From the perspective sketched above, a wide range of policy analyses are potentially relevant to a particular policy discussion. At one end of the spectrum, there are studies focused on the particular policy op-tions under immediate consi deration and on plausible alternatives to these. These kinds of studies are very dependent on the particular context and are designed to explore the qeustion: What should be the next move? At the other end of the spectrum are studies that broadly survey the terrain. They aim to help improve strategic planning. Be cause the narrower studies are so particularized, unless they involve a methodological breakthrough they seldom are of durable interest. The focus of discussion here, therefore, is not on a particular present policy issue, such as whether the federal government should now aid the aircraft industry in developing the next generation of commercial aircraft or whether it should join with the American automobile in dustry in the support of automotive R&D. Rather we offer here a background analysis of the general issue of the appropriate and fruitful roles that active governmental support of industri al R&D can play. Any particular policy study, after all, assumes such an analysis.
Such an analysis depends, of course, on assumptions made about industrial R&D. Here we will make basic assumptions about R&D that are reminiscent of the discussion in Chapter II, but somewhat different from those employed in the more stylized models. R&D is an activity separate from production . It is a highly uncertain activity, and reasonable people will disagree on the rankings of R&D projects. The outcome of an R&D project may include a technology ready for implementation, or nothing may be found or invented. In either case, the outcome of a project also includes revised knowledge about technological alternatives. In particular, a successful R&D project re veals that similar but not identical R&D projects may yield similar but not identical technologies . An unsuccessful project provides gen eral information about the location of “dry holes.” As a consequence, if the topography of innovation is sufficiently regular, technical ad vance will be cumulative in the following sense. The outcome of one round of research and development projects (which includes some successes and some failures) defines a set of “neighborhoods” where it is a good bet that further R&D will locate technologies similar to and better than the technologies developed previously. These neigh borhoods may not be in close proximity to each other; rather, the promising lines of search branch into an exploration of distin guishable subclasses of technology. Research and development pro jects within one class provide knowledge relevant to the next round of research and development projects within that same class. How ever, they do not contribute much understanding that is relevant to research and development activity aimed at another class of tech nology.
It also is i mp ortant to be explicit about where the information rele-vant to industrial R&D decision making resides. In general, it resides with the organizations that are engaged in producing and marketing the product. These are the organizations that know about the strengths and weaknesses of prevailing technologies and of the targets and opportunities for improvement. They know how custom ers react to different product designs. At best it is time-consuming and costly to relay the bulk of this information to an R&D organiza tion that stands significantly apart from the producing and market ing organizations. And without the cooperation of the firms in ques tion, it is impossible. Thus, in an economy that relies basically on profit-seeking private enterprise to provide goods and services, it is virtually inevitable that much R&D decision making will be decen tralized to private business firms, with returns to R&D internalized through secrecy, patent protection, or market domination.
The questions under consideration are these. First, what will be the strengths and weaknesses of leaving industrial R&D totally in the private domain? Second, what are the opportunities and limitations for governmental involvement in industrial R&D? We will argue that both the anatomy of market failure and the opportunities and con straints on governmental action depend on the character of market structure and competition in the industry, as well as on such institu tional variables as the strength and scope of patents and the extent of industrial secrecy.
Consider an industry consisting of a large number of competing firms, each doing its own R&D . There are several different kinds of “market failure” that need to be recognized. First, if firms have less than perfect ability to exclude other firms from using their technol ogy, there is the well-known “template externality,” which stems from the chances that a technology that is found (created) by one firm will be imitated by others . If patents prevent direct mimicking, but there is a “neighborhood” illuminated by the innovation that is not foreclosed to other firms by patents, the externality problem remains, though in modified form. Second, and more recently emphasized in the literature, there are problems akin to those of multiple independ ent tappers of an oil pool or of fishermen working the same fishing ground. Incentives to be the first to invent, to get the patent, may in duce many firms to try to invent early. Barzel (1968) and others have pointed out that, under certain assumptions, in such a competitive race too many resources are applied too early. Given a set of estab lished patents and imperfect license markets, individual companies can make money from projects that would not be worthwhile had they access to the best technologies developed by others- projects that yield little social value. The stronger the patent rights, the greater the importance of the oil pool problem relative to the template problem. The template problem tends to hold total R&D spending to a level below a social optimum. The oil pool effect may spur R&D spending, but toward an allocation of effort that is socially inefficient.
Still another allocational problem emerges if technological ad vance is cumulative. In the competitive situation there would appear to be a problem regarding R&D that is similar to the one described by Hotelling in the case of location decisions. Where the returns to a firm from a technical advance must be assessed against the technol ogy it currently is using rather than against the best technology in the industry, and where the rough location of the best available tech nology is known and the neighborhood looks both promising and unprotected by patents, there are incentives in the system for every body to cluster around the same broad opportunity . In the develop ment of technology over an extended period, too much attention is focused on particular parts of the technological landscape and not enough real diversification of effort is achieved. If a firm explores new terrain, it is less likely to come up with some thing. And if it does, it knows that other firms will soon cluster around.
Consider now a monopolized industry, noting first the differences in its incentive structure relative to that of the competitive case. In a monopolized industry, neither template externality nor the oil pool externality exists. And the knowledge externalities that come from successful exploration of uncharted regions of the set are internal ized. There may be cost- side advantages as well. For many kinds of R&D there are economies of scale, at least up to a point, arising from several different sources. Certain kinds of R&D inputs and outputs are lumpy: a significant quantity of R&D effort must be directed to a project if there is to be any hope of success. A small-scale R& D effort may not be able to achieve success at all, and, if it does, will achieve it significantly later than an effort that is funded at a higher rate . There also are diversification advantages of a large-scale research and development effort. Multiple attacks on particular objectives can be mounted. A large and diversified range of projects can help guard a company from the economic disadvantages of a long dry spell between R&D successes. And to the extent that the rate of growth of capital or sales of a firm is limited, there is an economy of scale asso ciated with the fact that a big firm can quickly apply a new tech nological development to a larger quantity of output and capital than can a small firm.
What are the debits of monopoly to be charged against these cred-its? Traditional theory would argue that the size of output in the in dustry would be lower. This causes the traditional triangle loss. It also feeds back to R&D incentives by reducing the size of the output to which R&D applies. It is hard to say whether there would be more or less R&D undertaken in the monopolized case than in the compet itive case. The greater degree of internalization and the smaller scale of output pull in different directions. In the monopolized case there will be less incentive to do the kind of R&D that is profitable in a competitive case only because someone else has a patent. While this is another factor that acts to lower the R&D level in the monopolized case relative to the competitive case, it suggests that the most impor tant difference in the two regimes is the efficiency of R&D allocation. If the monopolist can be assumed to be a profit maximizer and if the consequences of choosing any particular R&D project are more or less obvious, there are strong arguments that monopoly would gen erate a better portfolio of R&D projects than would a regime of com petition.
However, this tentative conclusion looks less compelling if we note that different people see alternatives in different ways and that organizations have tendencies to adopt parochial viewpoints and simplified decision-making styles . Then a central ized regime looks less attractive in terms of the portfolio of projects it would be likely to cany, and a competitive regime looks more attractive. The argument here against monopoly and for competition is not the standard one of textbook economic theory. It does not derive from the logic of maxi mizing choice or from arguments akin to the proposition that it is so cially desirable to set the level of output at the point where marginal cost equals price. Rather, the argument is in part that differences in perception as to what are the best bets will in a competitive regime have a greater chance to surface and be expressed in a diversified portfolio of R&D projects than they would in a monopolized regime. The argument also is that large, sheltered organizations tend to be stodgy and uncreative or narrowly messianic in the R&D they do, rather than ingeniously and flexibly creative. It is not just that monopoly limits the sources of new ideas, but that an industry domi nated by a large, secure firm is not a setting that spurs the generating and sensitive screening of good ideas. Any regime of competitive R&D is bound to involve some waste and duplication. The costs and dangers of monopoly are principally those of reliance on a single mind- unlikely to be an agile one-for the exploration of tech nological alternatives.
One is tempted to look to a regime of oligopoly-involving neither the R&D incentive problems of a multitude of small produc ers, nor the pricing and single-source reliance problem of a true monopoly-as the most desirable institutional structure. Many prominent economists, from Schumpeter to Galbraith, are associated with this position. And, interestingly, oligopoly tends to be the market structure that naturally seems to evolve in industries where the funding or inventing of new technologies has proceeded rela tively rapidly. An oligopolistic structure has the potential of com bining the best aspects of competition and pluralism and of R&D benefi t internaIization .
But such a structure also has the potential for combining the worst features of monopoly and competition. In many oligopolistic indus tries, a considerable amount of R&D done by firms seems to be “de fensive” and aims to assure that a fi rm has available a product similar to that developed by a competitor, rather than aiming to come up with something significantly different. Small numbers may yield considerable duplicative R&D without any real R&D diversity.
And economists who tout oligopoly as p rogressive should be more alert to the possibility that, where oligopolistic rivalry in R&D does involve firms exploring significantly different p arts of the range of technological alternatives, oligopoly may be unstab le . A monopo lized structure may gradually evolve. A central feature of Schumpe terian competition is that the profits th at are the reward of successful innovation provide both moti vation and the funds for firm growth. And there are social economic advantages of having the firm with a better technology (the lower cost or the better product) supply a growing share of the market. However, to the extent that firm re search and development expenditure s are keyed to size and to the ex tent that there are advantages of scale of any of the sorts discussed above, a successful innovator may reduce i ts rivals to a point where they can be effective competitors no longer. Where oligopolistic Schumpeterian competition has the merits that some observers as sign to it, our simulation studies suggest that the structure may tend to self-destruct.
The “failures” of market- induced R&D may well be serious, at least if performance is judged against the standards of an ideal plan ning model. There is indeed a fundamental dilemma in using profi t-seeking firms and competitive markets as the organizational device for stimulating and guiding R&D. If the problem were simply 1/ externality,” as some economists seem to beli eve, it could be re solved through tightening p atents or p roviding simple R&D sub sidies. But th e problem is much more complex than that, involving overspending on certain types of R&D as well as underspending on others, the warping of R&D s trategies, and constraints on the use of what is essentially a public good: knowledge. Market failure regarding R&D is not neatly resolved by giving a small a djusting twist to conventional policy instruments or by introducing a few new ones.
Of course, it is possible to take R&D or some component of R&D largely out of the market system. This is what has happened regarding basic research, which is conducted mostly at universities rather than at profit-seeking firms and is funded largely by the gov ernment. This strategy has been relatively successful for two reasons (if the analysis above is accepted) . First, the information needed to guide basic research decision making is not located in the operating parts of organizations that produce goods and services, but rather in the minds and experience of basic research scientists. Relatedly, the opportunities and problems guiding allocation are signaled by the logic and values associated with advancing scientific understanding, rather than by the profit objectives of enterprises. Indeed, these dis tinctions form a basis for defining basic research. Second, basic research decision making has been largely decentralized and plural istic; the proposals come mostly from research scientists and institu tions and are subject to a peer review system or something equiva lent.
Society in effect has a choice regarding what arenas of research it will define as basic research, to be funded publicly and guided by the tenets of a scientific discipline, and what arenas it will regard as ap plied and to be guided (if not necessarily funded) by criteria close to the values of the organizations using particular technologies . The lesson of history is that the former approach has large long-run prac tical payoffs, when a field thus defined as a science can advance pro gressively and when the scientific understanding illuminates tech nological options and their connections with economic values. It should be recognized that many of the same kind of inefficiencies of market decentralized R&D- allocating mechanisms reside in this de centralized “Republic of Science,” which allocates basic research resource s.6 No hidden-hand theorems obtain for either system. But so long as it pays to have some bodies of research guided by the logic inherent in the natural unfolding of certain bodies of understanding, it is viable to have scientific criteria (as contrasted with profitability criteria) guide R&D allocation in those areas. So long as good deci sion making needs detailed access to the particulars of ongoing re search and so long as there are dangers of a single central mind any how, decentralization seems far preferable to a more centrally planned system.
But for the bulk of R&D that bears on advancing industrial tech nology, much of the relevant information is located in the production enterprises, and good R&D decision making involves attending directly to economic benefits and costs . The Republic of Science is not an appropriate system for governing the problem-oriented R&D work aimed at advancing production technology. Let us ignore here the cases in which government itself is a heavy purchaser of the prod uct in question or is its provider (where special considerations ob tain) and focus on private industry selling products largely to other private parties . In these circumstances, government is severely con strained in terms of what it can do to guide and support R&D. There are, first, informational constraints; second, constraints imposed by the requirement for “fairness”; and, third, constraints arising from bureaucratic politics .
The first two constraints turn out to be closely connected. Where the suppliers of goods and services are not rivalrous, the govern mental i nformation access problem can be resolved. Agriculture and medical practice are good examples. In these cases governmental in formation gathering and R&D support are not vi ewed as helping one part of the industry at the expense of another, but as helping the whole industry. (Whether this conception is justified or not is an other matter. ) Not only governmental R&D support but public insti tutions to allocate these funds, and even public R&D undertakings, generally are welcomed in these arenas.
The difficult problem of information access arises, along with real problems of “fairness,” where private suppliers are rivalrous. This is the situation that characterizes much of American industry. In such a regime, the kind of information that enables good R&D decisions to be made is the kind of information that gives one firm a competitive advantage over another. Since R&D often is an important instrument of competitive policy, firms are not likely to be cooperative when governmental programs are proposed that might upset the competi tive balance. Governmental or other outside interests may conjecture that the risks and limited capturability of certain technological ven tures are deterring private investment, but they are likely to have great difficulty in finding out exactly how much private firms are spending on these endeavors. Proposals that companies share their technological knowledge are likely to unify the companies and the antitrust division in resistance. It is hard for public policy to fill in the holes in the portfolio when there is no solid information as to what that portfolio actually is .
Governmental policies not only are limited by information access constraints, but are limited to those actions that industry considers as generally supportive, neutral, and unthreatening to the status quo. Thus, industry has long advocated even-handed tax credits. Support of cooperative research institutions run by industry and guided to keep out of fields of proprietary interest has been em ployed widely in Europe, less widely in the United States. Not sur prisingly, cooperative R&D tends to concentrate on techniques of common interest not likely to give any firm a competitive advantage. Support of industry-specific basic research and pilot development of certain technologies at universities, nonprofit institutions, and governmental laboratories has been used occasionally by the United States. This seems to be viable politically, so long as governmental funds go into projects far enough away from actual practice so that there are no obvious likely gainers and losers among private compa nies. Support of atomic energy and civil aircraft technology are good examples . The experience in these fields has been quite mixed. A good part of the difficulty certainly has been that, unless the govern ment completely takes over industry research on the fronti ers of tech nology, the informational and fairness constraints in a sense force the government to explore alternatives that no private firms think are worthwhile funding themselves. In some cases, there may be a real “market failure” problem that governmental funding is resolving. But all too often, what industry was not funding was not worthwhile funding (even by broad social criteri a) , at least not at that particular time.
The aircraft and atomic energy cases also signal the “bureaucratic politics” problem. While governmental support of R&D activities in these fields initially was justified in terms of the “public knowledge” nature of frontier-probing R&D, over time there developed within government a constituency for particular technological options and R&D proj ects. As suggested above, this does not seem inevitable regarding governmental R&D programs, but avoidance seems to re quire building in pluralism either through geographic and political decentralization (agriculture) or the use of outside peer reviews ( Na tional Institutes of Health, National Science Foundation) . This is dif ficult to do if the relevant long-run criteria are commercial, if much relevant knowledge is industrial rather than open and scientific, and if the relevant industries are rivalrous.
So, although the “market failure” may be serious if the basis of comparison is R&D allocation under an idealized optimized plan, surely this is the wrong basis for comparison.1 In economies where the production of goods and services is largely conducted through profit-seeking business firms selling their goods and services on rea sonably competitive markets, it is inevitable that these organizations be the locus of the bulk of R&D activity. Certain kinds of R&D can be established in other institutional regimes guided by other informa-tion and incentive systems. Perhaps the regime of academic basic re search is the best example. But a good share of industrial R&D must be guided by info rmation available in and criteria relevant to the fi rms who eventually use the technology. Government is quite lim ited in the extent to which it can effectively supplant the market. And government-business cooperation is severely constrained by bUsi ness rivalry.
The point of view on market failure, and limits on government ac tion, in industrial R&D lent by evolutionary theory is not totally di vergent from that which would be lent by positive orthodox theory. But the emphasis is different. In the first place, the current state of uncertainty regarding the range of things that can be done, and the consequences of doing various things, is stressed. Second, no at tempt is made to define an optimum policy; rather, the style of analy sis is to try to identify policies that should be avoided and others that appear more promising, and to focus attention on the latter. In part this represents carrying over to the arena of policy analysis our explicit recognition of bounded rationality. In part it represents a more general acknowledgment that notions like ” market failure” cannot carry policy analysis very far, because market failure is ubi quitious. Finally, it involves an explicit recognition that govern ments are quite limited in the things they can do well, and that there fore policy analysis should be concerned with these constraints as well as with the inefficiencies of private action. Third, flexibility, experimentation, and ability to change direction as a result of what is learned are placed high on the list of desiderata for proposed institu tional regimes.
Frankly, we do not know of any “orthodox” economic analysis of the fruitful and unfruitful roles of government in industrial R&D to contrast with our own. This is largely because those economists who are seriously interested in the question, while they often use ortho dox language and concepts, tend to adopt a point of view that is im plicitly, if not explicitly, evolutionary. See, for example, Marschak, Glennan, and Summers (1967) or Noll ( 1975). Our point is that analy sis of the problem is hindered, not advanced, by the assumption that firms literally maximize profit and industries are in equilibrium, and is advanced when bounded rationality and slow-moving selection are recognized explicitly .
As scholars who have drawn so much from Schumpeter, we find it interesting that o ur policy perspective on industrial innovation ap parently differs signifi cantly from his in at least one important respect. As part of his prognostication of socialism arising out of capitalism, he spoke of the p ending routinization of innovation and the decline of the entrepreneur. He seemed to argue that these devel-opments would tame technological advance, or the economic adjust ments required by it, but not hobble or badly distort innovation. Our analysis, perhaps influenced by our knowledge of the fate of such efforts to plan and optimize technological advance as the super sonic transport and the breeder reactor, leads us to a different posi tion. The attempt to optimize and accordingly to control tech nological advance will, according to the evolutionary theory we espouse I lead not to efficiency but to inefficiency.
Source: Nelson Richard R., Winter Sidney G. (1985), An Evolutionary Theory of Economic Change, Belknap Press: An Imprint of Harvard University Press.