Governments spend tens of billions of euros each year encouraging private research and development (R&D), mainly through R&D subsidies and tax credits. This public support for private R&D is traditionally justified by two arguments (both tracing back at least to Arrow 1962). First, firms may underinvest in R&D because they fail to appropriate its returns fully. Second, R&D is hard to finance externally, as it relies heavily on intangible and non-collateralisable assets. A third, increasingly important policy objective is to encourage firms that do not yet perform R&D to start doing so. 

Both tax credits and subsidies aim at boosting innovation by reducing firm’s marginal cost of R&D. Perhaps for this similarity, established economic models seldom differentiate between tax credits and subsidies. Often government-paid marginal cost reductions are called R&D subsidies even if they resemble R&D tax credits more. In practice, the two policies operate in profoundly different ways. Subsidies are discretionary: firms apply for support for specific projects, and a government agency decides which projects to fund and by how much. Subsidy policy thus allows for targeting support according to government preferences but involves an uncertain application process which is costly both for the firms and the government. Tax credits, in principle, apply a uniform rule to all eligible R&D spending. This rule-based policy is often simpler and more predictable but, as a result, it cannot distinguish between high- and low-social return projects.

Although there is substantial evidence that subsidies and tax credits increase R&D spending in recipient firms (e.g. Einiö 2014, Van Reenen and Nguyen 2016, Bloom et al. 2019, Galindo-Rueda et al. 2020), little is known whether these policies are worth taxpayers’ money. Our recent research (Takalo et al. 2026) suggests that application costs and the uncertainty associated with subsidies limit the reach and thereby their effectiveness. Tax credits, by contrast, are blunt but broad, so they reach more firms and generate higher overall gains, but only if their fiscal cost is not limited by monetary caps on the claimed amount. 

A framework for evaluating R&D support policies 

To compare these R&D support policies, we develop and estimate a model that captures three key features of R&D: societal impacts, financial frictions, and fixed costs of starting R&D. The model also explicitly incorporates the subsidy application and selection process. We estimate the model using detailed Finnish data on firm-level R&D projects and subsidy decisions 2000-2008. The data reveal substantial variation in support across projects, as illustrated in Figure 1, suggesting that policymakers attempt to tailor support to project characteristics.

Figure 1 Project-level R&D subsidy rates awarded in Finland

Note: The horizontal axis shows the percentage of an applicant’s R&D project costs subsidized by the government agency and the vertical axis shows the fraction of applications for each subsidy rate decision. 
Source: Takalo et al. (2026).

As shown in Figure 1, the Finnish agency awarding subsidies reject some applications (they receive zero support), accepts some but grants little support, while subsidising some R&D projects heavily (the maximum subsidy rate allowed by the agency rules is 70%). At the same time, only a fraction of firms (18% in our data) applies for subsidies, highlighting the limited reach of the policy.

Subsidies versus tax credits

In theory, subsidies have an advantage: they can be targeted at the R&D projects with the highest social returns. In practice, this advantage is undermined by the application and selection process. Applying for subsidies is costly and uncertain. Firms must prepare detailed proposals without knowing whether they will be funded or at what rate. The government agency awarding subsidies scrutinises proposals carefully before deciding on support but is nonetheless prone to errors; the agency does not necessarily always prioritise high social return projects and may encounter challenges identifying such projects. These costs and uncertainty discourage participation, especially among firms with less experience with public funding. As a result, the set of applicants and supported projects is imperfectly aligned with the set of socially desirable projects. The policy therefore fails to reach all the projects it should and sometimes supports the wrong ones. Crucially, the application costs of firms are unlikely to be internalised by the agency. For these reasons, we find that subsidies slightly reduce welfare relative to a situation with no policy at all, despite that subsidies substantially increase R&D in recipient firms.

In contrast, tax credits cannot be targeted to specific projects but, because they are more automatic and predictable, they can reach a much larger share of firms. We find that tax credits increase welfare by about 0.6%. They significantly boost R&D investment among firms that already conduct R&D and bring some new firms into R&D. Although the increase in participation is modest, it is substantially larger than with subsidies. Tax credits are more expensive for the government, reflecting their broader reach. Nonetheless, attempts to limit their fiscal cost, for example by making claiming tax credits difficult or capping total spending or the amount firms can claim, largely eliminate their welfare benefits according to our analysis. Making claiming tax credits difficult eliminates their key advantage of broader reach, whereas capping the maximum amount makes – whenever the cap is binding – a tax credit a pure transfer from taxpayers to recipient firms with no incentive effects. 

Implications for innovation policy and research

Our results suggests that R&D tax credits can be made more effective by ensuring sufficient uptake and removing firm-specific caps, whereas R&D subsidies can be made more effective by internalising the application costs and reducing its costs and uncertainty to firms. In practice, such application cost internalisation would likely to mean a more-rule based, transparent subsidy policy. More broadly, an optimal subsidy policy should employ mechanism design, as suggested by Neeman et al. (2017).

In academic research, the differences between R&D subsidies and tax credits should get more attention. R&D subsidies are based on a discretionary examination system, whereas R&D tax credits are based on a rule-based registration system. More research looking at the potential complementarities of these policies is badly needed. Economic growth models (e.g. Acemoǧlu et al. 2018) sometimes associate R&D subsidies with a uniform marginal cost reduction applying to all firms. This practice is not only misleading but can hugely overestimate the growth impacts of R&D subsidies. Overall, our results support the view that R&D tax breaks and subsidies hardly significantly boost productivity growth at national level (cf. Straathof and Gaillard 2015). Rather, they complement other innovation policy tools such as investments in education (cf. Akcigit et al. 2025). Furthermore, because the benefits of R&D do not remain within national jurisdictions, the R&D support policies could be much more effective if coordinated at the EU level.

References

Acemoǧlu, D, U Akcigit, H Alp, N Bloom, and W Kerr (2018), “Innovation, Reallocation, and Growth”, American Economic Review 108: 3450–91.

Akcigit, U, J Pearce, and M Prato (2025), “Tapping into Talent: Coupling Education and Innovation Policies for Economic Growth”, Review of Economic Studies 92: 696–736.

Arrow, K J (1962), “Economic Welfare and the Allocation of Resources for Invention”, in R R Nelson (ed.), The Rate and Direction of Inventive Activity: Economic and Social Factors, Princeton University Press.

Bloom, N, J Van Reenen, and H Williams (2019), “A Toolkit of Policies to Promote Innovation”, Journal of Economic Perspectives 33: 163–84.

Einiö, E (2014), “R&D Subsidies and Company Performance: Evidence from Geographic Variation in Government Funding Based on the ERDF Population-Density Rule”, The Review of Economics and Statistics 96: 710–728.

Galindo-Rueda, F, M Bajgar, C Chriscuolo, and S Appelt (2020), “Effectiveness of R&D Tax Incentives in OECD Economies”, VoxEU.org, 14. October.  

Straathof, B and E Gaillard (2015), “Will R&D Tax Incentives Get Europe Growing Again?”, VoxEU.org, 20 January.

Neeman, Z, M Schankerman, and S Lach (2017), “Government Financing of R&D: A Mechanism Design Approach”, VoxEU.org, 3. November. 

Takalo, T, O Toivanen, and T Wahlberg (2026), “Welfare Effects of R&D Support Policies”, forthcoming in the RAND Journal of Economics. 

Van Reenen, J and K-T Nguyen (2016), “Credit Where (R&D Tax) Credit’s Due”, VoxEU.org, 21 March.