Innovation is usually defined as the ‘invention, diffusion and exploitation of new ideas’. The innovation process is an important influence on long term economic development, and investment in research, development and new ideas is central to this. In particular, economists identify two key linkages from R&D to wider growth:
- First, firms conduct R&D to find ways to cut costs; to develop smarter ways of working; and to develop new goods and services. Those product and process innovations may, in turn, feed through to higher productivity, higher sales and profits for the firm. In turn, this helps recoup at least some of the cost of the original investment.
- Second, innovation in one firm may also spill over and benefit other individuals, firms or organisations. This means that the wider gains from R&D to society, which economists refer to as the ‘social returns’, may be greater than firms’ private returns.
These knowledge spillovers occur because new ideas permeate outside the firm: as key staff take new jobs, or set up new companies; through imitation and reverse engineering by competitors; and because forms of intellectual property protection, like patents and trademarks, don’t offer complete and permanent coverage. This wider diffusion process is often disruptive, as in Schumpeter’s notion of ‘creative destruction’.
The available evidence suggests that returns to private R&D are positive in most countries, and typically higher than regular capital investment. A 2010 survey suggests returns to R&D of 20-30% in more developed countries during the second half of the last century. Social returns are harder to estimate, but may be higher still: typically over 30% and in some cases even over 100% for studies over the same time period.
Public R&D activity is an important element in this mix. For example, a recent study of 15 OECD countries between 1980 and 1998 suggests that firms’ response to public R&D spending is higher than for private sector spending. In line with this, a 2013 study for the UK suggests substantial spillovers from academic research to private firms, while private sector R&D is almost wholly captured by the original investors.
These numbers help explain why national governments directly and indirectly support R&D, as part of a broader portfolio of innovation policies. If the firm that makes the R&D investment bears the cost, but others across the economy benefit from the new knowledge, then society would invest far too little in new knowledge if R&D activity was left only to the market. What is more, research at the knowledge frontier has highly uncertain payoffs and often requires expensive investment by firms, for example in specialist staff and equipment: these factors may also lead to sub-optimal levels of R&D. Some R&D activities may also exhibit ‘network spillovers’ due to their cost and complexity, which create further disincentives for firms.
The spillover argument implies that governments should support investment in R&D – for example by funding R&D directly or by complementing private sector activities through subsidies or making parallel public investments. For example, government can influence R&D activity by doing its own research; by encouraging collaboration between organisations, by funding universities and public research labs; or by funding private sector research through grants, loans and contracts.
At the most basic level, innovation is not a linear process. Pathways from R&D funding to innovation can be iterative and unpredictable. More broadly, firms’ and public sector opportunities may be shaped by previous decisions and trends (processes known as ‘path-dependence’). This can make identifying causal effects of interventions extremely difficult.
This has important implications for our evidence reviews, given their focus on impact evaluation. Preliminary sifts of the literature identified two areas for which there existed a sufficient number of impact evaluations to undertake a systematic review: R&D grants, subsidies and loans, including collaboration/networking interventions associated with these policies, and tax credits and other fiscal incentives.
Grants, loans and subsidies
What can we expect R&D grants and loans to achieve? There are multiple potential impact channels which may interact with each other. R&D support to firms should translate into ‘innovation outcomes’ like patenting, trademarks and new products/processes. In turn, that may feed through to higher productivity, higher sales/profits and increased employment in the investing firms – assuming they are able to effectively commercialise the knowledge. Knowledge spillovers should diffuse these benefits more broadly across the economy in a range of ways. These spillovers may, however, reduce the ability of individual firms to benefit from new R&D in terms of higher sales and profits (and related employment growth).
R&D spending in universities or public research labs can also have impact through multiple pathways: new knowledge and its applications; training and upskilling researchers; networks between researchers and firms; contract research and the generation of new spinout and startup firms. These wider economic outcomes are clearly harder to attribute to the original policy, making it easier (although not easy) to track effects for programmes that fund private firms/partnerships directly, compared with programmes that fund public science.
There are also crucial aspects of these interventions which further complicate evaluation. In particular, identifying the additional effect of programmes is challenging. For example, public R&D spending might crowd out investments that private firms would have made anyway. This is a big issue in areas like venture capital, where a market typically exists but government may wish to grow it further.
In addition, because R&D grants programmes for firms are often open to all, we might worry that the best (or worst) performing businesses might ‘select into’ the programme, so that participants are not representative of target businesses as a whole. This can lead evaluations to over (or under) estimates of the true effect of the intervention. The impact of grants and subsidies is also conditional on firms’ ‘absorptive capacity’ – for instance, the presence of qualified staff, suitable equipment, connections to experts or previous organisational experience. A recent review suggests that direct R&D support may have more impact when delivered in tandem with business advice or other support. This means that the impact of support may be quite heterogeneous across different types of firms. We will discuss these issues further, below.
More broadly, neither the private sector nor policymakers can predict exactly which experiments and new ideas will succeed; so public policies need to be able to identify promising areas of support without the ability to pick individual winners. At the same time, policies have to engage with industry – to ensure the programme reaches those who need it most – without being captured by vested interests. This means that governance, rules and processes may be just as important as policy content. As a result, policies that look similar (i.e. ‘give out grants’) may differ substantially in their design in ways that matter for impacts. Unfortunately, our ability to say much about these design elements is limited by the evidence available.
A final issue is the scale of policy effects. Knowledge can easily spill over local boundaries, benefiting firms across the economy. This may be good for national welfare, but will lessen the direct impact on local economic growth. This might still result in a net benefit for places implementing the policy, but spillovers would need to be taken into account in evaluating impacts. Such spillovers are one important reason why R&D grants and subsidies are often devised by national government; even if some aspects of delivery take place locally.
R&D Tax Credits
Government may also support R&D indirectly through tax credits or other incentives. For example, the UK has operated an R&D tax credits scheme for small and medium-sized enterprises (SMEs) since 2000, with an extension to larger firms in 2002. The programme was made substantially more generous for SMEs in 2008; one of our studies evaluates the impact of these changes.
Part of the appeal of tax credits is practical: the delivery apparatus already exists through the business tax system. This makes them relatively easy to target on certain types of firm (such as SMEs) or to make them accessible to all firms. (That said, the detailed design of tax credit schemes can be extremely complex.)
Tax credits also fit with some policymakers’ desire for a market-led, ‘hands-off’ approaches to innovation strategy: while grant programmes involve policymakers or experts selecting what they hope are the ‘best’ proposals, tax credits can reach a much larger number of businesses and avoid any suggestion of ‘picking winners’.
Understanding whether R&D tax credits are effective should also be of interest to local and regional policymakers. Most tax credit programmes tend to be designed by national governments, but not all: in more fiscally devolved countries than the UK, there are also regional-level fiscal incentives for innovation. Just looking at the biotech industry, for example, at least eleven US states have their own fiscal incentives.
Even in the absence of localised schemes, the fact that innovative activity is uneven and tends to cluster means that if R&D tax credits are effective, they are likely to have a local, as well as national, impact. Although knowledge spillovers are often physically bounded, information can also spill over local boundaries, benefiting firms across the economy. This may be good for national welfare, but will lessen the direct impact on local economic growth in a given area. For example, such spillovers are one important reason why R&D grants and subsidies are often devised by national government; even if some aspects of delivery take place locally. Getting a sense of the likelihood, importance and scale of these policy impacts is therefore very important for those interested in local economic growth.
As this short introduction makes clear, innovation policy can involve a wide range of very different interventions. As we discuss below, it is also an area in which comprehensive evaluation is challenging.
At the most basic level, innovation is not a linear process. Pathways from R&D tax credits to innovation can be iterative and unpredictable. More broadly, firms and public sector opportunities may be shaped by previous decisions and trends (processes known as ‘path-dependence’). In turn, this can make identifying causal effects of interventions extremely difficult. In addition, although formal R&D is an important element of innovative activity, only a minority of firms are R&D-intensive: there are many other forms of innovation that governments can seek to support.
This has important implications for our evidence reviews, given their focus on impact evaluation. Preliminary sifts of the literature identified two areas for which there existed a sufficient number of impact evaluations to undertake a systematic review: 1) R&D grants, subsidies and loans, including collaboration/networking interventions associated with these policies and 2) tax credits and other fiscal incentives. We also found some evaluation evidence on 3) public venture capital policies and 4) collaboration / networking initiatives, although neither of these is large enough to merit a full review.
What can we expect tax credits to achieve? As with R&D as a whole, there are multiple impact channels which interact with, and feedback on, each other. Tax credits should reduce the cost of research. More R&D should translate into ‘innovation outcomes’ like increased patenting, trademarks and new products or processes. In turn, that may feed through to higher productivity, higher sales/profits and increased employment in the investing firms – assuming they are able to effectively commercialise the knowledge. Spillovers should then help feed these benefits across the wider economy. These spillovers may, however, reduce the ability of individual firms to benefit from new R&D in terms of higher sales or profits (and related employment growth).
However, to initiate these effects, the tax credit has to offer a big enough cut in R&D costs for at least some firms to respond, something that is not easy to determine beforehand. And unlike R&D grants, which are directed at specific activities that administrators deem have a high social return, firms will use tax credits to fund R&D projects with the highest return to that firm – which might not be the activities of most benefit to society.
There are also crucial aspects of these interventions which further complicate evaluation. In particular, R&D tax credits might also run the risk of crowding out private investments that firms would have made anyway, or of distorting efficient investment allocations. This is an issue that especially relates to larger firms as they, amongst other things, face lower adjustment costs and, therefore, have a higher responsiveness to tax changes.
In addition, governments tend to deploy a number of innovation policies at the same time. For example, a number of tax relief schemes reviewed in this report are offered at the same time of R&D grants and subsidies. There are also overlaps with other policy agendas, notably business support and industrial policy.
This complexity makes it harder to identify the causal impact of single programmes. Compared to (say) public science programmes, which operate in a diffuse way, it should be simpler to trace the impacts of tax credits because they target firms directly. However, researchers still need access to high quality firm-level data, and need to be able to track firms through time. As we shall see, relatively few of our evaluations are able to do this.
Figuring out the additional effect of R&D tax credits is particularly tough. Because tax credit programmes often require a qualifying level of existing R&D activity, it is possible that qualifying firms might have made further investments without the programme, or that non-qualifiers might have benefited more. Without a counterfactual, we will over or under-estimate the true programme impact. The knock-on effect of cheaper R&D on firms’ innovative activity and economic performance is also conditional on ‘absorptive capacity’ – for instance, the presence of qualified staff, suitable equipment, connections to experts or previous organisational experience. Again, we need to find ways to control for these hard-to-observe factors when evaluating impact.
Evaluating the impacts of R&D investment (both as loans and as tax credits) is extremely complex, even if the policies themselves seem relatively simple. The likely economic outcomes are hard to predict, hard to measure and evaluate, and may differ substantially at local and national level. This is reflected in our two reviews. We find a number of impact evaluations that meet our minimum quality thresholds, but very few that can precisely identify the full range of policy effects (and none that can attribute this to specific aspects of programme design).
Definition of Innovation
By ‘R&D’, we mean investigative activity undertaken by the private sector (with or without academic participation), which has the objective of improving existing, or developing new, products or processes. Governments carefully define the scope of R&D inputs. Programmes aimed at commercialising R&D aim to assist the generation, diffusion and exploitation of these products and processes.
In our review of R&D grants, loans and subsidies, we included in our definition programmes that provided financial assistance for the purposes of R&D and to support R&D commercialisation and growth, where growth includes: increased private R&D expenditure; growth in number of patents and growth in productivity. Shortlisted programmes include:
- Innovation-policy schemes providing public funding for innovation projects
- National funds for research in science and technology
- Subsidised government loans for R&D activities
- Regional subsidies to support public and private R&D activities.
In our second evidence review, we looked at the effectiveness of R&D tax credits. By construction, the primary goal of tax credits is to increase R&D spending by reducing its after tax-costs and thereby influencing wider economic outcomes as well. In general, there are two main schemes for the roll out of tax credits:
- incremental-based, where a firm’s eligibility for the credit depends on current R&D spending that exceed historic figures;
- volume-based, where the credit is only based on the current volume of R&D expenditures.