We review the state of research and development (R&D) incentives, with the intent of comparing the diverse types of R&D incentives, from credits on spending or on revenue to grants and private-public partnerships in low-income countries. Untargeted incentives provide incentives for any firm that improves its products or processes, frequently in ways as simple as upgrading equipment. Because measuring direct impact of untargeted incentives is difficult, much of the literature on untargeted incentives has focused on whether they encourage private R&D spending, and our review finds a broad consensus that government expenditures do induce greater spending on R&D by private firms, but at rates that differ in consistent ways. Another thread of the literature seeks to link intellectual property (IP) outputs with untargeted R&D incentives, with mixed success, and we offer some explanations for the disconnect between R&D incentives and intellectual property productivity. Targeted incentives are There is exceptional diversity in targeted R&D incentives, which are available only for certain industries, firms or activities., and we provide several examples of how governments have targeted unique goals with unique incentives. We estimate about 20% of countries in the lower end of the development spectrum do not have untargeted incentives, which might be explained by that incentives that are carefully targeted to the goals of the country with measurable effects, are a more effective use of developing country resources.

1 Introduction

The heterogeneity across research and development (R&D) incentives is vast, and government supports of all types accounted for around 70% of R&D performed in OECD countries from 2000 to 2013 (OECD, 2016). The diversity is even expanding, as new forms such as patent boxes are gaining popularity. A country with a limited or without governments’ support to private R&D can draw lessons from the many existing implementations.

There are many papers that discuss one type of R&D incentive usually in high-income countries (e.g., Hall et al, 2010, David et al 2010, Correa at al. 2013). We offer a larger view across different types of R&D and their relative strengths and weaknesses, and to the extent possible across countries with different income level.

Even the definition of R&D varies, from narrow definitions that cover certain sciences to broad definitions that cover almost any industrial improvements. Every country tailors incentives to its own goals and situation, so it is no surprise that the term “R&D incentive” can refer to entirely different things in different locales. Section 2 goes over the various definitions codified in laws.

One can broadly divide R&D incentives into direct or targeted incentives, which are frequently in the form of grants aimed at certain industries or goals, and indirect or untargeted incentives, which are almost always in the form of tax credits granted to any applicant with qualifying R&D. There is no one best approach, either for a high-income or low-income country, but Section 3 goes over the various pros and cons of both, and offers some observations on the existing choices made by governments.

Section 4 reviews the literature on the efficacy of these incentives. There can be situations of an ‘R&D poverty trap’ (Galor and Weil, 2000, García-Rodríguez and Sánchez-Losada, 2014): in one equilibrium, R&D is too low-productivity to invest in, so no growth occurs; in another, R&D produces enough growth to be self-sustaining. In other words, if there is certain threshold under which firms would not invest in R&D, supporting those firms below this threshold could catalyze their R&D investment (Choi, 2017). Therefore, it is especially important to understand whether government R&D expenditures crowd out or increase private spending.

These incentives show outcomes more heterogeneous than indirect incentives, especially between lower-income and higher-income countries. Among higher-income countries, studies have found great variation in the effects of broad-based incentives (including general industrial development along with more narrowly-defined sciences), but have generally found conditions that more conducive to complementary private investment than conditions in lower-income countries.

Among indirect incentives, the consensus in the literature is that government R&D spending leads to some increase in private spending, but the degree of such effects varies greatly based on many considerations. Patent or Intellectual Property (IP) boxes are increasingly common, providing a lowered tax rate for revenue associated with patents or broader IP. There is little evidence that they have benefits benefit over an equivalent tax credit for R&D expenses in inducing private R&D expenditures.

In low-income countries, the focus is more on industrial development, and grants are typically more focused on specific projects. There is also great heterogeneity in results here: some projects show returns even in the thousands of percent, but many are simply failures.

Beyond general investment, there is the question of how these incentives affect the ‘inventiveness’ of recipients, usually measured by patent counts. This measure is common but problematic: there is little correlation between direct R&D spending by a government and patent counts in that country, and patent laws are constantly changing, sometimes drastically enough that any change in invention rates are dwarfed by policy shifts. There is some consensus that patents correlate to a company’s stock market returns, but it is an open question whether this indicates productivity gains or rent-seeking.

2 Definitions and Scope

This section considers the wide range of definitions of R&D. The breadth is so wide that some pairs of definitions refer to almost entirely different activities.

2.1Defining technology

The “technology” section in today’s newspaper focuses on new gadgets or scientific discoveries in biology or physics. We will refer to such topics, aimed primarily at new discoveries in certain fields, as narrowly-defined technology. Definitions in law, and therefore in practice, are frequently far more broad.

The OECD’s Oslo Manual gives these definitions (§24):

A technological product innovation is the implementation/commercialization of a product with improved performance characteristics…. A technological process innovation is the implementation/adoption of new or significantly improved production or delivery methods.

We refer to this conception of technology as broadly-defined technology. In the academic literature, one finds an even broader definition in Solow [1956], who defined technological improvement to be any change in the production function converting inputs to outputs, which might include better roads, a better-educated work force, or even better marketing.

2.2 R&D definition

The OECD’s Frascati manual provides standardized definitions of R&D in three components (§2.1(64)):

Basic research is experimental or theoretical work undertaken primarily to acquire new knowledge of the underlying foundation of phenomena and observable facts, without any particular application or use in view.

Applied research is also original investigation undertaken in order to acquire new knowledge. It is, however, directed primarily towards a specific practical aim or objective.

Experimental development is systematic work, drawing on existing knowledge gained from research and/or practical experience, which is directed to producing new materials, products or devices, to installing new processes, systems and services, or to improving substantially those already produced or installed.

The manual adds the key caveat that R&D must have “an appreciable element of novelty and the resolution of scientific and/or technological uncertainty.” (§2.3.1). Note that the definition is about the work done to generate some output, not the output itself, which may or may not be eligible for markers of novelty such as patents or journal publications.[1]

2.3 R&D definition in practice

Each country selects its own point on the narrow to broad spectrum. German deductions for researchers are especially narrow and focus only on basic research as defined by the Frascati manual, while Brazil, Canada, Japan, Malaysia, Russia, Spain, and the USA allow R&D to include product improvements and do not have a strong “resolution of uncertainty” requirement. Among others, the Austrian, Belgian, Czech, Croat, Dutch, and French R&D laws are based on the Frascati manual[2]. Some countries (including China, Ireland, Israel, and Poland) require that a firm register as focused on narrowly-defined technology, and then have license to claim broadly-defined tax credits. Many laws, such as the U.S. law, have an industrial application requirement, making basic research as defined in Frascati ineligible (26 US Code 41(d)). The Dominican Republic has an even more industrially-focused definition of R&D: “the process of improving the productivity of processes, products, and industrial services.”[3]

Thus, the R&D dentition varies across countries. In extreme example is that the R&D qualifying for German deductions has almost no overlap with R&D qualifying for U.S. credits.

This paper will make an effort to be clear about when the discussion is about narrow or broad regimes. But confounding of different types is common: it is difficult to find a statement about broadly-defined R&D by a politician that doesn’t confound it with narrowly-defined R&D. To give one example of many in the literature, Straathof et al (2014) motivate their study partly by the common justification for patents and trade secrets, to prevent one company from spending heavily on a new technique only to see another costlessly imitate (p 18). But the only way to imitate broadly-defined R&D such as implementing an ISO standard or upgrading software is by making the same outlay.

Scope This paper considers dedicated R&D funding in various forms. There are many other means of improving a country’s R&D position that can have a major effect, such as education and gender policies[4] and individual incentives such as scholarships and visiting fellow programs, but are a digression from the focus here. General industrial grants unrelated to R&D are also out of scope from this paper, but, as discussed below, the distinction can be sometimes difficult to make, and one sometimes finds general grants to certain industries in the guise of R&D grants. This paper limits the targeted support provided directly to the firms to improve its innovation input and ouput, but exclude the incentives to enhance R&D collaboration among firms and universities.

3 R&D incentives: Types and goals

This section reviews different types of R&D – targeted and untargeted, and their advantages or disadvantages. The optimal balance of targeted and non-targeted support for R&D varies by countries and objectives of such supports. This section shows the wide variation in the choice of incentives.

3.1 Targeted vs. untargeted

A targeted incentive is one which is provided to a specific firm, industry, project, technologies, or goal. Direct grants are used for diverse purposes, such as individual R&D projects within a firm, and can be allocated on an entitlement basis or a competitive basis.

An untargeted incentive is mostly administered via the tax system and is given to applicants for general R&D expenditure that meets often broad requirements. R&D tax incentives aim to reduce firms’ tax payment as firms invest on R&D. In recent years, some Governments have included the intellectual property income generated by R&D as tax incentives (the so-called “patent box”).

However, the division line between targeted and untargeted is not clear. For instance, some tax incentives can be very similar to targeted incentives when it is accompanied with high level of audit and pre-approval requirements for R&D projects (OECD, 2012). Also, some countries add special provisions to target specific firms to receive the tax incentives. For instance, France takes the ceiling on the credit allowed that leads the credit toward smaller firms. China provides differences in tax treatment that apply to R&D done abroad, domestic firms or R&D done in the country by foreign-owned firms (OECD 2016).

Figure 1: R&D tax incentives by countries’ income level

The horizontal axis of Figure xxx divides 131 countries across the full spectrum of development into deciles by GDP per capita. There is tendency that higher-income countries are more likely to have untargeted incentives than lower-income countries. The vertical axis shows the percentage of the count of countries in each decile that have an untargeted R&D credit. At the lowest decile, under 10% of countries have an untargeted R&D credit, while over 80% of the countries in the top deciles have one.[5]

However, this does not mean that higher-income countries’ R&D expenditures are primarily untargeted. Earlier predictions expected that countries will increasingly turn to the R&D tax incentives and away from direct grants (Hall and Van Reenen. 2000), but no clear trends in this prediction have materialized. To stress this point, figure 2 plots the levels of direct versus indirect R&D funding among a range of countries that have R&D tax credit, showing a diverse range of strategies.

Figure 2 plots direct and indirect funding for 41 countries in 2014 as measured by the OECD [OECD, 2017], showing wide variation in country strategies. The units of the plot are percent of GDP; for example, the OECD estimate of Ireland’s direct subsidies totaled almost 3% of GDP.

Figure 2: Direct versus indirect funding.

Note: The 45^◦line is shown for reference. Figure 1 used OECD data only, so most countries in that figure have an R&D tax credit.[6]

To give a detailed example, the United States leans heavily toward targeted R&D subsidies over untargeted support by providing almost 18 times larger amount for direct supports. For instance, in 2016, direct incentives through grant awards were $36.74 billion (covering 86,403 grants, including $29 billion awarded by the National Institutes of Health), while an estimate of the expenditure on the R&D tax credit of $2.06 billion.[7]

3.2 Why do governments choose targeted vs. untargeted?

Within a country, targeted incentives and tax incentives are substitutes (Montmartin and Herrera, 2015). The level of targeted and non-targeted support for R&D varies by countries and the objectives of such supports. For instance, targeted incentives are more suitable in countries with a low capacity for designing and administering tax policies. Moreover, targeted incentives might be a better tool to enhance young and small firms’ R&D while tax incentives might be better for attracting international R&D to locate jurisdictions.

The main benefits of targeting is that it allows governments to select those projects with the highest expected social returns and those industries which are countries strategic industrial policy priorities, while tax incentives may, in contrast, be difficult to target and may support projects with very low (or even negative) rates of social return (WB 2010).

Another benefit of targeting is that it makes the recipients to be accountable. For example, grants can be provided conditional on the accomplishment of targets. Moreover, targeted incentives can have longer-term demonstration effect. Obtaining the targeted support is a good signal of the firms’ performance, which can help the firms to access to finance even beyond the period of the incentives provide (Bravo-Biosca et al 2012, OECD 2016).

However, selecting the projects and firms with the highest returns provide a number of challenges. The major determinants of high growth firms and the successful ideas/projects are not well known (McKenzie. 2014, Kerr et al, 2014), implying identifying those firms that would provide the largest effect of R&D incentives are difficult, or not possible.

Moreover, government’s decision to adopt the direct incentives needs to go through inherently political process, so the choice of incentives may be driven by the best lobbyists rather than the degree of social returns. R&D incentives So I deleted it.

can be influenced by the political powers and economic cycle, therefore, businesses can’t plan to receive such supports in the longer-term. During the financial crisis of 2008–9, businesses significantly reduced their research activities (Campello and Harvey, 2010). Targeted R&D incentives has been used as a ‘stimulus packages’ to support those reduced research activities (OECD 2012). It also has administrative challenges. Given the complexity involved in selection and disbursement mechanisms, it pose much administrative burden both to the authority manage the incentives and the applicant firms, particularly for SME. (OECD 2016, WB 2010).

Tax incentives are usually neutral with regard to field of research and type of firm, and designed to allow firms to decide where to spend the R&D investment, so they have a wider reach than R&D grants (OECD 2010).

The disadvantages of direct incentives also work as advantages of indirect R&D incentives. Compared to the targeted support, tax incentives usually involve simpler administrative procedures than targeted incentives. Moreover, they are more predictable instruments for firms when run on a long-term legal base (Köhler, 2012).

Also, R&D tax incentives generally are compliant with competition and rules of the international trade agreements, such as the WTO’s TRIPS (Bora et al., 2000), which makes tax incentives popular over other targeted incentives.

Another motivation for their growing use is tax competition. Large multinationals can engage in sophisticated tax planning, reducing the effectiveness of R&D tax incentives (Guimón, J., 2009). R&D tax incentives can help attract internationally mobile R&D; see Section XXX for discussion of patent and IP boxes, which are an increasingly popular tax-based tool to attract IP revenue.

However, tax incentives might not be proved to the best firms and projects that would maximize the effect of R&D. For instance, firms may not choose R&D investment in the areas  with high social returns but choose the areas where they can maximize their profit (WB 2010).

Moreover, tax incentives are difficult to design and implement as it requires specific capacity (eg. Audit) within the government. It is difficult to monitor the spending of firms are not abused, and it is even more difficult to evaluating the direct effect that the tax subsidies generated. In fact, design and implementation of audits requirement by tax authorities have e a significant effects on firms’ take-up of R&D. (Bloom et al. 2001, WB 2010, OECD 2016).

Lastly, introducing a tax incentive is usually a significant policy and budgetary commitment while it is  difficult to forecasting and managing the fiscal impacts as they can’t anticipate the exact amount of tax incentives to be provided each year.  Meanwhile, as tax incentives are proportional to spending, it may favor larger firms, while creating administrative burden that will weigh relatively more on smaller firms.

Figure XXX: Summary of advantages and disadvantages of targeted vs untargeted R&D incentives

	targeted	untargeted incentives
pros	suitable for promoting strategic sectors and projects. create incentives for recipients to be accountable Can be used to mitigate the short term reduction of R&D investment (eg. Economic crisis)	Less distorting of market/private sector Less risk of mis-management Tax incentives more easily compliant with competition and international trade rules Can attract internationally mobile R&D.
cons	Depend on discretionary decisions by government, thus, potential corruption and mis-management of the process High administration cost Difficulties in targeting right beneficiary firms/technology Heterogeneity impact of direct supports            Vulnerable to the political powers and economic cycle, thus, difficult for business to plan in advance to use such supports in the longer	Limited scope for promoting strategic sectors/projects/technologies Higher risk of supporting non R&D related expenses Requires certain capacity of tax offices to audit and checks the R&D related spending Challenges in forecasting and managing the fiscal impacts Evaluating the direct effect may be impossibleCould favor larger (and multinational) than foreign firms

4 Empirical evidence on the effects of R&D incentives

This section discusses efforts to measure the effect of R&D incentives to firms and countries from the existing literature in countries with different income level. Such effects include both direct and indirect effects, and are organized in three groups, from more direct and short term effects to longer term effects; private R&D investment, innovation outcomes (eg. patents), firm performance and spillover effect.

4.1 Effect on R&D input additionality

Most studies on R&D incentives’ effect investigate on the change in business R&D expenditure resulting from R&D incentives (R&D input additionality), specifically test whether public R&D incentives complementary or substitute to private R&D investment. Here we review such literature and summarize the finding by targeted and untargeted incentives, and investigates whether those effects are different between targeted and untargeted incentives.

Targeted incentives

 

Are R&D incentives by governments a complement to private spending by firms or are subsidies paying for research that would have been done even without subsidies (that is, does government funding crowd out private)?

While the answer varies across studies, depending on the data and econometric strategies used, there are more evidence of complementarity that substitutes.

For instance, a previous meta-analyses find a mixed result. David et al. (2000) explained such mixed outcomes depend on market demand and supply conditions, and explained the larger substation effects at the micro-level studies, such as at the firm level, than aggregate-level studies is due to the industry differences in technological opportunity, and the effect of government funding of R&D raising the cost of R&D inputs. Meanwhile, more recent meta-analysis and country specific studies support the positive effect of targeted incentives on R&D input additionally. Correa et al. (2013) conducted a meta-analysis of 37 studies for the 2004-11 period and the effect of public investment in R&D is significant (the coefficient of additionality impacts on R&D ranges from 0.166 to 0.252).

While direct competition is not possible, R&D input additionality with is also shown with the country specific studies. For instance, firms show 1.34 times R&D investment than incentives received in Flanders, (Aerts and Czarnitzki, 2006). Firms that received R&D investment spent four percentage points higher R&D investment than other firms without crowing out the private R&D investment in East German (Almus and Czarnitzki 2003, Czarnitzki and Hussinger 2004).

Broadly-defined R&D in developed countries includes some amount of narrowly-defined research and some amount of industrial development.  Although one finds high-tech incentives in some developing countries (for example, Mongolia offers some incentives for the establishment of nanotechnology and biotechnology factories[8]), far more focus is on industrial development, including training and extension services. What would be called an education and outreach grant in a developed country becomes a contract for research extension services; see Andersen and Feder (2003) and Rivera, et al. (2000) for overviews of such contracts and the range of government involvement.

Note that studies on targeted R&D incentives in high-income countries typically measure the effect of government expenditures on changes in private R&D investment, while papers regarding low-income countries are typically measure the effect of government expenditures on final improvements in productivity due to government expenditures. In the other direction, spending on R&D is sometimes used as a proxy of the “absorptive” or “national learning” capacity of firms and countries to adopt existing technology (At the firm level, see Cohen and Levinthal (1990), Forbes and Wield (2000), Griffith, Redding and Van Reenen (2003), Pavitt (2001) at the national level see, for example, Baumol, Nelson and Wolf (1994)).

There are growing number of literature that evaluate the specific targeted R&D incentives’ impact. To give one representative example, Tsusaka et al (2016) evaluated the returns on a targeted program to research, train use of, and facilitate adoption of improved methods for groundnut cultivation, and found a return of 22% over the base cost of capital. Like many programs in developing countries, the project was a collaboration between government, NGOs, and some private sector companies, and the expenses were totaled across the entire project. Counting only government expenses would lead to a higher return on government funding.

Dlamini and Liebenberg (2016) measured the costs and returns from an R&D project on maize cultivars, including both government research and grants to seed producers, from 1993 to 2012, and found a return of 37 rand for every rand spent, or an expansion of 19.9% per year.

For comparison, a classic paper by Griliches (1958) indirectly measured the returns to research on hybrid corn in the established U.S. farming context from 1910-1955, thanks in part to U.S. government development programs, and found at least a sevenfold increase, or an expansion of 4.3% per year. But Malla et al. [2004] use yield and quality measures of the Canadian canola crop, and finds that “the combined effect of IP [rights] and public incentives has driven the quantity of research beyond the socially optimal level.”

The goal for Alston, et al. (2000) was to gather every agricultural R&D impact evaluation from 1953 to 1997, including the full range from pure research to pure outreach; they found 1,128 estimates of return rates, from all levels of development but with a bias toward the United States.  330 estimates were from papers with government employees as first authors, indicating government funding or outreach. The mean return from the subset of 219 studies used in their analysis was 65.75%, but the standard deviation of 82.47% with some very large returns and many failures.

Both the studies of the effect of government expenditures on private investment in high-income countries and studies of the effect of government expenditures on final output in mid- and low-income countries show wide variance, confirming that targeted grants can be a lottery that sometimes returns a high payoff and sometimes a total loss.

Untargeted – general tax incentives

 

Several studies consider the question of whether untargeted R&D incentives are substitutes or complements to private R&D. For instance, Hall and Van Reenen (2000) provide a survey of the efficacy of R&D tax incentives and the user cost of R&D in OECD countries. Köhler et al. (2012) summarize the findings of the effectiveness of such incentives in more than 20 studies in 12 countries from the 1970s to the 2000s. Both surveys find positive effects of tax incentives on private R&D, even if they sometimes also subsidize R&D activities that firms would conduct anyway.

Cross-country studies have also supported the positive effect of R&D tax incentives. Using data on tax changes and R&D spending in nine OECD countries from 1979 to 97, and a range from broadly-defined R&D credits (USA, Japan) to narrowly-defined (Germany), Bloom et al. (2002) found that “a 10 percent fall in the cost of R&D stimulates just over a 1 percent rise in the level of R&D in the short-run, and just under a 10 percent rise in R&D in the long-run.” R&D incentives have a non-linear (convex) effect on private R&D additionality in 25 OECD countries from 1990 to 2009 (Montmartin and Herrera, 2015). The positive effect of tax incentives on R&D input additionality has increased as R&D tax incentives usage increased. For instance, OECD countries with R&D tax incentives has increased from eighteen in 2004 to twenty-six in 2011 with the number of OECD countries using them rising (Correa and Guceri 2013).

Country studies also provide consistent message. In the United States, the R&D tax credit produces roughly a dollar-for-dollar increase in reported R&D spending on the margin (Hall and Van Reenen, 2000). In France from 1993 to 2003, one Euro of the incremental R&D tax credit would give slightly more than one Euro of total R&D (Duguet 2012). In UK from 2006 to 11, tax price elasticities are as large as 2.6 on relatively narrowly-defined R&D. It was projected that during the same period, aggregate business R&D would be around 10% lower in the absence of the tax relief scheme (Dechezleprêtre et al. 2016). The author explains this large effect by the fact the treated group is from a sub-population of smaller firms which are to financial constraint. In fact, a study used different subset of firms also find a positive effect but the user cost elasticity was between −0.88 and −1.18 (Guceri, 2017).

While the above mentioned studies find a positive effect of R&D incentives on R&D additionality, the results require a careful interpretation. For instance, the amount increase of R&D might not necessarily mean that the increase of R&D input in the case when the government support increase the price of R&D input, such as salary of R&D workers.

Also, the magnitudes of effect vary, and range from as small as 10% (in the OECD countries 1979-1997, in the short-run, although it increases to 100% in the long-run, Bloom et al., 2000) to as large as 296% (in the US, Klassen et al. , 2004) , depending on the countries and data used, which is very large even considering different methods and time period used for each study. Then, why does the effect of R&D incentives vary so much?  Here we provide the sources of these high variance effects, which depend on the number of years after the implementation of such incentives, and industry and firm characteristics. Also, the design and implementation know-how matters to the impact.

The first factor seems to be related to how long the tax incentives have been provided. The effect of an R&D tax incentives increases over time (Bloom et al., 2000, Hall and Van Reenen, 2000, McKenzie and Sershun, 2010). This probably implies that it takes time for private firms to learn about the R&D incentives and its benefit and there might be significant time required for the private firms to materialize their R&D investment (eg., learning about how to hire and what to purchase).

The effects also vary depending on the industries. An increase in the R&D tax credit rate has a positive effect on the R&D spending of high-tech firms (Chen and Gupta, 2017). But there are inconsistent findings on which sectors show higher additionality. In countries with an incremental scheme, higher additionality are found for firms in service and low-tech sectors (Castellacci and Loe, 2015).

The effects also vary depending on recipient firm characteristics, such as firm size and being a multinational. Crowding out of private R&D can be avoided for small firms, but not for large firms in Netherlands (Lokshin and Mohnen, 2012). Further, multinational enterprises (MNEs) are major R&D investors but also have cross-border tax planning opportunities that small firms do not, such as the ability to shift assignment of intellectual property to the country with the most favorable tax treatment. We will discuss about this effect more detail in the section on IP box. This implies that an R&D policy that is neutral on paper may in practice are more beneficial to large and (multinational) firms than SMEs.

Also, the design of the tax incentives matters for the effects. Volume-based tax incentives, which holds for more generous schemes, produce higher additionality while incremental schemes have little effect when the market environment for R&D activities is unfavorable, such as economic downturn (Köhler et al. 2012). When the tax incentive is provided credit based on incremental R&D spending, firms opportunistically time their R&D investment patterns to obtain tax credits, which lead to the potential loss of tax revenues (Chen and Gupta, 2017).  This may imply that credits can only be taken to offset profits, so startups with no profits may not be able to take credits, yet they’re more likely to be building new processes.

Meanwhile, in terms fo impleemtation issues, sustainability and supporting activates also matter for the effect. Frequent R&D tax policy reversals greatly hempen the R&D tax incentives’ impact on private R&D additionality (OECD 2016). R&D incentives provide a better result when it was complemented by other supporting business environment and inputs, such as university research, -skilled worker, and R&D cooperation (Becker, 2014).

However, our knowledge about the efficacy of R&D tax incentives is limited to the incentives that developed in the 1980s and 1990s (Köhler et al., 2012). Little is known about the newly introduced incentives, although they may show different effects compared to the traditional ones.

 

 Untargeted – Patent Boxes

An increasing number of countries have introduced Patent boxes or Intellectual Property (IP) boxes, which provide a lower tax rate for income generated from intellectual property, such as a patent. This section discusses this class of untargeted incentive which currently has limited coverage in the literature.

Many corporations and universities have contracts with employees stipulating that inventors will assign the royalty rights to new patents to their employers, and some firms then assign ownership to a corporation in a tax haven such as the Cayman Islands or Luxembourg.

IP boxes are primarily intended to attract royalty assignments to within a country’s borders. For example, the UK government states that its patent box is intended “to ensure new and existing patents are further developed and commercialized in the UK.”[9] two bills introduced in the U.S. Congress to introduce patent boxes state an intent “to encourage domestication of intangible property”.[10] One of the stated intents of the OECD’s Base Erosion and Profit Shifting (BEPS) project is to reduce involved countries’ use of patent boxes to induce reassignments across borders.

But potential revenue gains from IP boxes are tenuous. Evers et al. (2015) raise some concerns regarding the technical details of implementation. For example, if firms can deduct expenses at higher ordinary rates and pay taxes on IP-related income at the lower IP box rate, they may be able to achieve a negative tax rate. Klemens (2017) shows that even though a country with a lower IP box rate may attract more IP assignments to within its borders, it is unambiguously worse off than with a single tax rate for all revenue.  Griffith et al. (2015) empirically estimate the effect on revenue estimates, and find that “[p]atent Boxes are likely to attract patent income, but our estimates suggest they will also lead to substantial falls in tax revenues.”

Setting aside international tax competition, are patent boxes the right tool to spur inventive activity? There is some conceptual distance between lower rates on assignee royalties and inventor incentives, so the effect on patent rates is difficult to predict. Bradley et al (2015) gathered data on patent applications and assignments from 71 countries, and found that “virtually all countries appear to exhibit declining rates of aggregate patent activity immediately following the start of their patent box regimes.” [p 14] and “the propensity for domestic inventions to be patented is significantly lower after a patent box regime is implemented.” [p 15] However, because many countries have only recently implemented IP boxes, the data may not yet be conclusive.

4.2 Effect on firms’ narrowly-defined outcomes

R&D expenditure are not only expected to induce private R&D spending, but also some output of greater “inventiveness” or “innovativeness”. Private R&D spending has been used as a measure of inventiveness, but there is a wide gap between the broad legal definitions of R&D in Section 2 and the far more narrow definition of what can be patented: both basic scientific research into “laws of nature” and general industrial development are typically unpatentable.

Measures of narrowly-defined technology outputs, therefore, go to the source and use patent counts to measure the inventiveness of firms that received R&D subsidies, and sometimes measuring the efficacy of policy shifts (e.g., Bloom, et al., 2013; Bottazi and Peri, 2003; Bronzini and Piselli, 2016; Bosch, et al., 2005; Brahmbhatt and Hu, 2010; Dechezleprêtre, 2016; Furman et al., 2002; Graham et al., 2009; Griliches, et al., 1987; Jaumotte and Pain, 2005; Lerner et al., 2008; World Bank, 2010, p 150ff)

Figure 4: Patent counts versus percent of direct investment in R&D out of all R&D (left) and amount of direct investment in R&D

Figure 4.4 compares direct R&D investment to patent counts, and shows no evident relation.

Jaffe et al. (1992, p 8) state that “The main advantage of patent data can be stated simply: they are easily available.

Despite their ease of analysis, their use is problematic.  First, with many legal definitions underlying R&D credits covering quotidian improvements not focused on novel invention, we can expect that any correlation between novel invention and such incentives would be weak. Figure XXX looks at the relation between direct investment in R&D (which tends toward being more narrowly-defined than indirect incentives), and the ratio of direct to indirect investment. Both show no discernable relation between these incentives and patent counts.

Cat Field 1994 2014 2016

304 Communications: Electrical 657 2,881 (+439%) 2,778 (-3.6%)

705 Data Processing: Financial… 171 5,902 (+3,451%) 2,136 (-53%)

Figure XXX: Applications received by the U.S. Patent and Trademark office for two categories: “Communications: Electrical” and “Data Processing:

Financial, Business Practice, Management, or Cost/Price Determination”.

https://www.uspto.gov/web/offices/ac/ido/oeip/taf/cbcby.htm

Further, policies for what is or is not patentable change across counties and years. Figure XXX shows counts of patent applications received by the U.S. Patent and Trademark Office (USPTO) in two categories. To give a baseline, the field of “Communications: Electrical” shows healthy gains during the advent of the Internet era, rising 439% from 1994 to 2014, and holding steady into 2016. By contrast, applications received under “Data Processing: Financial, Business Practice, Management, or Cost/Price Determination” showed an increase of 3,451% over the same period, followed by a 53% drop in applications by 2016.

Even if there truly was a renaissance in business practice technique from 1994 to 2014, it is difficult to explain why it collapsed so suddenly in 2016. The most likely explanation lies in a series of court rulings, beginning in 1998 when an appeals court ruled that business methods were within the scope of patentability,[11] and concluding in 2014 with a Supreme Court ruling that a large class of business methods are not patentable.[12] (see Klemens [2005, 2008] for a timeline and discussion at the peak of activity).  For the purposes of using patents to measure productivity, there was no collapse in growth among U.S. firms that make use of business methods comparable to the 53% decline in applications.

Business methods underwent a large-scale shift, but in 2016, the USPTO posted 65 notices, rules, and proposed rules, each of which may change the decision process for an inventor considering a patent application.[13]Cockburn et al. [2002] provides an interesting look at how patent examiner practice can affect citation counts. For example, some patents, often from academia, are so thoroughly written that examiners repeatedly cite to them in subsequent applications.

In the five years from 2011 to 2015, the Chinese domestic patent count more than doubled (from 415,829 to 968,252); procedural changes that indicated a greater concern for IP, including the establishment of a central IP court in 2014, likely had some part in the doubling of patent counts.[14] In Morocco, in the one-year span from 2012-2013 resident patent counts jumped 60% from 197 to 316, which may be related to a new fee schedule that took effect in late 2012.

A number of papers find a correlation between citation-weighted patent counts and stock market value, including Nicholas [2008], Hall et al. [2000], and the survey of Griliches [1990]. Some, like Kogan et al. [2017], directly link this growth to aggregate economic growth and advances in total factor productivity. An alternative hypothesis to the claim of aggregate growth may be that the exclusive rights granted by a patent allows rent seeking by patent holders against firms that independently practice the patented item, as one firm gains exclusive rights to an invention that may otherwise have been in common use. One could get some evidence on the likelihood of zero-sum rent seeking by considering how often patented items are independently reinvented by competitors. Cotropia and Lemley [2009] looked at U.S. lawsuits for patent infringement, in which the patent holder may be granted triple damages If the alleged infringer directly copied the patent-holder (rather than independently invented). Despite this strong incentive to make any reasonable claim of direct copying, the authors found that “only 10.9% of the complaints studied…contained even an allegation that the defendant copied the invention.”

Although they do not address it, Kogan et al provide some indirect support for the rent-seeking hypothesis, as they find that stock market returns for a firm are negatively correlated with the count of competitor patents.

In the context of countries that are developing a patent regime, from Albania (with 14 domestic patents in 2015) to the USA, incentives to file patents and the procedures to grant them are constantly changing, making it nearly impossible to use patents as proxies for narrowly-defined invention. This is especially true in the case of low-income countries and cross-country studies. Given these difficulties, the remainder of this paper, and the bulk of this literature, focus primarily on broadly-defined R&D, which can be more readily measured via expenditures and overall productivity measures.

4.3 Effects on firm performance and spillover effects

Evaluations of firm performance of R&D incentives are rather scarce.

although the importance of R&D in enhancing firms’ productivity and profitability are widely documented (Sougiannis, 1994; Green et al., 1996).

There is no clear evidence that R&D incentive raises measures of firm performance, such as TFP.

Köhler et al. 2012 explains “R&D tax incentives stimulate R&D projects with a lower marginal rate of return so that productivity impacts of these projects are minor”

the evaluation on the government’s direct incentives on enhancing recipient firms’ innovation performance were concluded as failed in 1990s (Guan and Yan, 2015).

Meanwhile, many literature investigate the social return of R&D.

Social return

Nelson [1959] puts forth the theory that basic research generates “social profit”—public externalities in excess of public spending on R&D. The essay raises the question of whether the externalities gained from research are greater than the externalities gained from other allocations of public and private resources, but provides neither a formal model nor data to answer the question, so other researchers have worked to fill this gap.

The “return” to R&D is the outcome of a complex interaction between firm capability, market competition and other external factors, such as economic crisis. Therefore, these returns may significantly vary over time or across sectors or countries. Specifically, for the social returns, even the concept of cost of capital is somewhat uncertain. However, estimating such returns can still provide insightful information as it allows comparison across times for the same R&D incentives provides. Also, as Hall et al (2010) noted “these estimates can still be useful for making comparisons between various financing systems, sectors, or countries, and can also be a guide to policy-making toward R&D”.

Here we provide the estimated returns on R&D from different countries.

R&D has high social return. While the return to private R&D is higher than to fixed investment (7 times higher in the case of in a cross section study of 53 countries, Lichtenberg, 1994 and slightly higher in Halle et al. 2010),  the  social returns are even higher than return to private R&D (Hall et al. 2010). Extremely high social rates of return ranging from 71% (Griliches and Lichtenberg 1984) to over 100% (Terleckyj 1980 and Scherer 1982). However, the definition and measurement of social returns can be somewhat subjective, making these results potentially difficult to replicate. For example, Tewksbury (1980) finds comparable social returns to the above authors but gives no description of the methods by which social returns were coded.

Estimates using cross country data, thereby presumably capturing intra-country spillovers, find rates of return to R&D of 123% for the G7 and 85% for the remaining 15 OECD countries; returns of 68% in the G7 and 15% for a subset of the remaining OECD countries (Coe and Helpman, 1995, van Pottelsberghe de la Potterie and Lichtenberg, 2001). Positive spatial spillovers among private R&D investments (Montmartin and Herrera, 2015, using 25 OECD countries, for the period 1990–2009).

Other Spillover (bullet points to be used or deleted)

• The smaller returns to R&D at the firm level than at the industry or national level, may imply that there is substantial spillover effects that one firm’s R&D may contribute to another firm’s knowledge (Bosch, et al. 2005).

• There are positive inter-industry productivity spillovers from R&D, in Chinese manufacturing sector for 1998–2001 (Wei and Liu, 2006).

• Wei and Liu, 2006. Productivity spillovers from R&D, exports and FDI in China’s manufacturing sector. Journal of international business studies, 37(4), 544-557

• Cf) the R&D generated by the tax policy creates positive spillovers on the innovations of techno-logically related firms in U.K, 2006-11.( Dechezleprêtre et al. 2016).

4.4 Developing countries vs developed countries

Here we investigate whether there are common trends and differential effects of R&D incentives between developing and developed countries.

Developing countries invest less on R&D. R&D investment is strongly correlated with income level, measured as GDP (emerging economies showed very high R&D level) and rises with level of development at an increasing rate (Lederman and Maloney, 2003).

Developed countries tend to use both targeted and untargeted incentives, while developing mostly used targeted incentives only. As seen in figure 1 above, R&D tax incentives share rises with income level. The upward slope indicates that countries with higher GDP per capita are more likely to offer an untargeted credit than lower-productivity countries. Also, patent box – specific form of tax incentives- are also only a few developed countries.

Then, why do developing countries not provide tax incentives to promote R&D, although they are motivated enough to provide targeted R&D? We summarize the reasons as below.

First, developing countries use targeted R&D incentives to tackle specific targets, such as promoting innovative firms in strategic sectors. This is contract to developed countries that have a very broad approach to R&D.

Second, developing country may have a lower private demand for R&D while having skill (lack of R&D workers, for example) and budget constraints. Small scope of targeted incentives could be sufficient to cover the low demand for R&D, while  tax incentives could require more budget.

Third, capacity constraints of the government may limit the use of untargeted tax incentives in developing countries. As seen above, tax incentives are complex to design, and they require a robust and skilled audit capability within government to ensure they are not abused.

While there is limited evidence on returns on R&D incentives in developing countries, our review concludes that returns on R&D are smaller in developing countries than in developed. While effects of tax incentives in OECD countries are growing, documented effects of such policies in developing countries are only a few although there is increasing interest in developing countries in granting R&D incentives. The studies we review show that R&D incentives’ effect on R&D investment increase from 100% to 296% from developed countries, while less than 100% for developing countries. This is consistent with previous cross country analyses that find that R&D investment produce differing result across countries and the returns to R&D diminish with countries income level (Lederman and Maloney, 2003, Camagni and Capello, 2013).

(I will look for more evidence – will delete this table alter )

From developed
a 10% fall in the cost of R&D 1% rise in the level of R&D in the short-run 10% rise in R&D in the long-run	OECD countries, 1979–1997, Bloom et al., 2000
Dollar in tax credit a dollar of additional R&D	Hall and Van Reenen, 2000
R&D policies have a non-linear (convex) effect on private R&D	Montmartin and Herrera, 2015, using 25 OECD countries for the period 1990–2009
1 Dollar R&D spending average $2.96 of additional R&D spending in U.S. system $1.30 under the Canadian system	Klassen et al. , 2004)

From developing
R&D tax incentives has increased firms’ innovation investment However, when innovation investment is divided into innovation related capital goods expenditures and only R&D, the results suggest that the absolute value of the elasticity for the R&D component of the innovation investment is less than 1.	Argentina from merged data from the National Innovation Survey collected by the National Institute of Statistics and Censuses. Crespi et al, 2016).
tax incentives (i) are incentives used as tools of tax competition and (ii) how effective are incentives in attracting investment?	40 Latin American, Caribbean and African countries for the period 1985–2004 no robust evidence, however, for competition over investment allowances and tax credits. Using dynamic panel data econometrics to answer the second question, we find evidence that lower CIT rates and longer tax holidays are effective in attracting FDI in Latin America and the Caribbean but not in Africa. None of the tax incentives is effective in boosting gross private fixed capital formation. (Klemm, A., & Van Parys, S. (2012). Empirical evidence on the effects of tax incentives. International Tax and Public Finance, 19(3), 393-423. But this is about general tax icnentives, not specific to r&D incentives

Weak business environment and institution contribute to the low return of R&D in developing countries. For instance, Roper (2010) identifies the weak R&D system in Western Balkan countries prevents R&D investment from producing innovation outcomes. Lederman and Maloney (2003) identifies that “financial depth, protection of intellectual property rights, government capacity to mobilize resources, and the quality of research institutions” are the constraints that prevents R&D effort rising with the level of income.

However, the low effect of R&D could also be from measuring the wrong thing. For instance, outcome measures should be broader than patents, particularly in developing countries.

Adoption/adaptation, rather than creation, are crucial for developing countries. Aghion et al. [2013] and Delbono and Lambertini [2017] propose a Shumpeterian perspective that firms far away from the technological frontier are better off improving their processes by imitation of pre-existing methods, while those at the frontier should push toward novel research. At the country level, less developed countries (far from the frontier) might be better off by promoting technology transfer from more advanced countries while more developed countries (close to the frontier) might be better off for providing incentives for firms to innovate and create new things.

Promoting R&D could serve as a tool for industrialization and economic growth for developing countries. There may be industrial policy goals, such as a desire to focus resources on specific industries as part of a larger economic strategy. An extensive literature documents how product sophistication increases as a country get wealthier, and the structure of industry changes over time according to their level of incomes [Hausmann et al., 2005, Rodrik, 2013]. Governments can apply targeted incentives to promote strategic sectors and products which are exportable, upgrading the countries knowledge assets, and/or high spill-over to other sectors.  In fact, Korea and Taiwan’s rapid growth in the global electronics industries during 1980–2001 coincided with the dramatic increase in both countries’ R&D expenditures during that period (Chen and Gupta 2017). Governments of Korea and Taiwan have played a central role in forming a business group-centered R&D strategy and in strengthening public R&D infrastructure that contribute to the growth of firms’ R&D productivity (Nagano, 2006). Some rapidly industrialized countries, such as Finland, Israel, Korea, and Taiwan, and countries that experienced a “take off” such as China and India were radically deviated from the predicted trajectory and displayed impressive R&D take-offs (Lederman and Maloney, 2003).

 5 Policy implications for development

R&D incentives in high-income countries are partially about high tech, cutting-edge research, and partly about the day-to-day problems of improving production in all fields. These R&D incentives are not fundamentally different from R&D incentives in countries with limited potential to be on the forefront of the sciences, though the balance of basic and applied research will differ. This means that the lessons learned from high-income countries with a long R&D incentive history may still be applicable in the development context.

We structured the story around two categories of incentive: directed incentives like grants, and undirected, almost always implemented via general tax law. Within our sample of 131 countries, low-income countries are less likely to have untargeted tax incentives—in the lowest decile of GDP per capita, about 8% of countries had some R&D incentive in its tax law; among the top deciles, about 80% had R&D tax incentives.

There is consensus in the literature that government R&D incentives induce greater R&D spending in the private sector, rather than crowding out private spending. But the size of the effects differs in consistent ways: laws that have been in place for decades allow planning in a manner that constantly shifting laws do not, and high-tech fields are more likely to show greater private spending while lower-tech industries may show substitution away from private spending. As a final point, these untargeted incentives are handled through tax law, so they will require good predictions before implementation about expected uptake and budget implications and an effective audit system to prevent widespread abuse after implementation.  Each of these points favors higher-income countries with a reliable and stable bureaucracy. Because lower-income countries are less likely to have R&D tax credits, it remains to be seen whether adoption by low-income countries would induce the complementarities in private spending higher-income countries have shown, but the patterns above indicate that this is far from certain.

Targeted incentives based on government funding appear across the development spectrum, though they are typically called grants in high-income countries and R&D contracts, projects, or collaborations in low-income. These incentives showed similar results across contexts: some programs produced exceptional returns to output, while others failed. This does not imply that low-income countries should avoid targeted R&D programs, but that it is important to maximize the odds of success, picking programs with known records in similar contexts, maintaining focus on bottlenecks in existing production systems, and enlisting the expertise of NGOs or private firms when such expertise is needed.

As a side-note, given the perception of R&D as being about innovation and inventiveness, there has been much effort put into correlating patent counts with R&D incentives as a proxy for the innovative force of the incentives. Even in the United States, with its well-established patent system, changes in patent law, both minor and large, made certain patent counts unreliable. In countries that are beginning to develop a patent system, such changes in policy, funding, and enforcement for patents are even less stable. Although R&D incentives are frequently paired with pitches about innovation and invention, their most reliable means of evaluation are based on simple productivity gains.

Seen in this context, divorced from pictures of people in lab coats, R&D incentives across the development spectrum are frequently aimed at furthering productivity of nearly all types of production and processes, giving countries developing new R&D policies abundant opportunities to study existing systems that have much in common.

Some policy issues I can think off – will add more bulet points here before writing and merging them with yours above.

Cf) government considering R&D incentives should choose the right instruments based on their objectives, capacity and private demand

Cf) R&D incentives matter for R&D additionality, and spillover

Cf) but the crowing out effects do exist and results vary across programs/incentives

Cf) less evidence of R&D output and firm performance might be to do i) with measurement problem of R&D outcome and 2) a challenge to attribute the R&D incentives’ direct effect on longer term and

Cf) the design and implementation know-how matters and the mechanism to correct design over time is important.

Cf) Use mixed methods of targeted and untargeted: I wrote on this a while ago but it has been disappeared.

Cf) contrast to the prediction of growth theories, R&D (incentives’) return lower in developing countries, i) which may suggest that complementary policies and environment significantly matters => so developing countries should implement complementary support, and ii) measurement issues => developing countries should not focus on only narrow definition of R&D.

Cf) should do better job for evaluation – particularly for longer-term effect

Cf) concern about cross-country analysis or even for a meta-analysis : R&D definition varies significantly across countries. Therefore, some of the existing studies quantify the effect of R&D incentives probably didn’t compare “apple to apple”, although they provide useful information. A complementary case studies and careful interpretation is useful.

[1] Under U.S. law, the result of systematic work to seek out new knowledge may be excluded from patentability for being “obvious to try” [KSR v Teleflex, 127 S. Ct. 1727.] The unpatentability of new scientific discoveries has a long history; see the Supreme Court’s ruling in Gottschalk v. Benson, 409 U.S. 63, 175 USPQ 673 [1972] for an extensive history.

[2] http://investinamericasfuture.org/PDFs/Global_RD_Survey_September_2012_FINAL.pdf

[3] Law 392-07 (amended in 2014 by Law 542-15), §46.

[4] Education policy can incentivize science, technology, engineering, and math (STEM), or certain subfields within that range. Given that women are often under-represented in STEM education (see e.g., Beede et al. [2011]),

[5] Data coded by the authors based on Price Waterhouse Cooper summaries of tax conditions in each country; see http://taxsummaries.pwc.com.

[6] Source: 2016 GDP per capita from World Bank data bank. Presence of tax credits from Price Waterhouse Cooper reports, mostly updated in early 2017, coded by the authors.

[7]  Source : NIH Federal RePORTER, and the US trensury. Available at  https://www.treasury.gov/resource-center/tax-policy/tax-analysis/Documents/RE-Credit.pdf.  Note that this is not a comprehensive listing of all possible R&D incentives; for example, there are also targeted tax credits for drug development for rare diseases and renewable energy research. However. no estimate of the costs of these targeted credits was immediately available, and  the costs are known to be small.

[8] http://taxsummaries.pwc.com/ID/Mongolia-Corporate-Tax-credits-and-incentives

[9] https://www.gov.uk/government/consultations/patent-box-substantial-activities

[10]Schwartz proposal: https://www.congress.gov/bill/113th-congress/house-bill/2605/text

Boustany-Neal proposal: http://waysandmeans.house.gov/wp-content/uploads/2015/07/Innovation-Box-2015-Bill-Text.pdf

[11] State Street Bank & Trust Co. v. Signature Financial, 149 F.3d 1368, 1375 (Mass. 1998).

[12] Alice Corp v CLS Bank, 573 U.S. __, 134 S. Ct. 2347 (2014)

[13]  https://www.regulations.gov/searchResults?rpp=25&so=DESC&sb=postedDate& po=0&a=PTO&pd=01|01|16-12|31|16

[14] World Bank Data Bank.