Do Economists Understand Science and Technology?
Partha Dasgupta

Recent years have seen a disturbing slowdown in the growth of productivity in Western industrial countries. Perhaps as a consequence the theory of market structures in general, and of technological competition in particular, is today one of the most active areas of economic research. Economists have, of course, long recognized that prices are but one weapon that firms use to compete with one another and that they are a dominant mode of competition only in the mature phase of an industry. It is only in the last decade, however, that economists have begun to understand the complex range of strategies used by firms in a capitalist society, not only to compete against one another, but also in many instances, to prevent other firms from competing. The use of such strategies as research and development (R & D), advertising, share issues, capacity creation and product 'quality' result in market structures and behaviour that differ markedly from the textbook descriptions of competitive markets.

Much of the new industrial organization theory is concerned with the world of technology, not science; the processes studied by the theory have been those that result in technological change, not scientific progress. It is therefore not surprising that if economists continue to understand technology less well than they would wish, they understand science even less. This is disturbing because it implies that the science policies pursued by governments will continue to be based on an even more fragile understanding than will their technology policies.

Both science and technology, as social organizations, are concerned with the production and use of information. Is it then useful for economists to distinguish between them? Can a reasonably precise analytical distinction be drawn between the search for scientific knowledge and those forms of technological research and discovery to which the writings of Joseph Schumpeter first drew attention? Historians and sociologists of science have considered the criteria for such a distinction already. Some have argued that scientific and technological research are to be distinguished by the characteristics of the 'research technology': for example, the allegedly greater risk involved in scientific research. Others have drawn the contrast on the basis of the nature of the commodity produced by the research: for example, by whether or not the addition to knowledge has practical applications (often contrasted by the labels 'basic' and 'applied' research). Finally, attention has been drawn to a distinction based on the way science and technology, as social organizations, treat the information they produce, in particular, the attitudes, behaviour and institutional arrangements relating to the communication of research findings. Science, it is argued, insists on complete disclosure, while technological research is kept secret or is protected by patents.

In CEPR Discussion Paper No. 74, Paul David of Stanford University and I argue that these distinctions are of limited relevance in the design of policy. For the purposes of economic analysis, the essential difference between science and technology lies in the respective goals the two communities have set for themselves. Roughly speaking, the scientific community appears concerned directly with the stock of knowledge, and is concerned to foster its growth. The technological community, on the other hand, is concerned with the private returns, or the economic rents, that can be earned from this stock of knowledge. In the social role of 'scientist', a researcher views the stock of knowledge ultimately as a public consumption good; in the social role of a 'technologist; however, the researcher regards it as a private capital good. Each community, quite naturally, seeks to inculcate in its members, through training and incentives, attitudes concerning research procedures that tend to further its particular objective. One manifestation of this process is the greater urgency shown by scientists in disseminating newly acquired information throughout the research community. This emphasis is not shared by the technological community, whose members are free to adopt information strategies ranging from disclosure (in the case of patent acquisition) to total secrecy in regard to their discoveries and inventions.

This distinction between science and technology enables us to explain why, despite the strong similarities between the two communities, individuals generally move from science to technology but not in the other direction. We also analyze the role of priority of discovery as a basis for allocating rewards among scientists, its compatibility with the norm of disclosure in science, and the ambiguous status of patent systems in technology. There are certain ineluctable conflicts between the goals of the two research communities. These conflicts help us to understand why the position of science in modern industrial societies is so exalted yet so financially precarious. Although the contributions of scientists and technologists to the search for knowledge are interdependent, and although science and technology enjoy a symbiotic relation, we find that science as a social entity is today in danger of being undermined by the technological community's conception of knowledge as a form of productive capital. If the position of science is undermined today, however, that of technology will be undermined tomorrow. This will result in slower growth of both public and private knowledge, which will further constrain the possibility of economic growth in the future.

The distinction that we draw between science and technology makes the diffusion of scientific knowledge somewhat easier to understand than that of technological knowledge. Many years ago Joseph Schumpeter constructed a theory of capitalist economic development based on a chain consisting of individual sequences of invention, innovation and diffusion; or, in other words, on the origins, development and use of technology. Valuable subsequent research by Edwin Mansfield and others has given us a better idea of the nature of diffusion processes. This earlier work, which was based on 'epidemic' learning models, has been supplanted by research which delves deeper into the workings of such learning processes. This is based on the hypothesis that individuals choose rationally whether or not to adopt a new technology.
The research has followed two routes. One has concentrated on the fact that individuals typically do not know how fruitful it will be for them to innovate; they learn from observing the success or failure of others who have already adopted the new technology. The second route ignores uncertainty and concentrates instead on the differences in the costs and benefits of the new technology. In this approach firms adopt a new technology at different dates because the costs they incur and the benefits they enjoy from adoption differ. As the technology becomes cheaper over time it is diffused among those firms which were initially more reluctant to adopt it. A central aim of this research has been to explain the familiar S-shaped 'logistic' curve which describes the diffusion of new technologies. In CEPR Discussion Paper No. 49, Paul David and CEPR Research Fellow Paul Stoneman have integrated these two approaches. They have at the same time allowed the profitability of innovation to be determined within their model and have thus directly linked the sequence of invention, innovation and diffusion through their analysis.

One focus of the new industrial organization theory has been microeconomics of technological change, but it is applicable more generally. It departs from the earlier theory in its attempt to explain not only industrial performance but also industrial structure. The structure of an industry, which earlier theory had taken as given and exogenous is now seen as endogenous. In particular, empirical regularities such as the relationship between industrial concentration and R & D expenditure which have traditionally been given causal explanations are no longer necessarily seen in these terms. The new theory treats both as jointly endogenous and subject to explanation. This altered - and deeper - perspective has profound implications for policies regarding research and development, and anti-trust.

A much-discussed issue in recent years has been the ability of potential competition to serve the same role that actual competition serves in textbook economic models. The theory of price competition, for example, has traditionally assumed that active firms are actually engaged in such competition. Actual competition therefore provides a discipline for firms. Recently several economists, most notably William Baumol of Princeton University, have argued that one does not need actual competition, that under certain circumstances the incumbent firms in an industry can be disciplined as effectively by potential as by actual competition. If the incumbent firm feels threatened by rivals 'standing in the wings', it will, under certain circumstances, act in as disciplined a manner as society would wish it to. It has been shown that even when the incumbent enjoys large economies of scale, the threat of entry by rivals can force it to expand its production to the point where it prices its product at the average cost of production and earns only normal profits. This implies that government regulation may well be unnecessary even when an industry is a natural monopoly; that is, even when the size of the market, relative to the economies of scale in production, is not large enough to warrant more than one active firm in the industry.
A key assumption in this theory is that the scale-economies are due to indivisibilities, or fixed costs, but not sunk costs. A firm's cost is 'fixed' if it is independent of the quantity produced so long as production is undertaken, but is not incurred otherwise. A cost is 'sunk' if it cannot be recovered even if the firm subsequently decides not to produce anything. A model that incorporates sunk costs must necessarily incorporate time explicitly. This may appear to be an obvious requirement of a model, and it is. But most elementary textbooks ignore the temporal sequence of production. A major achievement of the new industrial organization theory is its explicit handling of time and thus of history.

Now, it can be argued that all production involves some sunk costs. In a forthcoming CEPR Discussion Paper entitled 'Welfare and Competition with Sunk Costs', Joseph Stiglitz of Princeton University and I demonstrate that the presence of sunk costs, no matter how small, may serve as an effective barrier to entry for new firms. A potential entrant, contemplating entry, will wish to estimate what its profits will be subsequent should it enter. Suppose that it has good reasons for believing that competition with the incumbent will be fierce; or in other words, that the incumbent will not be accommodating, and so profits will be negligible. If entry involves a cost that must be sunk the firm will not enter because it expects that entry followed by competition will result in losses. The incumbent firm can anticipate such reasoning by its rivals and knows that the rival will not enter the market. Safe from competition, the incumbent can thus charge a monopoly price. Potential competition is here totally ineffective, and this provides an argument for government regulation.

The foregoing argument suggests strongly that the more fierce post-entry competition is expected to be, the less incentive there will be for firms to enter. This in turn implies less actual competition and greater security for incumbent firms. This tension between pre-entry and post-entry competition has important implications for policy. There are circumstances in which government policies which restrict post-entry competition actually serve to increase welfare, by encouraging entry! In our Discussion Paper, we show by means of a simple example that minimum price regulations (or in other words the imposition of price floors) can result in lower actual prices by encouraging entry. The converse of this is also true: policies that are intended to encourage competition, by making firms compete more vigourously within a market, may actually reduce both competition and social welfare by discouraging entry.

The extent to which costs are sunk is often a decision taken by the firm. Under some circumstances firms will take actions which they would not take in the absence of potential competition. They may sink costs earlier, or to a greater extent, than they otherwise would, simply in order to deter entry, and not for any other purpose. An important example of such discretionary sunk costs is R & D expenditure. An incumbent firm may have incentives to accelerate its R & D programme so as to obtain a patent, not necessarily for the purpose of producing the new product, but for preventing other firms from doing so. Thus a patent may be taken out but not used. In extreme cases as has been noted in some recent work, the incumbent may accelerate to R & D programme so as to obtain a sufficient research lead and only that. That is it may then decide not to complete the programme. The lead if sufficiently large will discourage entry because the potential entrant will know that if it tries to complete it will not succeed. These entry-deterrence activities on the part of the incumbent reduce welfare, because they are wasteful: profits are reduced and prices remain high. We present examples in which potential competition results in everyone being worse off than they would be were there legal barriers to entry.

An important form of technological progress is learning-by-doing, in which a firm's productivity increases with production experience. There is now extensive empirical evidence of this learning effect, for example in the semiconductor industry. One way of formulating this effect is to assume that the firm's cost per unit of production declines with its cumulative output. Of course, learning effects may spill over to other firms, so that a firm's unit cost tomorrow may also to some extent depend on the output of rival firms today. Industry managers call this the learning curve. Learning provides an interesting form of increasing-returns-to-scale, one which operates only through time. It in fact creates a natural monopoly, since industry-wide unit cost falls fastest if production is restricted to a single firm. Learning also involves sunk costs. A firm cannot 'unlearn' and recoup the expenditure it incurred acquiring (past) production experience!

Learning-by-doing provided economists such as Mill and Bastable with a loose argument for protecting domestic 'infant industries' against foreign competition, an argument which has been revived in recent years in the United States and Western Europe. But to date economic theory has told us very little about the industrial structure one might expect if there is a learning curve. Presumably, such an understanding is a pre-requisite for policy debates.

In another forthcoming CEPR Discussion Paper entitled 'Learning- by-Doing, Competition and the Infant Industry Argument', Joseph Stiglitz and I explore the effect of learning-by-doing on the structure, conduct and performance of an industry.

We find that learning provides a nationalized industry with a reason for pricing its product below current marginal production cost, for learning provides the link through which today's higher production leads to tomorrow's lower costs. We also develop formulae for giving optimal pricing policy when the nationalized industry must break even over its planning horizon. Setting price below current marginal (or unit) cost is still the right policy in the infant phase, though in the more mature phase price must exceed the then current unit cost so as to cover the losses incurred in earlier years. We also derive production and pricing rules that a protected private monopolist would follow and show that monopoly output is less in each period than the optimal output of a revenue-constrained nationalized industry.

The central case we examine, however, is one where market structure is not given, but is influenced by the learning process. We consider the case of an incumbent firm which is threatened by a rival possessing the option of entering the market now or at any time in the future. Entry involves an arbitrarily small fixed cost that must be sunk. We produce examples in which, as in our earlier paper, the presence of the potential entrant has absolutely no effect on the behaviour of the incumbent. In the presence of sunk costs, the invisible hand is not merely weak, it is paralysed! These examples also provide a formal demonstration that small historical accidents (such as which firm was the first to innovate) can have a large cummulative role in the temporal unfolding of an industry.

We then examine government policies for such an industry, exploring the implications of two policies often advocated, both reflections of anti-trust legislation in the United states. We first analyse the effect of the imposition of a ceiling on the market share that a given firm can enjoy - an application of the 'per se' doctrine. We also examine the effect of the government's setting current marginal costs as the lower bound on the price that the incumbent firm can charge - the rationale being that a lower price is symptomatic of predatory pricing. We find that both these policies can reduce social welfare in a wide variety of circumstances, because they severely reduce the extent to which learning in fact occurs in the resulting market.

We also examine optimum trade policies in the presence of learning-by-doing. Learning provides a government with an argument for subsidizing domestic production, not for imposing an import tariff on foreign competition. In fact if foreign firms face a learning curve and if domestic production is possible only at a high cost there is a case for subsidizing imports! Import subsidies coax greater output from the foreign exporter and thus reduce future production costs. Provided that import subsidies are suitably chosen, future import prices may be lowered to such an extent that the importing country is better off!

Recent developments in the theory of industrial organization have been for the most part in the field of positive economics. As noted earlier, our understanding of appropriate regulatory policies is vastly imcomplete. This is especially so in the field of technology policy. For this reason CEPR is organizing a major conference on The Economics of Technology Policy in September 1986. The conference agenda will address the following questions:

(1) What might one mean by appropriate rates of technological advance?

(2) Under what circumstances will the optimal rate of technological advance be generated?

(3) How should governments intervene to correct sub-optimality? (4) How does government policy aimed at one part of the Schumpeterian sequence of invention-innovation-diffusion affect the other parts?

(5) What are the links between trade and technology policies?

This is one of a series of articles describing research relevant to economic policy. Partha Dasgupta is Professor of Economics at Cambridge University and a Fellow of St John's College. He is Director of CEPR's programme on Developments in Applied Economic Theory and Econometrics. Further details of the research described in this article can be found in CEPR Discussion Papers Nos. 49 and 74. CEPR is organising a major conference on The Economics of Technology Policy in September 1986. Further information on this and other CEPR initiatives in science and technology policy can be obtained by contacting Professor Dasgupta at the Centre.