Innovation is widely seen as central to the growth of developing countries, and available evidence suggests that the returns to R&D investment should be extremely high. Yet low-income countries invest very little. We propose an explanation for this paradox that has significant implications for theories of economic convergence, how innovation performance is benchmarked and targeted, and the conception and scope of the national innovation system.
Defined as the introduction of new products, technologies, business processes, as well as the invention of new ideas, innovation drives Schumpeter’s creative destruction process which underlies modern growth theory, and is the critical ingredient in historical accounts of how countries achieve prosperity. In turn, the radiation of ideas, products, and technologies to developing countries represents an externality of truly historic proportions. In the long run, productivity improvements can account for half of GDP growth (Easterly and Levine 2001), with adoption of technologies making up a sizable share of those. This implies that the wealth transfers from North to South are in the hundreds of billions of dollars per year, dwarfing international aid flows.
Such technological catch-up, however, requires investment in absorptive capacity broadly conceived, which is frequently proxied by R&D spending (Cohen and Leventhal 1990). Estimates of social returns to R&D for the advanced countries are above 40% (most recently, Lucking et al. 2017, Doraszelski and Jaumandreu 2013), and Griffith et al. (2004) further confirm, based on an OECD sample, that returns in fact rise with distance from the technological frontier reflecting the gains to catch up. Simple out-of-sample predictions for low-income countries suggest that their returns to R&D would reach well into the triple digits. To paraphrase Lucas, given these returns, it would be hard for policymakers to think of investing in anything else.
Yet, poor countries invest far less in innovation than rich countries, not only in R&D (Figure 1) but in virtually every type – technology licensing, managerial technologies, training, etc. Developing country firms and governments appear to be leaving billions of dollars on the table, uncollected. Indeed, Pritchett (1997), among others, documents a ‘Great Divergence’ of the last two centuries where, instead of poor countries catching up, with few exceptions rich countries continue to pull ahead. Comin and Hobijn (2004) and Comin and Ferrer (2013) argue that it is precisely the differences in the rate of adoption of new technologies that drives the magnitude of this Great Divergence. This counterintuitive juxtaposition of high returns and low investment motivates our recent volume, The Innovation Paradox: Developing Countries and the Unrealized Promise of Technological Catch up (Cirera and Maloney 2017).
Figure 1 Investment in R&D as a share of GDP versus GDP per capita
Note: Country level data on Gross domestic expenditure on research and development (GERD1966-2010) from the series are constructed combining data published by UNESCO, the OECD, the Ibero American Science and Technology Indicators Network (RICYT) and the Taiwan Statistical Data Book, following the definitions convened in the OECD Frascati Manual.
One of the background papers for the report points to a possible resolution of the paradox of high returns and low investment and provides evidence that, in fact, low-income countries don’t see such stellar returns (Goñi and Maloney 2017). Using country-level panel data and in the spirit of Fan and Zhang (1999) and Cai et al. (2006), who accommodate endogeneity in the conditioning set, we derive estimates of a standard production function, but allow all parameters to vary across the development process. The results suggest that while the contributions of physical and human capital-augmented labour are approximately constant across the development process, the returns to R&D trace an inverted U-shape relationship (Figure 2). The right vertical axis represents the technological frontier and income per capita falls as we move left along the horizontal axis. Using both internal lags as well as a measure of intellectual property rights as instruments, we confirm Griffith et al.’s (2004) finding of increasing returns to R&D up until about the income level of modern Argentina. However, as per capita income continues to fall, returns to R&D investment sharply decline with further distance from the frontier, including potentially negative values being for the countries at the very bottom of the income distribution.
Figure 2 Returns to R&D trace an inverted U-shape across the development process
Note: Figure uses quinquennials of cross-country data from 1960 to 2010 to estimate the rates of return to research and development (R&D) across the development process: 0 is the income frontier, and moving left represents progressively less developed countries. The points do not correspond to particular countries, but rather represent the average return at that distance from the frontier.
Source: Goñi and Maloney (2017).
Our proposed explanation for this pattern is analogous to that of Lucas (1990) when he asked “Why Doesn’t Capital Flow from Rich to Poor Countries”, namely, the increasing scarcity of a wide array of complementary factors. Innovation or technology adoption is not a free-floating activity, but can be thought of as the accumulation of knowledge capital, decisions around which will be made jointly with accumulation decisions of other types of capital or factors. Shortages of these factors will lead to low expected returns to innovation. For example, government subsidies to R&D in the absence of high level technical human capital will yield little return; good ideas generated by a university or think tanks, if not picked up by firms with high levels of managerial capability1 will not generate value added; and an entrepreneur with a good idea, but who cannot access credit markets, import necessary complementary machinery, or find the necessary qualified workers, will find that idea of limited value. As we show, each of these complementary factors – quality of research, managerial quality, physical and human capital – become increasingly scarce with distance from the frontier, arguably eventually offsetting Gershenkron’s (1962) advantages of backwardness.
These findings are important for several reasons.
- They provide a mechanism to explain lack of convergence of low-income countries to the technological frontier. These differential returns provide a new mechanism to explain multiple income convergence clubs, for instance, Quah’s twin income peaks (Quah, 1996 Maasoumi et al 2007) – firms in poor countries face low returns to investing in technological adoption while those in middle income countries face high incentives to converge rapidly to the frontier.
- They imply that we need to revisit innovation performance measures. Benchmarking innovation or targeting levels of R&D requires taking into account the stock of available complementary factors. In their absence, more R&D is not necessarily better. Hence, while it is not unusual to find unfavourable comparisons of a particular country’s gross domestic expenditure on research and development relative to that of frontier countries, and, on that basis, argue that more resources should be directed in that direction, this is only the case if the country also has similar levels of accumulated human and physical capital. Put differently, we’d like to know if a low level of R&D truly reflects innovation-related market failures. A low level of R&D investment may not reflect an innovation problem at all that would require, for example, government subsidies to remedy. It may rather reflect problems in the accumulation of all or other factors – ranging from poor school systems to absent credit markets. Simulations of innovation shortfalls that take into account complementary factors in fact show very little correlation between gross domestic expenditure on research and development and simulated barriers to innovation per se.2 The same logic applies to EU R&D targets.
- They suggest a broad conception of the national innovation system in developing countries. Given the above, the national innovation system (e.g. Nelson 1993) needs to include not only the usual innovation-related market failures, such as appropriability externalities, but also the functioning of the markets for all complementary factors. It also needs to be concerned with stimulating the demand for innovation on the part of firms and, again, with ensuring the capabilities of firms to respond to that demand. Put differently, while innovation policy in Sweden may reasonably focus largely on resolving innovation-related market failures because supporting markets work, the developing country innovation policy agenda must be much broader, balancing support to R&D with addressing the market failures that affect the accumulation of complementary factors.
This last point, in turn, raises what we term the innovation policy dilemma – at the same time that moving further from the frontier increases the number of market failures needed to redress in order to profitably innovate, the capabilities of governments to resolve them diminish. Hence, resolving the paradox requires increasing the capabilities of the state while narrowing the scope of interventions. For very low-income countries, this may imply focusing on, for instance, building firm capabilities or lower-level innovation, and perhaps leaving large investments in R&D until later.
-Xavi Cirera, Edwin A. Goñi Pacchioni, William Maloney