings are a little less positive for grants, loans and subsidies with only eight out of 18 evaluations finding positive effects on R&D. If we take the result at face value one explanation is that these programmes crowd out private sector R&D. In practice, we know that some schemes support firms, others support universities and some support public-private collaborations, so any actual crowding out may be more limited than this. It is also possible that public spending accounts for a small percentage of total R&D spending in supported firms, which might make it hard for some evaluations to detect relatively small positive effects that are statistically significant. Seven of the evaluations that have information on private funded R&D (rather than total R&D) are able to look at this issue directly. These seven studies therefore provide some reassurance on the extent of crowding out – in fact finding evidence of small ‘crowding-in’ effects, that is, public R&D spending encourages further private sector R&D activity. In turn, those results are consistent with the wider econometric literature, and with economic theory, which emphasises the need for government to partially fund firms’ discovery and commercialisation costs. That said, on the basis of the available evaluation evidence, the extent of crowding out remains an open question and it would be good to see further evaluation work that considers this issue.
For grants, loans and subsidies we can go a little further on aspects of programmes that might be linked to success. R&D subsidies are more likely to improve outcomes for small to medium size companies than for larger ones. In part this may be because for larger firms, public support makes up a relatively small amount of overall R&D spend, so positive effects are harder to detect. Smaller firms may also be more likely to formalise processes in anticipation of, or response to, a grant, so that some innovation-related spend is reclassified as R&D. The evidence also suggests that programmes that emphasise collaboration perform better than those that just support private firms (as well as those where the programme focus is unclear). Encouraging collaboration might have an additional positive effect on the likelihood that an R&D support programme generates positive effects on outcomes of interest. Finally, there is some evidence that programmes that target particular production sectors appear to do slightly worse in terms of increasing R&D expenditure and innovation. As is often the case, all of these findings are based on a relatively small number of studies and it would be good to see more work to better understand how these design features influence success.
So far, I’ve considered the effect only on R&D expenditures, rather than local economic growth. Unfortunately, the evidence does urge some caution on the role that more localised innovation policy could play in delivering the latter objective.
Local decision makers need to think carefully about their desired objectives. For example, as just discussed, our review shows that tax credits have a pretty good success rate in raising R&D spending (particularly for smaller / younger firms). Equally, R&D grants programmes which include a collaboration element seem effective at raising R&D activity. But in both cases we know much less about whether or how this increased R&D activity feeds through to greater innovation, better firm performance or longer term economic growth, particularly at the local level (because relatively few evaluations look at these effects). These broader outcomes are the things most local economic decision makers ultimately care about.
There are also good reasons to think that many of these broader economic benefits are likely to ‘spillover’ beyond the immediate area in which the policy is implemented. This might still result in a net benefit for the place implementing the policy, but such spillovers reduce the economic benefits to individual areas and strengthen the case for national policy.
Local R&D support programmes could also result in inefficiently high levels of support if footloose firms are able to extract more generous support from competing local areas regardless of any net beneficial impact. Any moves to devolve policy in the UK would need to test for these issues.
Overall, then, it is important to remember that evaluation of the impact of innovation policy is still limited and our reviews raise as many questions as answers. The limited evidence base, particularly in terms of the impact on local economic outcomes, highlights the need for realism about the capacity and evidence challenges of delivering innovation policy at a more local level.