How to manage public R&D spending in times of budgetary austerity
The dangerous cocktail of high debt and low growth in Europe calls for smart means of public spending. The need is for public investment that fosters long-term growth while minimising the potentially negative short-term effect on public finances, and R&D (Research & Development) is an area typically identified as a good candidate for smart public spending. Recent growth models make a strong case for R&D and innovation being areas which can contribute to growth, but this does not make the case for public investment which has to lead to sufficient innovation and growth to cover the opportunity costs of using public funds for R&D. To see whether these costs can be covered, we first look at the evidence from micro-economic analysis of the impact of public spending on private R&D and innovation. Public budgets for R&D are spent in a variety of ways, including funding universities and public research institutes, but the bulk goes to support private R&D, sometimes in collaboration with other public actors. These private R&D support programmes are the focus of this review (based on Veugelers, 2016). We also look at how public R&D affects GDP growth and jobs in currently applied macro-models. Although we can recommend designs for smart public R&D programmes from the reviewed micro and macro evidence, we need more evaluations as we still know too little of their actual effects.
1. The theory behind public intervention in R&D
The basic justification for government support of research is the classic argument of market failure: markets do not provide sufficient incentives for private investment in research. New knowledge created from R&D is “non-rivalrous”, meaning it is publicly accessible and can only partially exclude access: others may learn and use, without necessarily paying for, the knowledge, considered “spillovers”. Ignored by private investors, these spillovers lead to social rates of return above private rates of return, and private investment is below the socially desired levels.
Apart from spillover, another market failure follows on from the highly risky and uncertain nature of R&D outcomes. This uncertainty, coupled with imbalance of information between capital markets and R&D investors, causes financial market flaws, blocking access to finance for risky innovation projects. This holds particularly for small, new innovators.
The wide scope for market failure in the case of R&D investments for growth theoretically justifies government intervention, in order to bring private R&D investments closer to socially optimal investment levels. However, there are several reasons why these interventions may not be effective. First, public funding may directly substitute funding of projects that would otherwise have been privately financed. Second, further R&D generated by public funding may indirectly crowd out private R&D by increasing the input demand for research, leading to higher costs. This means that R&D support may result mainly in higher wages for researchers. Third, policy ideally triggers research projects with the highest social rates of return: but this assumes that the government is sufficiently informed about these rates of return, which is notoriously difficult. And finally, there is the problem of political capture, biasing project selection away from social value.
2. The evidence for the gap between private and social rates of return
Estimates of social rates of return are scarce. Most of the available empirical evidence is from specific cases. By and large, the literature indicates that the social or economy-wide returns of R&D are much higher than the private returns to individual firms (Hall & Mairesse 2009). A well-known reference is the work by Mansfield et al. (1977) who, computing the rates of return of 17 industrial innovations, found an average social rate of return of about 56% against a private rate of about 25%.
Rather than looking at social rates of return, it is also possible to look directly at the evidence on knowledge spillovers. Nowadays, cross-country and cross-sector technology spillover matrices, based on information from patents citing other patents as prior reference, are the most popular. For instance, Dechezlepretre et al (2014), looking at patent citation information to assess knowledge spillover from clean versus dirty technology, found that clean patented inventions received 43% more citations than dirty inventions. These results demonstrate why R&D for clean technologies should receive higher public support.
3. Government support for R&D
The evidence on high social returns and on technological spillovers can justify public intervention, but still does not make the case for public R&D funding: potential government failure needs to be analysed. In the following sections, we look at the evidence from econometric analyses on the effectiveness of subsidies and tax credits for R&D, two major tools for Research and Technological Development (RTD) policies.
R&D tax credits
A virtue of R&D tax credits relative to subsidies is that it lets the firm choose the projects. As any firm satisfying the criteria can use it, it economizes on bureaucratic decision making. Currently, tax credit schemes abound, applied to the total amount of R&D expenses (volume-based) or only additional R&D over a period of time (increment based).
Stimulation of private R&D by fiscal incentives is typically measured by the tax price elasticity: the additional R&D generated by one euro of tax deduction. In a review of the literature, Mohnen (2013) concludes that the existing evidence about the effectiveness of R&D tax incentives, although it is mixed, seems to tilt towards the conclusion that they are not terribly effective in stimulating more R&D than the amount of tax revenues foregone. Tax price elasticity is higher for incremental than for level based R&D schemes, suggesting that its power as a policy instrument lies more in stimulating new R&D projects rather than supporting existing ones.
R&D tax credits are also less effective because they are biased in favour of large, established R&D firms, even though small firms are often given higher rates of credit. For small and new firms, or those active in R&D for the first-time, lacking information and experience, the high fixed costs means they do not bother to apply. This is particularly unfortunate, as tax elasticity is higher for small firms.
Subsidies for private R&D
Most econometric research evaluating R&D subsidies focuses on the effects on private R&D spending: whether public R&D spending is “additional”, “crowding in” private R&D spending, or whether it substitutes and tends to “crowd out” private funding. This focus on the private additive aspect, however, neglects that of the social value, which should be the ultimate rationale of public intervention.
In a survey of the literature, David, Hall & Toole (2000) conclude that "the findings overall are ambivalent", although on average there is more evidence in favour of positive effects on private R&D. But most of the early literature is plagued with the problem of identifying the true causal effect of public funding. A major issue is selection bias: are positive effects associated with R&D subsidies because better firms are selected for subsidies, rather than subsidies causing better performance? The difficulty is in identifying what would have happened in the absence of support. More recent studies have come up with better methodologies to construct these counterfactuals. Although the conclusions are still ambivalent, positive effects still seem to prevail more often.
An interesting approach is used in SIMPATIC (2013), an EC 7th Framework Programme (EU-FP7) funded European network bringing micro- and macro-econometricians together to evaluate RTD policies. Its modelling approach estimates the costs for firms applying for a subsidy and the social benefits derived from a given project. The model and the estimated parameters can then be used to compare the costs and benefits of the existing policy with alternative scenarios. SIMPATIC does this on a comparable cross-country basis (currently for five countries: Belgium (Flanders), Finland, Germany, the Netherlands and Spain) so that country-specific differences in effectiveness of R&D policy can be looked at.
The first important finding from SIMPATIC, common to all five countries, is that a very small fraction of firms apply for subsidies. This is true even for the R&D-active firms. This is the single biggest obstacle for R&D subsidies to have an effect on a potentially larger scale. So why do, or don't, firms apply for subsidies?
For the additionality effect, i.e. whether subsidies crowd in or out private R&D investment, the results show a substantial difference across countries. In the most robust specifications, there is a crowding-in effect for Belgium and Spain, but the opposite is true for the other countries.
The SIMPATIC approach can also identify social value, revealed by government agencies’ choice to fund which projects at what rate. Not surprisingly, the results are again heterogeneous across countries, but do confirm a social value (welfare) of funded projects for all countries, substantially (but not massively) above private value (profits). If we assume that government agency choices reflect true welfare differences, admittedly a rather unrealistic assumption, social returns (welfare) would, e.g. in Germany, be around 20% higher than private returns (profits).
The most interesting results from SIMPATIC (2013) are the comparisons of different policy scenarios. Compared to no government support, current policies lead to large gains in R&D for firms who would do R&D anyway, but next to none in terms of enticing new firms to start R&D. This is the major reason why there are only modest welfare gains: firms that already invest in R&D spend so much that returns from extra R&D are quite low. Pushing them to invest more requires substantial public support. Most of the potential for welfare gains are in enticing new firms to start investing in R&D, but this is precisely what current policies are not very good at.
4. The impact of R&D policies in applied macro models
Ultimately, any extra R&D and innovation from public funding needs to translate into economy-wide GDP growth and jobs. To assess this, we need to resort to macro models. Most applied macro models either have no explicit difference in treatment of investment in R&D compared to other capital investments, or they treat R&D as “given” and consider public R&D policies as “technology” shocks (eg Worldscan). More recent models describe the incentives for firms to invest in R&D. We now look at two of these macro-models presently in use by the European Commission.
The NEMESIS model has been used by the European Commission’s DG Research and Innovation to provide an ex-ante assessment of the impact of the EU-FP7 2013 budget of € 8 billion on GDP and employment. The model starts by using an average leverage effect of 0.74 euro of extra private R&D from any public euro spent, leading to 13.9 billion euros extra R&D from the initial public investment. From this, NEMESIS estimates a total cumulative extra GDP of 86 billion euros after 20 years, a multiplication factor of around 10 from the initial 8 billion of FP7 funds, along with an estimated extra 38,000 jobs each year in the EU, after a 15-year period. But these positive effects require time. Initially, the stimulus is absorbed in higher wages for researchers and there is job destruction from increased productivity. Only in the longer term, at least five years, is the growth capability of the additional private investments in R&D leveraged into positive growth and job effects.
The European Commission’s DG ECFIN uses another applied macro model, QUEST-III. Simulations using this model show that subsidies for R&D give a permanent increase in GDP levels but not in the GDP growth rate (Roeger et al 2008). Like NEMESIS, the positive effects from public R&D instruments only play out in the long term, with negative effects initially: reallocation of highly-skilled employees from production to R&D and job losses associated with improved productivity. Major obstacles for trnsforming R&D into growth and jobs are the fixed stocks of skills as well as entry barriers, and market power in the intermediate and final goods sectors.
Beyond the simulation of the overall effects on GDP and jobs using QUEST III, the results help to identify structural reforms which can enhance RTD policy effectiveness. The results confirm the complementarity of RTD policies with particularly product market reforms, and labour market and education reforms. More competitive product markets and a policy to increase the base of highly-skilled workers would increase the efficiency of any R&D policy instrument.
Roeger et al (2013) used the QUEST III model to analyse the effects of various structural reforms in Southern European countries (Italy, Spain, Portugal and Greece). R&D tax credits have positive effects on GDP in the long run, but only minor, ranging from 1.4% for Greece to 0.1% for Spain. In comparison, the structural reforms that yield the most significant results in the long run are education policies which decrease the proportion of low-skilled workers. This gives an increase in employment: 11% for Italy, 10% for Spain. But the highest economic gains are realised from product market reforms. These reforms lead to significant economic gains in the long-term, as for instance a 16% increase in GDP for Spain.
5. Wrapping up the insights from micro and macro-evidence for the R&D policy agenda
There is positive evidence that public R&D can be part of smart fiscal consolidation, but with caveats.
A first important issue to deal with is the lack of empirical evidence on the (relative) effectiveness of different policies, based on sound evaluation studies with proper counterfactuals. Particularly lacking are studies with a (quasi-) experimental design to nail down the causality effect of public funding. This holds not only for the public programs which support private R&D projects, but also those for R&D at universities and public research organisations. It is difficult to draw general conclusions from individual studies, as the effects of a particular program depend on the specific program design and management. And even for similarly designed programs, the SIMPATIC evidence shows substantial heterogeneity in effects across countries. As we still do not understand what causes this heterogeneity, transferring best practices from one member state to another should be handled with great care at this stage.
Nevertheless, the current evidence suggests that, by and large, R&D grants and tax credits have scope for positive effects, but only if they are targeted towards firms that are impeded to develop R&D projects where social rates of return substantially exceed private rates of return. That is an important challenge for policy to identify and select those projects of higher social rates of return. Potential candidates are basic research efforts and collaboration between industry and science.
The evidence from SIMPATIC seems to indicate the government’s selection closely follows the market selection, looking at the private rates of return. Public funding mostly goes to firms which are already spending on R&D, inducing them to spend more. This is costly as these firms are more likely to be in the area where marginal returns to R&D are diminishing and so need to be compensated substantially to raise their R&D investments further. High public budget costs are a relevant issue for policy making, particularly in times of fiscal consolidation. A more promising target for public R&D programs would be to entice new firms to engage in innovative projects: the evidence indicates that this group is not being effectively reached in current, standard public R&D programs.
Another important insight from the evidence is the low number of private firms applying for government R&D programs. This may seriously impede the effectiveness of these programs, particularly if those firms with the projects that have the highest social rates of return are not applying. The low rate may be due to the high application costs for firms. In order to attract more, public R&D programs should keep the application procedures as clear and transparent as possible.
Campaigns to attract more applications should be targeted particularly towards those firms that are likely to produce the type of R&D the government wants to subsidize, i.e. with the highest social rates of return. But firms may not be applying because they lack attractive innovative projects that generate sufficiently high private rates of return even when subsidized. Perhaps the strongest policy to get more applications and improve the effectiveness of public R&D programs is to increase the private rates of return from R&D investments. This calls for complementary policies addressing the framework conditions for innovation.
Complementary framework conditions needed for higher private and social rates of return from innovation can be identified using macro-models. Unfortunately, few of these applied in policy evaluation explicitly model the R&D growth process. Those that do indicate that only a long-term view shows how the positive effects of public R&D support fully play out on GDP growth and jobs.
Macro-models are as yet underexploited in assessing which framework conditions need to be in place to improve the impact of public R&D funding instruments such as grants and tax credits. This is also where they would be a very useful R&D policy instrument. The interaction with product market reforms, improving competition, labour and education reforms, and improving the stock of skills, seem to be the main structural reforms which could improve the impact of R&D policies, particularly in Southern Europe.
So, for the question on whether public R&D can serve in smart fiscal consolidation strategies, the answer can only be a timid yes at this stage. Public R&D certainly has the potential, but we know too little of its actual effects. More micro and macro-evaluations are needed.
Reinhilde Veugelers, full professor at KULeuven, the Department of Management, Strategy and Innovation (MSI), Senior Fellow at Bruegel
Aghion, P. (2006): “A primer on innovation and growth”, Bruegel Policy Brief, issue 2006/06.
David, P.A., B.H. Hall and A.A. Toole (2000): “Is public R&D a complement or substitute for private R&D? A review of the econometric evidence”, Research Policy, 29(4).
Dechezleprêtre, A., R. Martin and M. Mohnen (2014): “Clean innovation and growth”, SIMPATIC WP24 available at www.simpatic.eu
Griliches, Z. (1958): “Research cost and social returns: hybrid corn and related innovations”, Journal of Political Economy, 66.
Hall, B.H., and J. Mairesse (2009): “Measuring the returns to R&D”, in B.H. Hall and N. Rosenberg (eds.): Handbook of the economics of innovation, vol. 2, Elsevier: North Holland.
Mansfield, E., J. Rapoport, A. Romeo, S. Wagner and G. Beardsley (1977): “Social and private rates of return from industrial innovations”, Quarterly Journal of Economics, 77.
Mohnen, P. (2013): “R&D tax incentives”, i4g Policy Brief 2013 no. 25, European Union.
Roeger, W., J. Varga and J. in ’t Veld (2008): “Structural reforms in the EU: A simulation-based analysis using the QUEST model with endogenous growth”, European Commission, DG ECFIN, Economic Papers 351.
SIMPATIC (2013): R&D subsidy applications. E-book on i) firm application behaviour; ii) R&D agency decision making; iii) structural modelling, SIMPATIC e-book available at www.simpatic.eu
Varga, J., W. Roeger and J. in ’t Veld (2013): “Growth effects of structural reforms in Southern Europe”, EC-ECFIN Economic Papers 511.
Veugelers, R. (2015): “Matching research and innovation policies in EU countries”, Bruegel Working Paper 2015/16.
Veugelers, R. (2016): Getting the most from public R&D spending in times of budgetary austerity: some insights from SIMPATIC analysis, Bruegel Working Paper 2016/01.