Innovation, and the knowledge externalities that underpin it, are the primary source of economic growth and are at the heart of policy debates over how to promote sustained growth and competitiveness. Governments have adopted a variety of policy instruments to support innovation. Some countries rely on indirect fiscal incentives such as R&D tax credits, while others prefer direct instruments such as grants, loans and equity finance.
Empirical research finds pervasive knowledge spillovers. If an R&D project generates externalities this is not, in itself, a justification for government support. Some of those projects might be financed by private capital markets, so that government support would also be redundant. Since public funds are costly, government support that is redundant would be welfare-reducing. On the other hand, even if capital markets were perfect, there could be R&D projects that could not get market financing based on their private returns, but which would generate sufficient social returns to justify themselves. In this context, we can think of government finance as a way of purchasing the (expected) social returns from projects that would otherwise not be realised.
Government support programs typically vary along three main dimensions:
- Whether grants or loans are used and, if loans are used, whether repayment is conditional on the project’s success or not;
- The interest rate charged on loans;
- The co-financing requirements from the applicant for both grants and loans.
All the loan schemes that we examined for OECD countries had either a zero or nominal interest rate, and many of the grant and loan schemes required the applicant to co-finance the cost of the R&D project (rates varying from about 20% to 80%).
Recently some empirical studies have shown that existing R&D subsidy programmes generate some 'additional' innovation relative to the counterfactual of no support (e.g. Klette et al. 2000, Lach 2002, Takalo et al. 2013, 2017). To our knowledge, however, there is no research on how the design of policy to support R&D influences its effectiveness, or how loan programmes should be optimally designed to maximise welfare. Understanding this would be critical when formulating effective, cost-efficient support policies.
In our recent research, we apply the techniques of mechanism design to analyse the optimal structure of R&D loan programmes (Lach et al. 2017). We develop a model in which projects generate positive externalities, and the government has limited information about the projects. To focus on externalities, we abstract from capital market imperfections, though we do incorporate a constraint on the ability of an entrepreneur to self-finance. In addition, loans have to be repaid only if the projects succeed. This mitigates the impact of any capital market imperfections. In our model, entrepreneurs have R&D projects characterised by three features:
- A probability of success;
- Private returns (assumed, for simplicity, to be common to all projects, with this restriction relaxed later);
- An externality.
Both the success probability and social returns vary across projects. We assume that the success probability is known to the entrepreneur and the competitive capital market, but not to the government. Since the competitive market interest rate depends inversely on the project's probability of success (or entrepreneur type), the private cost of funds varies across projects but is not known to the government. We assume that the government knows (or has a signal on) the social returns, and has two instruments at its disposal: the interest rate on the loan, and a co-financing (matching funds) requirement.
We show that the welfare-maximising policy under the assumption that the government does not know the success probabilities of projects, but there is no moral hazard (exogenous success probabilities), achieves 'first-best' efficiency and involves selecting exactly the projects that are socially profitable but will not be financed by the capital market. The optimal policy is to set the interest rate at close as possible to the ex post rate of return of the most successful project that would not be supported by the private market, with a co-financing rate that approximates zero (we call this the ‘epsilon policy’).
The basic trade-off in the model is between making government funding 'additional', and inducing the implementation of all projects that generate positive expected social returns (netting out the social cost of public funds). Projects with high success probability (low interest rates in the private market) need to be screened out to ensure additionality, but projects with very low success probability also need to be excluded because the expected social returns do not justify undertaking them. In this sense, the optimal policy will need to 'target the middle'.
Using a high interest rate – in the limit, the ex post rate of return – ensures additionality, and a low co-financing requirement maximises the set of projects with sufficiently high social returns from which the government can select.
This result differs sharply from typical R&D loan programmes in the real world, which have significant co-financing requirements but zero or negative interest rates. The optimal policy is also very different from the most common pure grant schemes (equivalent to a loan scheme with an interest rate of -100%).
When we introduce moral hazard – that is, we allow the entrepreneur to exert costly effort to increase the probability of the project's success – the optimal policy now involves offering at most two contracts: the epsilon contract, and another with a lower interest rate r to provide incentives to the entrepreneur to undertake effort, together with a higher co-financing requirement b to reduce the amount of public funds committed by the government. We refer to this as the ‘(b, r) contract’. The optimal interest rate is decreasing, and the co-financing requirement is increasing, in the degree of moral hazard. Thus, as moral hazard concerns become more serious, the optimal policy moves in the direction of policies that are observed in practice.
From a theoretical perspective, the optimal policy is notable for two reasons:
- It describes a mechanism design problem for which the optimal solution is 'simple', in the sense that the optimal menu offers at most two alternatives (at most one for each level of induced effort). This is very unusual in mechanism design. This simplified mechanism arises because the optimal policy involves targeting the middle – the high types (high probability of success) will be funded by the private market and low types do not justify public financing because the expected social gains are negative. Because of the incentive compatibility constraint, offering another contract to the high type involves leaving more rent to the entrepreneur, which has no social payoff. Simulations show that, for reasonable parameter values, the optimal solution consists of just one alternative.
- The optimal solution, with at most two contracts, is robust to the introduction of two-dimensional uncertainty, where there is asymmetric information about both the project probability of success, and the private return when successful. This is also very unusual. We are not aware of any other example of this in mechanism design.
We use calibrated simulation analysis to illustrate how the optimal policy varies with the shadow price of government funds and the level of the externality, and to assess how alternative policies, commonly observed in practice, perform relative to the optimal design in terms of additionality and welfare. The simulations highlight several key features.
- First, the optimal policy depends critically on the cost of public funds, λ, and the externality generated by the project, σ. If λ is high, it is optimal to use only the epsilon contract. Otherwise, the optimal policy is to offer the (b, r) contract. This is because the epsilon contract produces no redundancy, and the (b, r) contract generates some redundancy which is socially costly when λ is high.
- Second, the (b, r) contract is optimal if σ is large. Even though this contract involves redundancy, it widens the set of projects covered and induce some partially implemented projects to switch to implementation with full effort. Moreover, the optimal interest rate in the (b, r) contract is increasing in λ and decreasing in σ.
- Third, we show that the optimal policy generates substantial welfare gains for a wide range of parameters. The welfare gains (relative to no intervention) if the (b, r) contract is optimal will vary from about 7% to 29%. The optimal policy will always increases the fraction of projects implemented (even if just the epsilon policy is used), and generate substantial additionality – roughly 20-30% more projects than are implemented under the optimal policy without support.
The simulations show that neither (maximising) additionality nor (minimising) redundancy is an appropriate criterion for evaluation of R&D support policies. Projects with less additionality or higher redundancy may generate greater welfare. We also show that both full grants and interest-free loans – which are typical of observed policies – perform much worse that the optimal policy, unless the cost of public funds is very low, or the externality is very large. Shifting from full grants to interest-free loans generate a substantial welfare gain, unless public funds are very cheap.
This research has two key policy implications for the design of R&D support policies. First, optimal policies should ‘target the middle’. Very low-risk projects would be funded by the market, so government support is redundant, and very high-risk projects, while not supported by the private market, is not likely to generate sufficient expected social payoff to justify support. This important message is not widely appreciated.
Also, because the optimal policy depends on the cost of public funds and the size of the project externality, no single design is likely to be appropriate for all settings. In particular, the optimal policy is likely to differ across technology areas, and between industrialised and emerging economies. For example, if emerging economies are characterised by higher cost of public funds (or lower externalities), the epsilon contract – which is also relatively easier to implement – should be favoured.
Klette, T J, J Moen and Z Griliches (2000), “Do Subsidies to Commercial R&D Reduce Market Failures? Microeconometric Evaluation Studies," Research Policy 29(4-5): 471-495.
Lach, S (2002), “Do R&D Subsidies Stimulate or Displace Private R&D? Evidence from Israel," Journal of Industrial Economics L(4): 369-391.
Lach, S, Z Neeman and M Schankerman (2017), “Government financing of research and development: A mechanism design approach”, CEPR Discussion Paper No. 21299.
Takalo, T, T Tanayama and O Toivanen (2013), “Estimating the Benefits of Targeted R&D Subsidies," Review of Economics and Statistics 95(1): 255-272.
Takalo, T, T Tanayama and O Toivanen (2017), “Welfare Effects of R&D Support Policies," CEPR Discussion Paper 12155.