«Industrial Policies, the Creation of a Learning Society, and Economic Development1 B. Greenwald and J. E. Stiglitz Industrial policies—meaning ...»
Industrial Policies, the Creation of a Learning Society, and Economic
B. Greenwald and J. E. Stiglitz
Industrial policies—meaning policies by which governments attempt to shape the sectoral
allocation of the economy-- are back in fashion, and rightly so. The major insight of welfare
economics of the past fifty years is that markets by themselves in general do not result in
(constrained) Pareto efficient outcomes. (Greenwald-Stiglitz, 1986).
Industrial policies seek to shape the sectoral structure of the economy. This is partly because the sectoral structure that emerges from market forces, on their own, may not be that which maximizes social welfare. By now, there is a rich catalogue of market failures, circumstances in which the markets may, say, produce too little of some commodity or another, and in which industrial policies, appropriately designed, may improve matters. There can be, for instance, important coordination failures—which government action can help resolve.
But there are two further reasons for the recent interest in industrial policy: First, it has finally become recognized that market forces don’t exist in a vacuum. Development economics routinely emphasizes as central to growth the study of institutions All the rules and regulations, the legal frameworks and how they are enforced, affect the structure of the economy, so unwittingly, government is always engaged in industrial policy. When the U.S.
Congress passed provisions of the bankruptcy code that gave derivatives first priority in the event of bankruptcy, but which said that student debt could not be discharged, even in bankruptcy, it was providing encouragement to the financial sector. Secondly, it has also been realized that when the government makes expenditure decisions—about infrastructure, Paper presented to the International Economic Association/World Bank Industrial Policy Roundtable in Washington, DC, May 22-23, 2012. The authors would like to thank the participants in the seminar for their helpful comments. This paper is based on Greenwald and Stiglitz (2006, forthcoming) and Stiglitz . Greenwald and Stiglitz  provides a sequel to this paper, focusing on the implications of learning for industrial policy in the context of Africa.
education, technology, or any other category of spending—it affects the structure of the economy.
This paper is concerned with one particular distortion: that in the production and dissemination of knowledge. Markets, on their own, are not efficient in the production and dissemination of knowledge (learning). Sectors in which learning (research) is important are typically characterized by a wide variety of market failures.
Both econometric and historical studies highlight the importance of learning and innovation.
Maddison’s (2001) research, for instancedocuments that from the origins of civilization to the early 1800s, there was essentially no increase in incomes per capita. The economy was close to static. The subsequent two centuries have been highly dynamic, leading to unprecedented improvements in standards of living.
Since the work of Solow (1957), we have understood that most increases in per capita income—some 70%-- cannot be explained by capital deepening; for the advanced developed countries most of the “Solow residual” arises from advances in technology. At least for the past quarter century, we have understood that a substantial part of the growth in developing countries arises from closing the “knowledge” gap between themselves and those at the frontier. Within any country, there is enormous scope for productivity improvement simply by closing the gap between best practices and average practices. (Greenwald-Stiglitz, forthcoming.) Knowledge is different from conventional goods; it is, in a sense, a public good (Stiglitz, 1987a, 1999)—the marginal cost of another person or firm enjoying the benefit of knowledge (beyond the cost of transmission) is zero; usage is non-rivalrous. Markets are not efficient in the production and distribution of public goods. It is inevitable that there be, or that there ought to be, a role for government.
Moreover, as Arrow (1962a) pointed out fifty years ago, the production of knowledge is often a joint product with the production of goods, which means that the production of goods themselves will not in general be (intertemporally) efficient.
If it is the case that most increases in standard of living are related to the acquisition of knowledge, to “learning,” it follows that understanding how economies best learn—how economies can best be organized to increase the production and dissemination of productivity enhancing knowledge-- should be a central part of the study of development and growth. It is, however, a subject that has been essentially neglected. That would, by itself, be bad enough.
But Washington Consensus policies based on neoclassical models that ignore the endogeneity of learning often have consequences that are adverse to learning, and thus to long-term development.
Creating a learning society
Not only is the pace of learning (innovation) the most important determinant of increases in standards of living; the pace itself is almost surely partially, if not largely, endogenous. The speed of progress has differed markedly both over time and across countries, and while we may not be able to explain all of this variation, it is clear that government policies have played a role.
Learning is affected by the economic and social environment and the structure of the economy, as well as public and private investments in research and education. The fact that there are high correlations in productivity increases across industries, firms, and functions in firms suggests that there may be common factors (environmental factors, public investments) that have systemic effects, and/or that there may be important spillovers from one learner/innovator to others. But the fact that there are large, persistent differences across countries and firms—at the micro-economic level, large discrepancies between best, average, and worst practices-- implies that knowledge does not necessarily move smoothly either across borders or over firm boundaries.
All of this highlights that one of the objectives of economic policy should be to create economic policies and structures that enhance both learning and learning spillovers: creating a learning society is more likely to increase standards of living than the small, one time improvements in economic efficiency or those that derive from the sacrifices of consumption today to deepen capital. 2 And this is even more so for developing countries. Much of the difference in per capita income between these countries and the more advanced is attributable to differences in knowledge.
Policies that transformed their economies and societies into “learning societies” would enable them to close the gap in knowledge, with marked increases in incomes.3 Development entails learning how to learn.4 Market failure and learning. While the fact that knowledge is a (global) public good means that the production and dissemination of knowledge that emerges in a market economy will not, in general, be efficient, there are several other market failures that inevitably arise in an important way in the context of a learning economy.
The first set is related to the fact that those who produce innovation seldom appropriate the full value of their societal contributions. There are large externalities, and these externalities will play a pivotal role in the analysis below. Even when an innovator becomes rich as a result of his innovation, what he appropriates is sometimes but a fraction of what he has added to GDP. But even more, many of those who have made the most important discoveries—those who regularly contribute to the advances of basic science and technology—receive rewards that are substantially below their social contributions: Think of Turing, Watson and Crick, Berners-Lee or even the discovers of the laser/maser and the transistor.5 But externalities are more pervasive. Individuals who learn about better ways of doing business transmit that knowledge when they move from one firm to another. (We’ll discuss these spill overs at greater length below.) As Solow (1956) pointed out, an increase in the savings rate simply leads to an increase in per capita income, not to a (permanently) higher rate of growth.
See Stiglitz 1998, which describes development as a “transformation” into a society which recognizes that change is possible, and that learns how to effect such.
Stiglitz 1987b One should, perhaps, not put too much emphasis on the fact that these individuals did not appropriate the full benefits of their innovations: there is little evidence that they would have worked any harder with fuller appropriability. Discussions among economists focus on economic incentives; these may be far from the most important determinants of learning/innovation.
The second set is related to our imperfect attempts to provide incentives for innovation, through intellectual property. The result is that private rewards are typically not commensurate with (marginal) social returns, in some cases exceeding the social returns (metoo innovations, innovations that are designed to lead to “hold-up” patents)6, in other cases being markedly less. The fact that the distortion which industrial policy may be attempting to partially “correct” arises from a government policy highlights an aspect of industrial policy upon which we comment further in the concluding section of this paper: it is not just market failures which lead to “distortions” in the economy, but also “government failures.” (One could argue that it would make more sense to eliminate the government failure than to introduce another intervention in the market. But for one reason or another, typically related to political economy, it may not be easy to eliminate some government policies; it may be easier to introduce a new countervailing policy.) A third source of inefficiency which industrial policies may address arises from capital market imperfections (themselves endogenous, arising from information asymmetries). But capital market imperfections can be particularly adverse to learning: Because R & D investments (or “learning investments”7) typically cannot be collateralized, unlike investments in buildings, machines, or inventories, it is more likely that there will be credit and equity rationing, leading to underinvestment in these areas, compared to others.8 There are other important interactions between traditional market failures, like imperfect competition, and learning: sectors in which innovation is important are naturally imperfectly competitive—research expenditures are fixed costs, and give rise to increasing returns.
Because sectors in which competition is limited, output will be lower, and accordingly returns to cost-reducing innovations are lower. (Arrow, 1962b) Still another market failure arises from imperfections in risk markets. Innovation is highly risky—research is an exploration into the unknown. But firms cannot purchase insurance The social return is related to the arrival of an innovation earlier than would otherwise be the case. For a more extended discussion of these issues, see Stiglitz (2006, 2008, 2013).
Optimal learning may involve producing at a loss, necessitating borrowing. See Dasgupta and Stiglitz (1988).
This is an explanation of the high observed average returns to investment in technology. See Council of Economic Advisers, 1997.
against these risks (because of well-known problems of moral hazard and adverse selection).
But because of imperfections in capital markets, firms act in a risk-averse manner, particularly in the presence of bankruptcy costs (Greenwald-Stiglitz, 1993), and this discourages investment in riskier innovation.
Problems of appropriability of returns and imperfections of capital markets (including the absence of good risk markets) result in barriers to the entry of new firms (entrepreneurs) and the exploration of new products—products or processes that might be particularly appropriate for a developing country. Consider an “experiment” to discover whether conditions in a country are particularly suitable for growing a particular kind of coffee. If the experiment fails, those who conduct the experiment lose money. If it succeeds, there may be quick entry. The country benefits, but the “innovator” can’t capture much of the returns. In short, an experiment that is successful will be imitated, so the firm won't be able to reap returns; but the firm bears the losses of an unsuccessful experiment. As a result, there will be underinvestment in this kind of experimentation (Hoff, 1997).
A similar argument holds for why private markets will lend too little to new entrepreneurs. The borrower who becomes successful will be poached by other lenders, so the interest rate which he can charge (after the entrepreneur has demonstrated his success) will be limited to the competitive rate. But Stiglitz-Weiss adverse selection and adverse incentive effects limit the interest rate that can be charged in the initial period, which implies that there will be limited lending to new entrepreneurs. (Emran-Stiglitz, 2009).
In the absence of lump sum (non-distortionary) taxation, there is a fundamental tension:
research is a fixed cost, and there is no marginal cost to the use of an idea, so that knowledge should be freely provided. But that would imply that the producer of information (knowledge) would receive no returns. Thus, it is inevitable that there be an underproduction of knowledge (relative to the first best) and/or an underutilization of the knowledge that is produced. The patent system (in principle) attempts to balance out the dynamic gains with the short-run costs of the underutilization of knowledge and imperfections of market competition.9 When the government finances research and disseminates it freely, there is still a static distortion (from the distortionary imposition of taxes), but no distortion in the dissemination and use of knowledge.