VoxEU Column Industrial organisation Productivity and Innovation

Three guiding principles for R&D policies in the digital era

Digitalisation is transforming human life – from the way we interact with each other to the way we work, relax, and create. R&D within companies is no exception. This column lays out pathways for policymakers to successfully adapt R&D policies to these changes based on three guiding principles: direct policies towards spillovers, make policies technology-neutral, and do not favour superstars over challengers.

Digitalisation is transforming human life – from the way we interact with each other to the way we work, relax, and create. R&D within companies is no exception. Across industries, digitalisation is altering the way R&D is conducted as well as the areas of research. In medicine, for instance, artificial intelligence promises to accelerate the quest for better and cheaper drugs (Fleming 2018), while automakers and tech firms invest billions in the development of self-driving cars. 

Three guiding principles

Most governments support private R&D because of the positive spillover effects on the rest of the economy. As R&D becomes digital, these government policies require adaptation to remain effective. Our discussion here draws on Bijlsma et al. (2018). In a recent OECD Policy Paper, Guellec and Paunov (2018) also signal the digital transformation of R&D and likewise stress the need for policy reform. 

To ensure that R&D policies are effective in the digital era, we suggest an update of R&D policies following three guiding principles:

1. Direct policies towards spillovers

As in the analogue world, the digital era requires policies to resolve market failures (Tirole 2016). In the context of R&D, knowledge spillovers are the main source of market failures. Good R&D policy helps firms to internalise these spillovers (through tax incentives, for example) and, as long as it enhances welfare, promotes them (for example, via banning non-compete agreements or requiring open access publishing, for example). This is not new. The novel insight is that digitalisation changes the nature of these spillovers – and thus requires a policy reconsideration. 

How does digitalisation change spillovers, or how they can be internalised? 

First, digitalisation disproportionately promotes two non-embodied spillovers: open source software and data. Spillovers emerge when firms reuse software code or data to create new products or knowledge. Tracking tools and reductions in the cost of data storage foster the collection and processing of data. Sharing platforms are instrumental in disseminating these data and act as ecosystems for computer code developers. Governments can play a leading role in making data more available and thereby lower the barrier for reuse in data-driven R&D.  

Second, digitalisation allows for more external collaboration, outsourcing, and more flexible (‘agile’) R&D. This means that restricting tax incentives to in-house R&D and requiring ex ante approval reduces the scope for policy to internalise spillovers. 

Third, digitalisation enlarges the labour market for digital talent. Talented programmers can easily move to another firm in another industry or another country if their rewards fall short. The geographical and industry range of knowledge spillovers thus increases as a result of digitalisation. Restrictions to labour mobility between firms, for example through the enforcement of non-compete clauses, also restrict knowledge spillovers between firms and lower the scope for a thriving ecosystem of innovative firms. In fact, Bessen (2015) points out that the open labour market has been one of the key factors explaining the rise of Silicon Valley. In the digital economy non-compete clauses thus seem to entail higher social costs.

2. Make policies technology-neutral 

When a policy’s aim is to foster R&D in general, instead of addressing a specific societal challenge, the policy should be general as well and not geared to a subset of possible technologies. This calls for technologically neutral innovation policy (Tirole 2016). Market players are often better informed when it comes to the success rate of R&D investments, and policy should determine the direction of R&D as little as possible. In theory, most developed countries seem to follow this principle, as they subscribe to the OECDs Frascati Manual definition of R&D. In practice, however, many countries struggle with the eligibility of big data research, software development, or prototypes, and R&D tax incentives differ across countries (Uhlíř et al. 2017). For example, as software develops rapidly, it is hard for governments to provide up-to-date eligibility guidelines for software R&D as example-based guidelines can quickly become obsolete. Moreover, the boundary between what constitutes a true R&D software problem and a mere application of existing technology is not always clear-cut. 

3. Do not favour superstars over challengers

Globalisation of sales via the internet, network externalities on two-sided platforms, and data-driven innovation are channels through which digitalisation increases economies of scale and are believed to give rise to ‘superstar firms’ (Autor et al. 2017). To ensure that new superstars do not enjoy a quiet life, policy can foster dynamic competition via vigilant antitrust enforcement and R&D policies that are accessible to smaller firms. Contrary to this principle, however, politicians sometimes seem to have a preference for national champions. In the context of innovation policy, patent boxes tend to be more attractive to ‘superstars’ than for challengers. 

This third principle calls for a neutral stance towards challengers and incumbents; it does not state that entry should be stimulated as much as possible. Rationales for favouring entrants (or SMEs) could be that those firms respond stronger to subsidies or that their R&D entails higher spillovers, but for both rationales the empirical evidence is mixed (e.g. CPB 2014).

The economic rationale for R&D policy is unchanged

These guiding principles reflect that the economic rationale for having R&D policies does not change due to digitalisation. Traditionally, several market failures leading to an underinvestment in private R&D have been put forward as a justification for government subsidies (see Arrow 1962 for an early discussion). Positive externalities (or knowledge spillovers) cause firms to not capture all the benefits of their R&D activities. Firms therefore invest too little compared to what is best for society as a whole. Recent research shows that, despite digitalisation, the magnitude of spillover effects has remained fairly stable (Lucking et al. 2018).

The impact of digitalisation on the type and dynamics of research 

Although digitalisation does not change the rationale for R&D policies, it does affect the nature and dynamics of R&D. Corporate research is increasingly focused on digital products and services, as illustrated by the dominance of tech firms in the corporate R&D landscape (see Figure 1). The focus on digital innovation is by no means limited to the largest tech companies, however, with automotive companies doubling down their efforts on autonomous and electric vehicles,1 for example, and oil and gas majors recognising that new promising research directions are enabled by digitalisation.2

Figure 1 Information technology companies have taken over pharmaceutical and automotive companies in terms of R&D spend


Notes: “Information technology” companies incudes internet companies. Each bubble shows the yearly R&D investments (y-axis) versus yearly revenue (x-axis) for the companies that appear in the top 10 R&D investment companies in either 2012 or 2018. Bubble size indicates market valuation at the end of December 2012 (left plot) and December 2018 (right plot). Dashed line indicates 10% R&D investment ratio. Source: Strategy&. 

Moreover, digitalisation of R&D is accompanied by a more open innovation model strengthened by new spillover channels and a faster commercialisation of successful research. Digitalisation has made knowledge and tooling more accessible via open-access publishing and repositories, open-source software, online knowledge markets, and open courseware. Consider, for example, the development of open source packages such as TensorFlow or Prophet. These packages resulted from extensive research activities by Google and Facebook, respectively, yet the beneficiaries are data scientists around the globe working for other organizations.

Policy ideas for digital R&D

We hope our principles will guide governments to reap the benefits of changing R&D dynamics for their economies. With the guiding principles in mind, it becomes more straightforward to identify those areas that need to be reconsidered. 

First, the changes in R&D and its spillovers demand that governments review existing R&D-instruments. Presumably unintentionally, these instruments are tailored to traditional R&D-intensive firms, such as pharmaceutical or chemical firms, with large (in-house) research departments and plannable R&D. These instruments fit less well with tech-intensive firms and hamper R&D activities that span across institutions and continents. The patent box, which currently exists in many OECD countries, is often an instrument that disproportionally favours profitable ‘superstar’ firms when it reduces corporate tax on profits. In these cases, the instrument targets the most appropriable part of innovation. Additionally, patent boxes typically require firms to have intellectual property (such as a patent). As a consequence, the patent box directs firms to favour a ‘closed innovation’ model.  Second, the change in spillovers also requires governments to consider new instruments such as centralised data pools, mandatory data sharing, or support for open source software.

In addition, evolving R&D policies along the lines of our guiding principles can be a useful instrument to tackle broader economic challenges associated with digitalisation. The mounting evidence of increased market concentration (e.g. De Loecker et al. 2018) forms one of these challenges. This increased concentration calls for a review of competition policy in the digital age (e.g. Furman et al. 2019). Crémer et al. (2019) recently laid out how competition policy can be adapted to the digital economy.  They argue, among other things, that analyses of market power should also include an analysis of the difference in data accessibility between a dominant firm and competitors. The inclusion of data in measuring market power has implications for the assessment of mergers and acquisitions. For example, acquisition of startups with low turnover, but with a rapidly growing user base, might then be deemed anti-competitive. A second, potentially related challenge is the decrease in business dynamism – as indicated, for example, by the declining share of young firms in the economy. Akcigit and Ates (2019) have put forward a decline in knowledge diffusion as one of the potential explanations. By focusing on knowledge spillovers, R&D policies can help change course and contribute to more business dynamism and competition.  

Of course, there is no silver bullet for creating a digital-proof R&D policy. In the spirit of digital R&D, it will take continuous experimentation with these three guiding principles in mind to get IT right. 


Akcigit, U and S Ates (2019), “What Happened to U.S. Business Dynamism?”, NBER Working Paper No. 25756.

Arrow, K (1962), Economic welfare and the allocation of resources for invention, NBER.

Autor, D, D Dorn, L Katz, C Patterson and J Van Reenen (2017), “The Fall of the Labor Share and the Rise of Superstar Firms”, IZA Discussion Paper No. 10756.

Bessen J (2015), Learning by doing; The Real Connection between Innovation, Wages, and Wealth.

Bijlsma, M and B Overvest (2018), “Digitalisering R&D”, CPB Policy Brief 13.

CPB (2014), “A Study on R&D Tax Incentives”, EC Taxation Paper No. 52.

Crémer, J, Y-A de Montjoye and H Schweitzer (2019), “Competition policy for the digital era”, Directorate-General for Competition, European Commission.

De Loecker, J and J Eeckhout,2017, “The Rise of Market Power and the Macroeconomic Implications”, NBER working paper No. 23687.

Fleming, N (2018), “How artificial intelligence is changing drug discovery”, Nature 557: S55-S57.

Furman, J, D Coyle, A Fletcher, D McAuley and P Marsden (2019), Unlocking digital competition, Report of the Digital Competition Expert Panel.

Gonzales-Cabral, A, S Appelt and F Galinada-Rueda (2018), “OECD review of national R&D tax incentives and estimates of R&D tax subsidy rates 2017”, OECD

Guellec, D and C Paunov (2018), “Innovation policies in the digital age”, OECD Policy Papers 59.

Lucking, B, N Bloom and J Van Reenen (2018), “Have R&D Spillovers Changed?”, NBER Working Paper No. 24622. 

Tirole, J (2016), Economics for the Common Good, Princeton University Press.

Uhlíř, D, B Straathof and C Hambro (2017), “Administration and Monitoring of R&D Tax Incentives”, Mutual Learning Exercise, Directorate-General for Research & Innovation, European Commission


[1] See, for example,,

[2] See, for example,

2,835 Reads