How can policymakers boost long-term growth in the post–COVID-19 global economy? This chapter looks at the role of basic research—undirected, theoretical, or experimental work. Using rich new data that draw on connections from individual innovations to scientific articles, it shows that basic research is an essential input into innovation, with wide-ranging international spillovers and long-lasting impacts. International spillovers are particularly important for emerging market and developing economies, where institutional factors—including better education and deeper financial markets—help convert innovation into economic growth, making rapid technology transfer, the free flow of ideas, and collaboration across borders key priorities. Model-based analysis reveals that advanced economies could raise long-term growth by increasing research funding, targeting basic research, and developing closer connections between public and private research. By lifting the growth potential and future tax base of the economy, these investments tend to pay for themselves within a decade. Investments in basic research may also have green benefits, as cleaner technological innovations rely on newer, more fundamental research.
Introduction
Few concepts have implications as far reaching for economic policy as long-term growth. Growth— namely, the increase in an economy’s potential to produce goods and services—is of central importance not only for improving living standards, but also for addressing inequality, debt sustainability, and the cost of climate change mitigation.
Yet, the past few decades have seen a long and persistent decline in long-term growth. Policymakers face an urgent and essential question: how can this trend be reversed to build a more buoyant post-pandemic global economy? Although this has so far been mostly an advanced-economy phenomenon, demographic trends in China and other emerging markets make the need for an answer more urgent. With fewer active workers, aging populations will require more output per worker to maintain living standards.
Addressing this question requires an understanding of the underlying drivers of growth. The earliest explanations emphasized the role of productivity—the ability to create more outputs with the same inputs.1 More recent work emphasizes the role of innovation—the emergence and adoption of new technologies that improve the production of goods and services—as a driver of productivity.2 But the data present something of a challenge to this idea. Productivity growth has slowed, even amid increased spending on research and development—a common proxy for innovation effort (Figure 3.1, panels 1 and 2). This apparent conflict with leading theories makes formulating policies to boost long-term growth rather difficult.
Measures of Research and Productivity
Sources: OECD Science and Technology Indicators; Penn World Table 10.0; Reliance on Science; United States Patent and Trademark Office; and IMF staff calculations.Note: In panel 1, labor productivity growth is reported as a three-year moving average. The shaded area denotes the 25th to 75th percentile. Sample is restricted to be balanced throughout the period. In panel 3, the figure shows the average difference in funding for applied minus basic research over time. In panel 4, average citations from patents to academic articles and other patents are shown by year of application. The spike in 1995 is likely associated with a legislative change prompting an increase in patent applications (Byrne 1995). R&D = research and development.One possible answer is that the type of research matters. Innovations, great and small, occur not in a vacuum but draw on the stock of basic scientific knowledge. The invention of the cardiac pacemaker required a scientific understanding of both human anatomy and electronics. The GPS technology familiar to many smartphone users relies on Einstein’s theories of relativity to account for how time passes at different rates on fast-moving satellites and the Earth’s surface. More recently, the extraordinarily rapid development of COVID-19 vaccines, based on decades of prior basic scientific research, has had the massive economic payoff of bringing forward the reopening of many economies, perhaps by years (Box 3.1). Growth in research inputs has been increasingly applied, even as innovation depends more on basic scientific advances (Figure 3.1, panels 3 and 4), which may help resolve part of this puzzle.
The character of basic scientific research also suggests that policies to encourage it might be particularly potent—something relevant to aspirations to build a better post–COVID-19 economy (see Chapter 1). In contrast to applied innovation, basic research can have very broad economic applications. While this likely means that social returns from basic research are high, it also means that firms may struggle to internalize the gains from basic science, undermining private incentives. No firm could fully capture the gains from the invention of, say, the jet engine or the internet. As a result, private firms are likely to underprovide the most basic, far-reaching, and economically impactful types of research (Nelson 1959)—suggesting a role for public policy to bridge this gap.
This chapter explores whether public policy should support basic scientific research to boost growth during the exit from the global pandemic, addressing the following questions:
What is the progression from basic science to innovation and productivity growth? How does basic scientific knowledge diffuse internationally? And how do the economic roles of basic and more applied research differ?
What is the global economic benefit of scientific integration? How might a reverse in scientific integration of major economies, such as the United States and China, affect global growth?
Is basic research under- or overprovided? Can policy intervene to correct socially inefficient levels of basic research? If so, what is the appropriate policy mix? How should these policies balance returns from public and private basic research? And what are the potential gains from such policies? Can basic scientific research help in the fight against climate change? And if so, how might those benefits manifest?
These are the chapter’s main findings:
Basic scientific research is a key driver of innovation and productivity, and basic scientific knowledge diffuses internationally farther than applied knowledge. A 10 percent increase in domestic (foreign) basic research is estimated to raise productivity by about 0.3 (0.6) percent, on average. International knowledge spillovers are more important for innovation in emerging market and developing economies than in advanced economies. Easy technology transfer, collaboration, and the free flow of ideas across borders should be key priorities.
A decoupling of basic scientific research between the United States and China could have big negative effects on global productivity, with an estimated first-round decline of up to 0.8 percent.
Basic scientific research in advanced economies is underfunded. As a result, policies that fund public research and subsidize private research will have positive payoffs. A model estimated on three advanced economies suggests that subsidy rates for private research should be approximately doubled and public research expenditure increased by about one-third. Targeting support to basic scientific research will deliver the greatest return but, where this is not possible, more public-private partnerships may be a partial substitute. While such policies pay for themselves in the long term, optimal research funding may be lower in countries with immediate fiscal constraints. Science also plays a larger role in green innovation than in dirty technological change, suggesting that policies to boost science can help tackle climate change.
Conceptual Framework
The chapter’s conceptual framework draws on innovation-driven endogenous growth theory (Romer 1990; Grossman and Helpman 1991; Aghion and Howitt 1992; Akcigit and Kerr 2018), in which knowledge creation plays a central role in driving productivity growth.
In its simplest form, economic output can be thought of as produced by two interlinked production functions (Figure 3.2). In the first, the production function for ideas, research inputs—both basic and applied—are combined with preexisting knowledge to produce economically relevant innovations that add to the stock of common knowledge. The key difference between basic and applied research is that the former is undirected, theoretical, or experimental, whereas the latter is aimed at bringing products to market. In the second production function (the one for goods and services), standard macroeconomic inputs (capital and labor) are combined to produce output. The productivity of this process depends on the current stock of ideas and other country-specific institutional factors. Thus, research increases knowledge, knowledge enhances productivity, and productivity determines how much final output is generated from real inputs.
Although the analysis in the chapter adds finer details to this picture, the basic structure remains the same throughout. The empirical analysis unpacks these two production functions and estimates the direct impact and international spillovers of investing in basic science. Subsequent model-based policy analysis complements the empirical evidence by allowing for richer interactions, including between basic and applied research in general equilibrium. Given that the analysis of the more basic types of research is novel, the chapter’s focus is naturally on basic research. For more on applied research, see the April 2016 Fiscal Monitor and the April 2018 World Economic Outlook.
Connecting Basic Science to Growth
This section presents an empirical investigation into the two production functions outlined in Figure 3.2, extending it to include an international dimension, distinguishing the impact not only of basic and applied research but also the extent of international spillovers. An important first step is to construct measures of the stock of foreign knowledge accessible to each country.
The Diffusion of Basic and Applied Knowledge
The relevance of knowledge in one country for an innovator in another may depend on a variety of factors, including proximity, language, and so forth, and might be different for basic and applied knowledge. Cross-country citations in patent applications, from the Reliance on Science database (RoS, for basic research) and from PATSTAT (for applied research), provide valuable clues about the drivers of the international transmission of knowledge.
The RoS database is a rich data set that tracks citations of some 38 million US and European patents to scientific articles (Marx and Fuegi 2020). By providing unique identifiers for patents issued by the US Patent and Trademark Office, RoS can identify the countries both of the patent s inventor(s) and of the authors of cited scientific articles. PATSTAT, maintained by the European Patent Office, provides global coverage of patent applications, with 105 million records from more than 190 patenting offices. These sources illuminate two inputs to the production function for ideas, basic and applied research, and are discussed in Online Annex 3.1.3
A key assumption in the empirical work is that citations to scientific articles capture dependence on basic research and that citations to patents capture reliance on applied research. This draws a sharp distinction, whereas reality is more blurred; some articles may cover applied topics, and patentable work may spur major scientific breakthroughs.4
Figure 3.3 shows the main patterns of international citations of basic knowledge, using cross-border citations in the RoS. The United States is the main source of cited works—a constant in recent decades. However, citations to Chinese science have grown strongly since 2005 (albeit from a low base), as have citations across Asian countries. In general, regions tend to exhibit home bias, citing their own scientific works more than others do. This suggests that diffusion of knowledge from its source is partial—a point explored more formally in the next section.
Geography of International Basic Knowledge Flows
(Citation share)
Sources: Reliance on Science; United States Patent and Trademark Office; and IMF staff calculations.Note: Bars correspond to the country or region of the citing patent; legend items correspond to the country or region of the cited research article.Across Space
To harness this information, the chapter estimates a gravity-type model of international knowledge flows. The outcome variable is the number of citations from one country to another. For example, for basic research, this would be the number of citations by, say, Malaysian inventors to scientific articles with Spanish authors (for applied research, the citations are to other patents). The explanatory variables are: whether the two countries share a border, whether they have a common official language, how specialization in their economies differs (scientific specialization for science citations, technological for patent citations), and geographic distance in kilometers. Citing and cited country fixed effects capture differences in the knowledge mass, intellectual property rights, and other factors that may influence a country’s propensity to patent or to cite other patents. Further details are in Online Annex 3.2.
Panel 1 of Figure 3.4 shows the estimated cumulative impact of these various barriers, calculated separately for basic and applied knowledge. These show that basic knowledge diffuses more strongly than applied knowledge, with the red line staying above the blue line across most barriers. Country borders, lack of a common language, and specialization distance all present a larger impediment to the diffusion of applied knowledge. The marginal effect of geographic distance is negative for basic knowledge but insignificant for applied knowledge. Patent-to-patent citation intensity for applied knowledge is instead likely more dependent on other factors, such as tough competition. One example is the recent 5G technology race among China, the European Union, and the United States. However, the cumulative effect differs only over very long distances. These findings are unaffected by a variety of robustness checks, including controlling for cross-country differences in scientific and technological output, as detailed in Online Annex 3.2.
Diffusion of Basic and Applied Knowledge
Sources: PATSTAT; Reliance on Science; and IMF staff calculations.Note: In panel 1, the baseline knowledge flow equals 100 in the absence of barriers. In panel 2, the sample is restricted to patents applied for during 2010–19. Axis truncated at 50 years. Specialization distance is measured as one minus the uncentered correlation coefficient between the specialization vectors of country i and country j, where the vectors are the share of patents falling within internationally classified scientific/technological fields. km = kilometers. See Online Annexes 3.1 and 3.2 for details.This sort of exercise has a long history in the academic literature on international trade. Earlier attempts to adapt the framework to knowledge diffusion typically focused on applied knowledge flows using patent-to-patent citations.5 The extension to basic knowledge flows using patent-to-science citations is new. Predictions of the estimated models can also be used as a measure of how relevant knowledge in one country is for research elsewhere. This point is important for the empirical analysis of the production function for ideas, which uses this measure to create country-specific aggregate foreign knowledge stocks for each country (more on this later).
Over Time
Knowledge diffuses over time as well as across space. Panel 2 of Figure 3.4 illustrates this point, showing the density of the age of scientific articles (red line) and patents (blue line) cited by various patents. As such, they approximate the influence of basic and applied knowledge over the years. Basic knowledge displays a long-lasting impact, with the density for the age of cited scientific articles reaching a peak at about eight years versus three years for cited patents. This evidence suggests that scientific ideas can still be economically influential for long periods of time.6
Of course, using patent-induced knowledge flows to understand innovation drivers is subject to some caveats. Some research and development may have a direct impact on productivity without necessarily resulting in new patents, and new patent applications may be more reflective of strategic patenting practices than of authentic innovation. Yet, when using only patents filed in at least two distinct national offices—a likely control for these effects—the findings are similar (Online Annex Table 3.2.3).
Knowledge Stocks and the Production Function for Ideas
The empirical production function for ideas explains how the flow of new productive ideas—as captured by patents—depends on foreign and domestic applied and basic research stocks.
Given that these stocks are measures of research expenditure (that is, research inputs), they are true inputs to a production function. Domestic stocks are computed by summing past expenditures, with 10 percent annual depreciation. Construction of the foreign stocks follows Peri (2005). For each country, a weighted average of the domestic research stocks in all the other countries is calculated, with the weights determined by the gravity model presented in this chapter. For example, Mexico’s constructed foreign basic research stock puts weight on the United States that is proportional to the average Mexican inventor’s citations to science from the United States, as predicted by the determinants of the gravity model— geography, language, and technological mix. In this sense, construction of the data measures how accessible foreign research stocks are to a given country.
The estimated impact of research and development stocks on innovation is plotted in panel 1 of Figure 3.5. The main estimates use dynamic ordinary least squares, which efficiently utilize the cointegration of the data.7 The point estimates show the effect of a 1-percentage-point increase in the respective research stocks on the annual flow of patents, along with 95 percent confidence bands. For “own” basic research, the impact is 0.67 percentage point, and for applied research 0.77 percentage point, each having tight confidence bands. This suggests that domestic basic and applied research each have positive effects on patenting activity and are of similar magnitudes.
Estimated Ideas Production Function
Sources: PATSTAT; Penn World Table 10.0; Reliance on Science; World Bank; and IMF staff calculations.Note: Panel 1 shows the response of patent flows (log scale) to a 1-percentage-point change in each covariate (log scale) along with the 95 percent confidence interval. Panel 2 shows the additional estimated effect of research stocks on innovation in emerging markets. See Online Annex 3.3 for details. EMs = emerging markets; R&D = research and development.Foreign basic research also has a sizable effect, leading annual patent flows to increase 1.36 percentage points. In contrast, foreign applied knowledge has a negative estimated impact on patenting activity. However, this is very imprecisely estimated. Indeed, the magnitude of imprecision prohibits any confidence about even the direction of the true effect. That said, a negative impact of foreign applied research on domestic innovation is not completely implausible and would at least be consistent with the idea that some applied research and development leads to “business stealing” by competitors (as opposed to the nonrival and nonexcludable nature of foreign basic research; see Bloom, Schankerman, and Van Reenen 2013).8
Online Annex 3.3 shows the estimates of alternative specifications of the ideas production function. While the details vary, the estimates consistently reveal a strong and significant relationship between basic research and innovation and positive spillovers from foreign research (although the relative roles of foreign basic and applied research are not always as clear).
Box 3.2 extends this analysis to look at a particular type of innovation—clean technologies—and finds that basic research has larger green spillovers, suggesting that spending on basic research can play an important role in combating global climate change.
Differences in the Ideas Production Function: Advanced versus Emerging Market and Developing Economies
The estimates presented so far reflect those for an average economy in the data set. However, the estimated effects of basic and applied research stocks on innovations may differ by country. To get a sense of the size of these differences and what drives them, Figure 3.5 (panel 2) presents the estimated difference between advanced economies and emerging market and developing economies (see Table 3.3.2 in Online Annex 3.3). Two findings are apparent:
First, access to foreign research has a larger estimated effect on innovation in emerging markets than in advanced economies. This is true for both applied and basic research. Consistent with this difference, inventors from emerging markets are also less likely to cite homegrown research (Figure 3.5, panel 3). The results suggest that foreign technology adoption is more important for emerging markets than for advanced economies, consistent with the April 2018 World Economic Outlook. Learning-by-doing is one possible channel; adoption of foreign technologies (for example, through trade links; see Chuang 1998) may provide local workers the opportunity to learn new processes, forming the basis for innovation.
Second, evidence for the role of domestic research is mixed. While the estimated effect of applied research on innovation is not significantly different across emerging markets and advanced economies, basic research seems to play a larger role in emerging markets.9 It is possible that this reflects the larger impact of basic science in niche fields that receive less attention in advanced economies but may be relevant in emerging markets.
Overall, these results emphasize the importance of foreign knowledge for emerging market and developing economies. Although domestic basic research is more productive than for advanced economies in generating innovation, the effect is even larger for foreign research.
The Production Function for Goods and Services
Building on the estimates of the ideas production function presented earlier, this section examines the link between innovation and productivity.
The analysis relies on a production function for output and estimates the long-term relationship between productivity (real output per worker) and the country-specific stock of innovation.10 This is the empirical analogue of the production function for output in Figure 3.2.
In this setting, the stock of innovations is measured using cumulated annual flows of new patents, assuming an annual depreciation rate of 10 percent. The regression also takes in the usual factors of production, such as capital per worker and human capital, along with country and time fixed effects. Finally, the regression includes interactions between innovation and institutional factors to allow institutions to affect the transmission from innovation to productivity. Constant returns to scale are imposed, and the estimation uses data covering 138 countries during 1980–2017.11
The estimated relationship between innovation and productivity is strong and significant (Figure 3.6). An increase in the stock of patents by 1 percent is associated with an increase in productivity per worker of 0.04 percent,12 in line with estimates reported in Ulku (2004) and dependent on the institutional features of a country (Figure 3.6). The relationship is stronger for countries with higher financial development and more years of schooling, consistent with the idea that deeper financial markets and more educated workforces help transform innovation into productivity. Together with the findings on strong spillovers from foreign research (Figure 3.5, panel 2), these findings are relevant for emerging market and developing markets, as these results suggest that financial market and educational reforms can allow countries to better absorb the stock of foreign research.
Estimated Output Production Function
Sources: PATSTAT; Penn World Table 10.0; Reliance on Science; World Bank; and IMF staff calculations.Note: Patent stock shows the estimated effect of a 1 percent increase in the stock of patents on productivity. The other coefficients show the additional estimated effect (estimated in separate equations) of innovation on productivity from moving from the middle to the upper tercile of countries in financial development and years of schooling, respectively. See Online Annex 3.4 for details.Putting It All Together
This section combines the exercises of the previous sections to trace the path to the final impact of increases in basic research stocks on productivity.
Specifically, Figure 3.7 (panel 1) shows that the estimated effect of a 10 percent permanent increase in the stock of a country’s own basic research is to increase productivity by 0.30 percent, while a similar increase in the stock of foreign basic research is estimated to have a larger impact, increasing productivity by about 0.6 percent. The impact on productivity of own applied research is estimated to be of the same order as the impact of own basic research, and international spillovers are insignificant. The differences are driven by the respective elasticities estimated from the production function for ideas (Figure 3.5).
Implications of the Empirical Findings
Source: IMF staff calculations.Note: Panel 1 shows the estimated effect of a permanent 10 percent increase in research stocks on real GDP per worker. An estimated elasticity of 0.674/1.358 for patents with respect to own basic research/foreign basic research is used. An estimated elasticity of 0.044 for productivity with respect to the stock of patents is used. Panel 2 shows the estimated effect on global innovation (measured as flow of new patents) and productivity of a given reduction (in percent) in citations between the United States and China. See Online Annex 3.5 for details.Overall, the evidence suggests that international productivity spillovers are significant, particularly from basic research. This is in line with the earlier evidence on the extent of international spillovers in Figure 3.4, which also suggests that basic knowledge diffuses more widely and for a longer time than applied knowledge. Hence, the type of research does seem to matter for productivity growth. Quantitatively, however, large confidence bands around those estimates suggest caution in interpreting these results, especially on the impact of foreign research (Figure 3.5). In addition, the linear regression approach measures only the direct effect of basic research on innovation and productivity growth. The true effect may be even larger due to nonlinear relationships linking applied research to the stock of basic knowledge.13
Policy Experiment: Scientific Decoupling between the United States and China
In recent years, concern has been growing that rising tensions between China and the United States could lead to technological decoupling, with detrimental effects on innovation capacity and growth at the global level. This section uses the empirical framework described in this chapter to do a back-of-the-envelope calculation of the cost for global innovation of increased scientific decoupling between the two countries.
The empirical framework can be used to model scientific decoupling, implemented as a reduction in the citation intensity between the two countries. This reduces the foreign stock of basic research available to each country, which in turn decreases innovation and productivity. This is consistent with, for example, differences in technology standards inducing changes across the two countries, such that research done in one becomes less relevant for the other. Limits on knowledge flows might also arise if ongoing geopolitical tensions make it harder for researchers in the two countries to interact or work together. For instance, restriction on travel might prohibit the all-important personal contacts that can occur at seminars, conferences, and the like.
Figure 3.7 shows the estimated impact on global innovation as measured by the annual flow of new patents for various degrees of scientific decoupling. As a purely illustrative example, full decoupling, as modeled by citations between the two countries shrinking to zero, is estimated to reduce global patent flows by 4.4 percent and global productivity by 0.8 percent.14
These estimates are likely a lower bound of the impact of decoupling, for two reasons. First, they assume that only foreign stocks of basic research, innovation, and productivity for the United States and China are affected in a decoupling scenario. In reality, stocks in other countries are likely to be affected too, creating an extra dimension to the shock. Second, these estimates are partial insofar as they do not include any general equilibrium effects that could affect the impact of the initial shock on global innovation and productivity. Given the evidence presented previously on the magnitude of global basic research spillovers, these could be substantial.15
Policy Analysis
Earlier sections established the empirical links between basic research, innovation, and economic activity. This raises an obvious question: how can public policy best exploit these links to boost living standards? An important aspect of this empirical work is that it measures only the direct part of these links, holding all else fixed. But in reality, many indirect channels exist. For instance, policies that boost basic science spill over to increase returns to applied innovation, and changes in productivity feed back into wages, driving demand and influencing research incentives. To assess the impact of policy, a framework articulating these links is required.
The Model
Recent work by Akcigit, Hanley, and Serrano-Velarde (2021) provides a theoretical framework for answering this question. It analyzes a setting in which firms conduct two types of research: basic, which builds the stock of knowledge; and applied, which converts knowledge into products. These correspond closely to the basic and applied expenditure concepts used in the empirical analysis. The government has three policy levers: subsidies for each of the two types of research; and direct funding for public basic research, such as universities and public research labs.
The key feature of this approach is that basic research is modeled as having applications in many different fields. This captures an essential aspect of basic research—that, because individual firms typically operate in only a few sectors, they cannot profit fully from the range of economic applications opened up by the most fundamental and basic discoveries. As a result, private incentives for basic research are outstripped by its social benefits. Without a public policy response, this will result in inefficiently low levels of innovation and productivity.
Despite the special character of basic research, it is not the only potential target of public policy in this framework. Applied research—which is complementary to basic research, adapting knowledge to produce marketable products—also generates spillovers, which could also motivate public support. This is because innovations that bring a product to market can be superseded by competitors’ innovations. This introduces a “quality ladder” mechanism: firms may not be able to fully internalize the social value of applied innovation, leading to underprovision of applied research as well. Whether applied or basic research is more desirable is not hardwired into the model but is instead a function of parameters estimated from the data.
The model is estimated for three countries: France, the United Kingdom, and the United States. Although estimating for more countries would be ideal, the data requirements needed to maintain the important distinction between basic and applied research preclude this. Still, this exercise gives some sense of the impact of country-specific factors, at least within advanced economies.
Optimal Policies
Figure 3.8 shows optimal policies and the resultant outcomes from several experiments. The first, shown in red, is the case when governments cannot subsidize applied and basic research separately and, so, must apply the same rate to both. This is not an unreasonable approximation of reality, as deciding which of firms’ individual activities are “applied” and which are “basic” is often challenging and, so, being able to target them separately may be difficult. Indeed, many data sources for such subsidies cannot make this distinction.
Optimal Policy
Sources: Organisation for Economic Co-operation and Development; and IMF staff calculations.Note: Range shows optimal policies across the model reestimated for France, the United Kingdom, and the United States. In the differential subsidies case, public research is assumed fixed at the level in the data. See Online Annex 3.6 for details.This exercise suggests that research, in general, is funded below its socially optimal level. Subsidy rates for private research should be doubled, and public research expenditure increased by about one-third. Although country-specific caveats (see “Policy Conclusions” below) might caution against a too-literal interpretation of these findings, they are at least broadly supportive of the notion that there are likely underexploited spillovers from research that can leave room for policy to make households better off. Increasing subsidies and public research expenditures as recommended would raise productivity growth in the order of about 0.2 percentage point a year. This would start to pay for itself within about a decade. If applied over the period shown in panel 1 of Figure 3.1, this would have resulted in current per capita incomes about 12 percent higher than in the data. Moreover, in an era of low real interest rates, small increases in economic growth can have very large impacts on debt sustainability.
Under this policy program, the stocks of both applied and basic knowledge increase. But because public expenditure is purely basic, the stock of basic knowledge increases by more—with an increase about several times the size of that for applied knowledge. This increase in the knowledge stock also varies across countries and is largest in the United States, where higher corporate entry and exit rates mean that firms do not internalize the social benefits of research, leaving more room for policy to play a positive role. The level of wages also rises under optimal policy, with increases of between 2.5 percent and 3 percent, depending on the country.
Of course, assuming that no scope exists for targeting subsidies might seem somewhat restrictive and, so, the results of separately subsidizing applied and basic research are also shown in Figure 3.8, in yellow. This policy clearly dominates the previous one, which implies that, where possible, governments should target subsidies aggressively toward basic research. This policy recommendation matches the earlier empirical evidence, which shows that basic research is an important determinant of productivity growth.
Although targeting has only a minor additional impact on growth, it reduces the cost of subsidies, lowering taxes and making households substantially better of. The intuition for this is that basic research is a smaller sector than applied research. Given that the subsidy is smaller, and growth spillovers from basic research are larger than for applied research, this achieves a similar growth effect but with a much smaller subsidy. Lower subsidy spending can translate into lower taxes, boosting household disposable income and consumption permanently.
Exploring the Assumptions
As with any model-based analysis, the results depend on the modeling assumptions. Here, two important assumptions are explored in detail.
The first is the substitutability of public and private research. In the baseline, this substitutability is imperfect; public research requires extra work to be useful for commercial innovation—the “ivory tower” effect. If this is turned of, public basic research can be commercialized more easily and can take on more of the qualities of a public-private partnership.
The most obvious effect of this experiment is that optimal research expenditure increases considerably, to about 3 percent of GDP (Figure 3.8, panel 2, in green). This is not surprising: a public sector that can deliver more commercially adaptable innovations means better use of resources. Optimal subsidies fall, and growth increases by an average of another 0.1 percentage point. The policy implication is that, even if discrimination between basic and applied research subsidies is not possible, governments might be able to achieve something similar by encouraging greater collaboration between public and private basic researchers.
The second experiment investigates how sensitive these results are to assumptions about private basic research spillovers. It is conceivable that spillovers from private firms may decrease if, for example, recent technological change allows for more market power or other abilities to privatize breakthroughs. To proxy this, the blue bars in Figure 3.8 show what happens if the spillovers from private basic research shrink by one-quarter. This limits public gains from research and, so, optimal public subsidy rates are increased only by half relative to the data (versus doubling in the baseline).
Policy Conclusions
The preceding experiments highlight four key policy lessons.
First, public funding for research is too low. Gains can be made from both subsidizing more private research and doing more public research.
Second, the ability to discriminate among various types of research is very valuable. If possible, governments could achieve similar outcomes to the baseline at roughly half the cost.
Third, better connections between public and private researchers might be able to substitute for targeted subsidies, which can be hard to implement.
Fourth, regarding firms’ ability to protect their discoveries, if basic research spillovers decline, then the social gains from research will fall. This suggests that reducing overbearing market power or excessively broad patenting can boost productivity and growth (Box 3.3 discusses this issue more broadly).
As with any model-based analysis, tractability demands that this assessment leave out a number of other factors that could affect the policy conclusions. As such, these conclusions should be treated as a baseline, from which country-specific considerations could require some deviation.
One such issue is the absence of distorting taxation. In this setting, taxes are raised by collecting a lump sum from households. In reality, though, most tax instruments, such as labor or capital taxes, induce some sort of inefficiency. Such instruments introduce an extra cost to policy interventions. Because these costs typically increase with the size of the tax, countries with high tax distortions may find policies to support basic research to be more costly. A similar caveat applies to countries with high debt burdens or inefficient revenue collection systems. In these cases, a better source of funding might be to reprioritize expenditure or improve revenue mobilization.
Moreover, these policy conclusions are perhaps most directly relevant to advanced economies: the model lacks a channel (such as trade) for the international diffusion of knowledge, which earlier sections show to be important in emerging market and developing economies. As such, these countries may find that policies to better adapt foreign knowledge to local conditions are a better avenue for development than investing directly in homegrown basic research (Acemoglu, Aghion, and Zilibotti 2006). Other unmodeled factors, such as political constraints, may also hinder the kind of tax-funded innovation-boosting policies presented here.
Conclusions: Investment in Basic Science Boosts Productivity and Pays for Itself over the Long Term
The development of COVID-19 mRNA vaccines acts as a stark reminder of the importance of science for innovation and growth. In common with other technological breakthroughs, past scientific discoveries in unrelated fields typically laid the foundation for today’s technological advances, driving future productivity and economic growth (Box 3.1).
Improving growth outcomes will be essential to post-pandemic economies, helping finance higher public debt and additional post-pandemic social expenditures. It is therefore worrisome that the share of basic research has been steadily declining over the past three decades.
That the private sector underinvests in basic research is not surprising. As shown in this chapter, the benefits of basic research are diffuse and long-lasting, making it an unattractive proposition for private firms. This creates an opportunity for policy intervention. The chapter shows that doubling subsidies to private research and boosting public research expenditure by one-third could increase annual growth per capita by around 0.2 percent. Better targeting of subsidies and closer public-private cooperation could boost this further, at lower public expense. Such investments could start to pay for themselves within a decade or so.
The chapter also shows that scientific knowledge travels far over time and distance and that it is a key driver of innovation in both advanced economies and emerging markets. Spillovers from advanced economies to emerging markets are particularly large. Deep financial markets and better educational systems are key facilitators for cross-border technology adoption.
It is also important to ensure the free flow of ideas and scientific collaboration across borders, especially for emerging markets. The technological trajectories of China and the United States have been closely linked in the past two decades. Rising political tensions could lead to scientific decoupling, with detrimental effects on innovation capacity and global economic growth.
Beyond its impact on growth, basic science is likely to be a key contributor to a greener future. The fight against climate change requires drastic cuts in global emissions. New clean technologies will be central to this effort. Evidence presented in this chapter suggests that investment in frontier science—especially in natural sciences and engineering—could help speed the transition toward a cleaner economy.
mRNA Vaccines and the Role of Basic Scientific Research
Vaccines using new mRNA technology are key to the fight against COVID-19; the most well-known are those developed by Pfzer/BioNTech and Moderna.1 This technology uses genetic code known as messenger RNA (or mRNA) to instruct human cells to make part of the virus’s protective shell. These fragments help train the body’s immune system to attack the real virus. Compared with conventional approaches, mRNA technology can deliver better-performing vaccines with shorter research and production times. Their social and economic impact has been enormous, likely shortening the pandemic by years, and looks set to revolutionize medical treatments in years to come.
This technology was built on waves of prior scientific discoveries. To track these discoveries, Figure 3.1.1 shows the publication dates of scientific articles cited by five of the seven Moderna COVID-19 vaccine patents (in blue). This distribution captures the direct dependence of vaccine development on past scientific discoveries and is concentrated around breakthroughs on the function of mRNA in the early 2010s. To measure the indirect influence of science, the yellow line shows the scientific citations of the vaccine’s “parent” patents—other patents referenced in the five original vaccine patents. These peak in the early 2000s, tracking discoveries in editing genetic codes. Earlier advances in reading genetic codes drove a similar wave of citations from “grandparent” patents in the early 1990s. These waves of scientific influence illustrate how policies that help incentivize advances in basic science today influence the building blocks of future technologies and yield long-lasting economic payoffs.
mRNA Technology Was Built on Waves of Previous Scientific Discoveries
(Percent of citations)
Sources: Moderna; Reliance on Science; United States Patent and Trademark Office; and IMF staff calculations. Note: The y-axis shows the scientific citations by Moderna’s mRNA patents and their ancestors. Parent patents are those cited by Moderna’s mRNA vaccine patents. Grandparents are those cited by parent patents.Developing mRNA vaccines relied on a broad base of scientific knowledge. On average, the Moderna vaccine patents are in the same technological category as only 55 percent of their parent patents—a number that falls further as citation chains lengthen (Figure 3.1.2). This shows how wide-ranging basic science contributed to mRNA vaccines, indicating that policies to develop a broad scientific base can pay off in many and unexpected ways.
mRNA Vaccines Relied on a Broad Base of Scientific Knowledge
(Percent)
Sources: Moderna; United States Patent and Trademark Office; and IMF staff calculations.Note: The y-axis shows the fraction of patents in the same technological categories as the seven Moderna vaccine patents. The blue line is the averaged percentage for each ancestor. The shaded area shows the range of each ancestor of citation across the seven Moderna vaccine patents. Total number of categories is 7,523 based on the International Patent Classification. Parent patents are those cited by Moderna’s mRNA vaccine patents. Grandparents are those cited by parent patents. Great-grandparents are those cited by grandparent patents.The development of COVID-19 vaccines was encouraged by unprecedented public support. This included regulatory forbearance (emergency use authorization of COVID-19 vaccines), at-risk up-front investment and subsidies for vaccine production (Operation Warp Speed), help in scaling up manufacturing (Indian government grants to vaccine producers), joint licensing agreements with local producers (India, South Africa), and advance public purchase commitments (Israel, United Kingdom, United States). A distinguishing feature of public support for a COVID-19 vaccine was its continuation throughout the development process. Typically, public funding is most generous for early trials, falling as products near market. For COVID-19 vaccines, public and academic funding for clinical trials stayed high, even at the latest stages of development (Figure 3.1.3). This highlights how support throughout the production process can incentivize research by forward-looking firms.
Unprecedented Public Support for COVID-19 Vaccine Clinical Trials
Sources: US National Library of Medicine; and IMF staff calculations.Note: The y-axis shows the fraction of clinical trials with no private support. The three bars on the left show the clinical trial data for the COVID-19 vaccine. Support may include activities related to funding, design, implementation, data analysis, or reporting. Funder type is defined as private if support comes only from organizations in industry. Phases are based on the US Food and Drug Administration definition.Global distribution of vaccines remains a challenge. Although reliable data are hard to come by, global supply seems sufficient. World production of COVID-19 vaccines is likely to hit almost two doses per capita by the end of 2021—slightly less than demand. Although supply disruptions and capacity constraints can hamper delivery of vaccines, even planned purchases are unevenly distributed, with outsized demand in the United States and Europe (Figure 3.1.4). Fair distribution of vaccines will require adjustment of planned allocations, irrespective of where they are produced.
Global Distribution of Vaccines Remains a Key Policy Challenge
Sources: Duke Global Health Innovation Center; Our World in Data; and IMF staff calculations.Note: Blue bars show the planned doses of manufactured vaccines by region by the end of 2021, which also includes doses in contracts under discussion. Red bars show the number of administered vaccines by region. Yellow bars show the differences between the number of planned purchase of vaccines by the end of 2021 and the number administered. Other Americas = Americas excluding the United States; Other Asia = Asia excluding China and India.Clean Tech and the Role of Basic Scientific Research
Avoiding catastrophic climate change requires a rapid reduction in emissions of greenhouse gases. This will be possible only if global energy consumption transitions to predominantly clean (zero carbon emissions) energy sources. Technological advances to drive down the cost of clean energy are a key part of any strategy to minimize the economic impact of that switch. This box shows how investment in basic research is especially important to foster innovation in clean technologies and thus spur emission reductions.
This question is addressed using the patent-level Reliance on Science data set. This includes detailed information on the industrial category of its constituent patents, which is used to classify the technology covered in each patent as a clean or a dirty innovation (following Dechezleprêtre, Muckley, and Neelakantan 2020). Clean innovations include renewable energy technology and electric vehicles; dirty innovations cover gas turbines, furnaces, and the like. Comparing the properties of clean and dirty innovations against all other patents (as a benchmark) can help uncover the relationship between scientific research and the direction of technical change.1
The first dimension for comparing clean and dirty patents is their relative citations to prior patents and scientific articles. This contains information on how various types of innovation depend on applied and basic knowledge stocks. Figure 3.2.1 summarizes the results of this exercise. The first panel shows that both clean and dirty innovations cite less prior research than other sorts of innovation. Clean innovations cite more research than dirty innovations, but mainly within scientific articles. With a sample of several million patents, these differences are very precisely estimated.
Clean Innovation Relies Relatively More on Basic and Newer Research
Sources: Reliance on Science; United States Patent and Trademark Office; and IMF staff calculations.Note: Panel 1 (panel 2) shows coefficients from regression of citations (citation lag) on dummies for patent type, year, and country of inventor. Error bars represent 95 percent confidence intervals. Because the sample is very large, confidence intervals are sometimes so small as to be narrower than the width of the marker for the point estimate.The second panel compares the age of the research used by clean and dirty innovation, which can be thought of as a proxy for distance to the technological frontier. Clean innovations cite newer patents and scientific articles than both dirty innovations and other types of innovations. However, the difference is largest for scientific articles, which are, on average, 0.8 years newer than those cited by dirty innovation. In other words, clean breakthroughs rely more on scientific research closer to the frontier than dirty innovation.
Figure 3.2.2 shows the fraction of scientific research in various fields, relative to other patents. It shows that clean innovation is particularly likely to rely on research in engineering and technology and unlikely to rely on medical research. Interestingly, dirty innovations cite the natural sciences much less frequently than do clean ones. Unsurprisingly, neither clean nor dirty innovation seems to depend much on research in agriculture, social science, or the humanities.
Clean Innovation, in Particular, Cites Engineering and Technology
(Fraction of citations; difference relative to other patents)
Sources: Reliance on Science; United States Patent and Trademark Office; and IMF staff calculations.Note: Figure shows coefficients from regression of research field dummies on dummies for patent type. Error bars represent 95 percent confidence intervals. Because the sample is very large, confidence intervals are sometimes so small as to be narrower than the width of the marker for the point estimate.Overall, the evidence presented here suggests that clean innovations depend more than dirty ones on frontier science, particularly natural sciences and engineering. Accordingly, basic research investment in these fields is likely to have a positive impact in the fight against climate change. That said, public promotion of basic research in these fields will be only part of the solution. Other factors, such as incentives to bring new clean technologies to market, as well as addressing stranded assets associated with dirty fuels, will also be important.
The authors of this box are Philip Barrett and Niels-Jakob Hansen. 1 This comparison is done via regression, allowing for results that account for third factors that might otherwise influence this relationship. This includes the year that the patent is issued and the country of the inventor.Intellectual Property, Competition, and Innovation
Intellectual property rights are among several public policy tools to foster private innovation. Innovation requires costly and risky up-front investments in research and development. Thus, would-be innovating firms may undertake them only with some guarantee that their ideas can be protected from potential imitators, at least for some time. Intellectual property rights are designed to do just that. By granting temporary monopoly power to inventors, intellectual property rights make it profitable to invest in research and development and incentivize a continuous flow of innovation. Strong intellectual property rights also complement growth-enhancing pro-competition policies, such as reduced market entry barriers and tougher antitrust frameworks (Aghion, Howitt, and Prantl 2015). Competition is generally good for innovation but, when too strong, it can weaken firms’ prospective monopoly rents and therefore their incentive to innovate (April 2019 World Economic Outlook; IMF 2021), unless these future rents are well protected by patent laws.
However, there is a limit to how strong intellectual property rights should be. If overly protective they can cement leading firms’ position and weaken their incentive to innovate, discouraging lagging firms from doing so as well (Akcigit and Ates 2021). This is particularly likely if patents excessively reward incremental innovations, or if market leaders use them as barriers to competition. “Patent thickets”—overly complicated legal setups that require a firm to seek agreements with many parties to use a technology—are an example (Shapiro 2001).
In sum, intellectual property rights should be neither too weak nor too strong and they should reward disruptive innovations far more than those that are incremental. Yet, even when well calibrated, intellectual property rights confer temporary monopoly power, which delays the widespread dissemination of innovation to competitors and the general public. This could, at times, run counter to society’s broader goals. In a pandemic, for example, any delay in widespread vaccine production has enormous human and economic costs. Therefore, during a public emergency, and when the use of a targeted innovation is clearly identified, governments should consider alternative, less distortive approaches. Tax credits for specific research and development, direct government support, and innovation prizes, in particular, have been proposed in such situations (Kremer and Williams 2010; Maskin 2020). These policies better align society’s goals with private incentives when the targeted innovation (for example, a new vaccine) and success criteria (such as effectiveness and safety) are well identified.
By covering costs and risks up front, Operation Warp Speed generated the necessary incentives for pharmaceutical companies to develop effective vaccines in record time. Intellectual property rights also likely helped stimulate the development of vaccines, but at the risk of slowing global production in the near future. In response, a proposal—supported by China, Russia, and the United States—to temporarily waive these rights for vaccines is currently under discussion at the World Trade Organization. In future pandemics, alternative policy support, such as well-designed innovation prizes, could be considered, which would stimulate vaccine development just as powerfully while also facilitating rapid vaccine dissemination.
The authors of this box are Romain Duval and Jean-Marc Natal.
References
Acemoglu, Daron, Philippe Aghion, and Fabrizio Zilibotti. 2006. “Distance to Frontier, Selection, and Economic Growth.” Journal of the European Economic Association 4 (1): 37–74.
Aghion, Philippe, and Peter Howitt. 1992. “A Model of Growth through Creative Destruction.” Econometrica 60 (2): 323–51.
Aghion, Philippe, Nicholas Bloom, Richard Blundell, and Rachel Grifth. 2005. “Competition and Innovation: An Inverted-U Relationship.” The Quarterly Journal of Economics, 120 (2): 701–28.
Aghion, Philippe, Peter Howitt, and Susanne Prantl. 2015. “Patent Rights, Product Market Reforms, and Innovation.” Journal of Economic Growth 20 (3): 223–62.
Aghion, Philippe, Ufuk Akcigit, and Peter Howitt. 2013. “What Do We Learn from Schumpeterian Growth Theory?” In Handbook of Economic Growth 2, edited by Philippe Aghion and Steven Durlauf. Amsterdam: North-Holland.
Ahmadpoor, Mohammad, and Benjamin F. Jones. 2017. “The Dual Frontier: Patented Inventions and Prior Scientifc Advance.” Science 357 (6351): 583–87.
Akcigit, Ufuk, and Sina T. Ates. 2021. “Ten Facts on Declining Business Dynamism and Lessons from Endogenous Growth Theory.” American Economic Journal: Macroeconomics 13 (1): 257–98.
Akcigit, Ufuk, Wenjie Chen, Federico J. Diez, Romain Duval, Philipp Engler, Jiayue Fan, Chiara Maggi, Marina Mendes Tavares, Daniel A Schwartz, Ippei Shibata, and Carolina Villegas-Sánchez. 2021. “Rising Corporate Market Power: Emerging Policy Issues.” IMF Staff Discussion Note 2021/001, International Monetary Fund, Washington, DC.
Akcigit, Ufuk, Douglas Hanley, and Nicolas Serrano-Velarde. 2021. “Back to Basics: Basic Research Spillovers, Innovation Policy, and Growth.” The Review of Economic Studies 88 (1): 1–43.
Akcigit, Ufuk, and William R. Kerr. 2018. “Growth through Heterogeneous Innovations.” Journal of Political Economy 126 (4): 1374–443.
Belenzon, Sharon, and Mark Schankerman. 2013. “Spreading the Word: Geography, Policy, and Knowledge Spillovers.” The Review of Economics and Statistics 95 (3): 884–903.
Bloom, Nicholas, Mark Schankerman, and John Van Reenen. 2013. “Identifying Technology Spillovers and Product Market Rivalry.” Econometrica 81 (4): 1347–93.
Blundell, Richard, and Stephen Bond. 1998. “Initial Conditions and Moment Restrictions in Dynamic Panel Data Models.” Journal of Econometrics 87 (1): 115–43.
Byrne, John G. 1995. “Changes on the Frontier of Intellectual Property Law: An Overview of the Changes Required by GATT.” Duquesne Law Review 34 (1): 121.
Cass, David. 1965. “Optimum Growth in an Aggregative Model of Capital Accumulation.” The Review of Economic Studies 32: 233–40.
Cerdeiro, Diego A., Johannes Eugster, Rui C. Mano, Dirk Muir, and Shanaka J. Peiris. 2021. “Sizing Up the Effects of Technological Decoupling.” IMF Working Paper 21/69, International Monetary Fund, Washington, DC.
Chuang, Yih-Chyi. 1998. “Learning by Doing, the Technology Gap, and Growth.” International Economic Review 39 (3): 697–721.
Dechezleprêtre, Antoine, Cal B. Muckley, and Parvati Neelakantan. 2020. “Is Firm-Level Clean or Dirty Innovation Valued More?” The European Journal of Finance, July 2. doi: 10.1080/1351847X.2020.1785520.
Etzkowitz, Henry, and Loet Leydesdorf. 2000. “The Dynamics of Innovation: From National Systems and ’Mode 2’ to a Triple Helix of University–Industry–Government Relations.” Research Policy 29 (2): 109–23.
Grossman, Gene M., and Elhanan Helpman. 1991. Innovation and Growth in the Global Economy. Cambridge, MA: MIT Press.
Jafe, Adam B., Manuel Trajtenberg, and Rebecca Henderson. 1993. “Geographic Localization of Knowledge Spillovers as Evidenced by Patent Citations.” The Quarterly Journal of Economics 108 (3): 577–98.
Koopmans, Tjalling C. 1965. “On the Concept of Optimal Economic Growth.” In Study Week on the Econometric Approach to Development Planning, 225–87. Amsterdam: North-Holland.
Kremer, Michael, and Heidi Williams. 2010. “Incentivizing Innovation: Adding to the Toolkit.” In Innovation Policy and the Economy 10. Cambridge, MA: National Bureau of Economic Research.
Mankiw, Gregory N., David Romer, and David N. Weil. 1992. “A Contribution to the Empirics of Economic Growth.” The Quarterly Journal of Economics 107 (2): 407–37.
Marx, Matt, and Aaron Fuegi. 2020. “Reliance on Science: Worldwide Front-Page Patent Citations to Scientifc Articles.” Strategic Management Journal 41 (9): 1572–94.
Maskin, Eric. 2020. “Mechanism Design for Pandemics.” Talk at the Santa Fe Institute Webinar on the Complexity of COVID-19, April 14.
Nelson, Richard R. 1959. “The Simple Economics of Basic Scientifc Research.” Journal of Political Economy 67 (3): 297–306.
Peri, Giovanni. 2005. “Determinants of Knowledge Flows and Their Effect on Innovation.” The Review of Economics and Statistics 87 (2): 308–22.
Ramsey, Frank. 1928. “A Mathematical Theory of Saving.” The Economic Journal 38 (152): 543–59.
Romer, Paul. 1990. “Endogenous Technological Change.” Journal of Political Economy 98 (5): S71–S102.
Santos Silva, João, and Silvana Tenreyro. 2006. “The Log of Gravity.” The Review of Economics and Statistics 88 (4): 641–58.
Shapiro, Carl. 2001. “Navigating the Patent Ticket: Cross Licenses, Patent Pools, and Standard Setting.” Innovation Policy and the Economy (1): 119–50.
Solow, Robert M. 1956. “A Contribution to the Theory of Economic Growth.” The Quarterly Journal of Economics 70 (1): 65–94.
Ulku, Hulya. 2004. “R&D, Innovation, and Economic Growth: An Empirical Analysis.” IMF Working Paper 04/185, International Monetary Fund, Washington, DC.
As opposed to population growth or capital accumulation; see Ramsey (1928), Solow (1956), Cass (1965), and Koopmans (1965).
See the April 2018 World Economic Outlook; Grossman and Helpman (1991); Aghion and Howitt (1992); Mankiw, Romer, and Weil (1992); and Aghion and others (2005).
All annexes are available at www.imf.org/en/Publications/WEO.
Ahmadpoor and Jones (2017) gives examples of how the two types of research mutually reinforce their role in innovation.
The spatial diffusion of knowledge spillovers using patent data has been widely studied, starting with Jafe, Trajtenberg, and Henderson (1993). See Peri (2005) for a more recent example. While advances in communication have improved accessibility to scientific articles, there is still evidence of the localization of scientific knowledge (for example, Belenzon and Schankerman 2013), partly explained by national policies aimed at fostering collaboration among local universities, firms, and government funding agencies (Etzkowitz and Leydesdorf 2000).
A back-of-the-envelope calculation of tail decay rates reveals that, in the long term, basic (applied) knowledge decays at 7 (11) percent annually.
See column (7) in Table 3.3.1 in Online Annex 3.3.
Note that foreign research stocks are an order of magnitude larger than domestic stocks and even larger for emerging market and developing economies. This affects the interpretation of the estimated coefficients: a 1-percentage-point increase in foreign research is a much larger change in the total knowledge. Further, the results in panel 1 of Figure 3.5 are robust to the exclusion of the United States (as a key driver of the technological frontier) from the sample.
Note, however, that the coefficient becomes insignificant (although still positive) when China is excluded from the sample (See Online Annex 3.3).
See also Ulku (2004) for a similar exercise.
Online Annex 3.4 reports the full econometric specification and details on the analysis.
Results from alternative specifications in Online Annex 3.4 show this to be robust to averaging over multiyear intervals, which is strongly suggestive of a long-term relationship.
See the “Policy Analysis” section for general equilibrium effects of policies stimulating basic research.
Online Annex 3.5 provides further details and a full breakdown of these effects.
See Cerdeiro and others (2021) for a more structural approach to the decoupling issue.
Chapter 3: Research and Innovation: Fighting the Pandemic and Boosting Long-Term Growth
Chapter 3: Research and Innovation: Fighting the Pandemic and Boosting Long-Term Growth: Data
