Economics of Technology Policy
A neglected step child?

It is widely appreciated that investment in research and development (R&D) has a strong influence on an economy's performance. It is therefore surprising that the economics of technology policy is only in its infancy. One reason may be that information, the output of research and development, possesses exceptional characteristics which complicate its production.

In the study of technology policy, as in other many fields of economics, there is too little contact between theorists and empirical researchers or between the academic community as a whole and policy-makers. CEPR therefore brought together academics, policy-makers and private sector practitioners from the United Kingdom, continental Europe and the United States at a conference in London on September 1-2, to discuss the economics of technology policy. The conference was held under the auspices of the Centre's Applied Economic Theory and Econometrics programme, organized by Research Fellows Partha Dasgupta and Paul Stoneman, and with financial support from the Department of Trade and Industry, British Telecom, and the Centre for Economic Policy Research International Foundation. The papers presented at the conference are to be published by Cambridge University Press in early 1987.

The conference was opened by Partha Dasgupta (St John's College, Cambridge, and CEPR), who presented 'The Economic Theory of Technology Policy: An Introduction'. He argued that five inter- related questions are at the heart of the normative economics of technology policy. First, what problems should be on the research and development agenda of the private and public sectors? Second, how many and what kinds of research projects or strategies should be pursued? Third, how should resources be allocated among the chosen research projects? Fourth, who should conduct research, and last, how should research personnel be compensated for their work? Although there are no conclusive answers to these questions, Dasgupta argued that recent research has increased our understanding of the issues involved.

There was an important economic difference between science and technology and between science policy and technology policy, Dasgupta argued (see his article in CEPR Bulletin No. Nine/Ten). Science and technology are distinct social organizations, guided by different goals. These differences explain why science insists on information disclosure and why technology encourages secrecy. Because of these differences, science plays a vital role in screening research personnel for technology.

Dasgupta focused on technology policy. He noted that economists approached the study of technology policy by treating research as the production of a good, 'information'. Economic analysis of technology policy must therefore focus on the characteristics of this good which, Dasgupta argued, possessed certain unique properties. First, there are economies of scale in the value of information at low information levels. It may not be worthwhile to invest a tiny sum to gain a tiny amount of information: it may only be worthwhile seeking information in large chunks. This gives the market for information some non-competitive features, Dasgupta argued. Second, information need only be produced once, and society benefits only from the first or best discovery of a new product or process. Research portfolios must therefore be analysed in a different manner from conventional financial portfolios. Third, information resembles a public good. A piece of information can be shared without eroding its value and can be used over and over again in production at any scale of operation.

These characteristics of information suggested to Dasgupta that markets may fail to allocate resources to research in an optimal fashion. Public policy, such as the patent system, may attempt to correct such market failures. Dasgupta argued, however, that such a 'winner takes all' system encourages excessive R&D investment and excessive risk-taking, and free entry into the research 'market' could lead to an excess of entrants. If all that counts in the R&D 'race' is victory, firms may chose insufficiently correlated research projects. There is an incentive for rival research teams to distance themselves from each other more than is socially desirable, since this may yield higher expected private benefits, given that the loser in the race gets nothing, or at best very little.

In the discussion that followed, participants questioned several aspects of Dasgupta's analysis. It was argued that it may not be useful to consider technological advance purely as the production of information, especially if this information can be transferred and used without cost. Some conference participants argued that Dasgupta's 'winner takes all' assumption was inappropriate and was rarely observed in practice; this led to a discussion of the problems of intellectual property rights in cooperative research. Participants also argued that research expenditures and development expenditures must be distinguished. It was argued that expenditure on development, which is the largest part of investment in R&D, is not governed by the same conditions as research expenditures. The final issue, which was to arise throughout the conference, concerned the principle of 'misplaced concreteness'. Were the theoretical results described by Dasgupta a satisfactory guide for policy-makers? Participants argued that the theory should be applied cautiously until we were certain that the results were robust to changes in its underlying assumptions.

Dasgupta's theoretical introduction was followed by two papers on the practice of technology policy. The first of these was by John Barber (Department of Trade and Industry) and Geoff White (HM Treasury), entitled 'Current Policy Practice and Problems from a UK Perspective'. Barber and White examined the scope and rationale of technology policy in the context of general economic policy, and summarized current DTI policies towards science and technology. The authors noted the growing emphasis placed on technology policy by the UK government and the increasing importance of collaborative research. They also stressed the diversity of the sources of technological change, and the number of steps involved in the introduction of new technology, often formalized as invention, innovation and diffusion. At present, Barber and White argued, the political objectives of the UK government are such that technology policies are directed at correcting failures in the market mechanism rather than at the pursuit of grand strategies.

Two aspects of Barber and White's paper provoked particular controversy in the discussion that followed. The first was the relationship between defence procurement and technology policy. Over half of current UK government support for R&D is for defence purposes, and the defence procurement budget is also very large. Some participants contended that this expenditure strongly influences the direction and organization of the UK technological effort. In order to offset this bias, they argued, defence R&D spending should be at least partly determined by the wider objectives of civilian technology policy.

The discussion of Barber and White's paper also focused on the relative merits of 'bottom-up' and 'top-down' policies. Should government establish detailed technological objectives and take action to ensure that these objectives are achieved? Or should government instead react to pressures from practitioners and consumers of R&D? It was agreed that there were problems with both approaches. 'Top-down' policies could be subject to bureaucratic failure, while 'bottom-up' policies could be too sensitive to the lobbying of articulate pressure groups, so that the allocation of government assistance might be unrelated to need.

Henry Ergas (OECD) then asked 'Does Technology Policy Matter?'. Ergas attempted to classify the practice of technology policy. He defined mission-oriented countries as those where the emphasis of technology policy was on large prestigious projects, often in defence. The United Kingdom, France and the United States were mission-oriented, he argued. His second category was diffusion- oriented countries, where policy emphasized measures to encourage the use of new technologies. These measures included the public provision of 'innovation-related' public goods such as human capital and technological standards, and the encouragement of collaboration among industries and between industry and academic researchers. Sweden, Switzerland and West Germany were diffusion-oriented. The third category, consisting of Japan alone, involved a mixture of mission-oriented and diffusion- oriented policies. Ergas argued that the mission-oriented countries were not particularly successful, although the sheer scale of resources employed and the flexibility of its market institutions has enabled the United States to be more successful than France or the United Kingdom. The diffusion-oriented countries have been more successful, in Ergas's opinion, but the implementation of these policies had made them sometimes slow to respond to major shifts in technology. The success of Japan is well documented.

The ensuing discussion initially centred on whether the patterns Ergas observed reflected only the significance of defence spending in the different economies. Ergas had defined defence R&D as an instrument of technology policy, in contrast to the view taken earlier by Barber and White. Participants could not agree whether the observed patterns were the results of deliberate technology policy decisions or merely the outcomes of expenditure patterns determined by other objectives. Zvi Griliches (Harvard University), for example, suggested that the United States was not mission-oriented: it had no technology policy at all!

The discussion also touched on a topic that arose throughout the conference, namely the question of ergodicity and the importance of history in analysis and policy formulation. Could a country easily switch from one set of technology policies to another? The choices available to a country today may be conditioned by the historical process by which it has reached its current position, and its choices tomorrow may be conditioned by where it is today. What is appropriate for one economy, therefore, may not be appropriate for others that have followed different paths.

The first day was brought to a close by Zvi Griliches and Ariel Pakes (Harvard University), who discussed 'The Value of Patents as Indicators of Inventive Activity'. Griliches and Pakes discussed the development of a data set compiled at the National Bureau of Economic Research, which correlated the number of patents issued to US manufacturing companies from the 1960s to the 1980s with their sales, employment, and investment and their R&D expenditure, in order to illuminate the process of innovation and technological change. Griliches and Pakes found a strong relationship across firms between the number of patent applications and the level of current and past R&D expenditure. The relationship between R&D and patents appeared to be almost proportional, especially for firms above a minimal size. There was also a statistically significant but weaker relationship over time between R&D expenditure and patents for a given firm. The bulk of the observable effect is contemporaneous. There is some evidence that history also matters, in the sense that there is a lag between R&D and patent applications, but this effect seems to be small. It might also imply reverse causality: successful research leads both to patents and to a commitment of additional funds for the development of the resulting ideas.

Griliches and Pakes also used data on patent renewals from selected European countries to estimate the private value of patent rights, their dispersion, and their decay over time. They found that the average value of patent rights is quite small, about $7,000 and $17,000 per patent in France and West Germany, but the distribution of values is very skewed. While most were worthless, 1% had values in excess of $70,000 and $120,000 respectively in France and West Germany, although the returns tended to decline rapidly. They also estimated the value of patent rights to be only about 10-15% of total national expenditure on R&D, so they are unlikely to be major factors in determining R&D spending. The total number of patent applications fell during the 1970s in these countries. Griliches and Pakes were able to construct a 'quality-adjusted' index of patents by using data on the value of patents. Their quality- adjusted index rose during the 1970s, demonstrating that patent numbers could be an unreliable indicator of R&D 'productivity', at least in the short run.

The discussion centred on the extent to which the NBER data adequately reflected the level of and returns from R&D undertaken overseas, the effect of changes in patent institutions, and the reasons why firms take out patents. It was, however, generally agreed that work such as this, seeking to measure technological advance quantitatively, can significantly contribute to our understanding of the process of technological change.

Why are some societies more innovative than others? In the opening paper of the conference's second day, entitled 'Learning to Learn, Technological Change and Economic and Social Structure', Joseph Stiglitz (Princeton University) outlined a theory of 'multiple social equilibria', in which some societies were characterized by high levels of innovation and others by low levels.

Stiglitz considered an economy in which there were two kinds of individuals, 'innovators' and 'bureaucrats'. The latter enjoy routine established patterns of behaviour, the former enjoy change. The behaviour of individuals can change, as a result of social interactions. For example, a junior worker whose behaviour conforms to the preferences of management may be rewarded and an innovator in a bureaucratic environment may come to enjoy routine. Stiglitz argued that in such a model multiple equilibria were possible: a society could settle into either an innovative or a bureaucratic equilibrium. The survival of a characteristic such as innovation depends on the environment, which is itself endogenous. Bureaucrats create an environment in which routine is rewarded; innovators create an environment in which there is an incentive to be innovative. The proportion of bureaucrats and innovators at time t is therefore a function of the proportion of bureaucrats and innovators in the older generation, i.e. at time t-1.

Stiglitz also explored how the level of technology already employed by an economy affects its rate of technological change. He noted that Adam Smith had first emphasized the economic importance of specialization, which arises essentially from increasing returns to scale. One of the most important examples of increasing returns was associated with learning: the process of learning can itself be learned, by developing a frame of mind associated with discovering how a task can be performed better.

There may exist technologies in which most learning possibilities are already exhausted and those using these technologies have little opportunity to learn; in these circumstances, technological change will be slow. Suppose that individuals have not had a chance to learn, as a result of the technology currently in use. If they are presented with a new technology which has a greater potential for technological improvement, they may lack the skill to make these improvements because they have not 'learned to learn'.

Stiglitz noted the relevance of this argument to development policies. LDCs have often been criticized for using technologies that are inappropriate if they are not cost-minimizing at current prices. This view may be misguided, according to Stiglitz, if these technologies have more learning potential and thus increase output in the long run.

Paul Stoneman (University of Warwick and CEPR) then presented 'Some Analytical Observations on Diffusion Policies', a paper which examined the issues that arise when investigating the impact of government policies designed to accelerate the adoption of new technologies. No single theory of technological diffusion seemed applicable to all situations, Stoneman argued. Instead there were a variety of hypotheses, each emphasizing a different aspect of economic behaviour, such as differences in the characteristics of potential adopters, the role of information and uncertainty, and the role of strategic behaviour.

Stoneman outlined these alternative models and the policy conclusions that could be drawn from each. The theory suggests, for example, that faster adoption of technology does not necessarily improve welfare, and that the characteristics and responses of suppliers of new technology are crucial in determining the impact of any particular technology policy. Expenditure on R&D and on the diffusion of new technology are inextricably linked, such that any technology policy aimed at one will affect the other. Stoneman also emphasized the importance of how expectations are formed, which affects both the desirability of particular technology policies and their impact.

In the ensuing discussion it was suggested that further insight could be gained by looking at recent empirical work and that the international dimension deserved more attention than it had received in Stoneman's paper. The issue of the availability of skilled labour as a hindrance to the diffusion process was also discussed. A brief but heated debate then followed on the private and social costs and benefits of the preannouncement of technological advances.

Bruce Lyons (University of East Anglia) opened the final session of the conference with his paper on 'The International Dimensions of Technology Policy'. Lyons surveyed the vast literature on the relationship between technology and international trade, and highlighted the policy conclusions that could tentatively be drawn from them.

Innovation or technological change taking place in a country can affect that country's terms of trade, the ratio of the price of its exports to that of its imports. It is possible, Lyons noted, that technological progress may be 'immiserizing', i.e. making a country worse off. This can occur if, for example, innovation reduces the price of those of a country's exports that are subject to inelastic demand: as the export price falls, revenues fall as well. If markets are competitive, for example, 'process innovation' in an export industry can be immiserizing as the terms of trade move against the innovator, although if all exporting countries collude an export tax can alleviate these problems. Lyons noted that 'product innovation' in export industries is much less likely to be immiserizing. The demand for the inputs required by the innovating industry increases, putting an upward pressure on export prices and improving the country's terms of trade. If royalties are charged on the transfer of process technologies, then the recipient of the technology can lose as the terms of trade move adversely. A tax on the transfer of product innovations to other countries can be justified on the grounds that it improves the country's terms of trade. For process innovations the gains from such a tax will depend on the extent of induced competition with imports, Lyons concluded.

If technological progress is exogenous, Lyons noted, the gains from trade are increased if the world contains countries which have very different economic structures. If progress is endogenous, on the other hand, it may be preferable to live in a world of similar countries, if this makes it more likely that technologies more appropriate to your requirements are invented. In the absence of the conditions for immiserization, rapid innovation benefits all countries, but a slowdown in innovation can lead in technologically advanced countries to an absolute reduction in the market-clearing wage level, and if wages fail to adjust it may cause unemployment. The pressures for a policy response in such circumstances are clear, Lyons argued, although the appropriate policy actions are not.

Economies of scale are inherent in the production of technology. This, Lyons observed, generates important gains from trade; it may also create imperfectly competitive markets. This may justify policy interventions, such as import protection or R&D subsidies, on the grounds that such policies shift profits to the home economy. Learning by doing may sustain the 'infant industry' case for protection, but if learning by doing is taking place abroad there may equally be an argument for an import subsidy! Lyons also observed that almost every conceivable policy intervention seems to beggar at least one of thy neighbours!

Some participants argued strongly that the other papers presented at the conference should have taken note of the international dimensions of technology policy. Lyons's paper, devoted to this subject, should have been unnecessary. Given that this was not so, his paper provided a good summary of the relevant literature, but it was generally agreed that the international economics literature does not satisfactorily deal with technology.

Standards have a significant bearing on the development and diffusion of new technologies and products, and the process of technological innovation obviously exerts a powerful influence on the structure of markets and the performance of industries. Technological standards and product standardization are therefore subjects of active concern for both business and government. So it is not surprising that issues of 'standards', although once quite neglected, have emerged since the mid-1970s as a focus of analytical and empirical attention among economists. When should a firm strive for standardization of product features and technologies, and when should it seek to thwart standardization? Can dominant firms choose technological standards so as to extend their own market power? How, if at all, will standards evolve from the interplay of competitive market forces, and how is the course of their evolution likely to be influenced by governmental intervention? What are the effects of a technological standard on the rate and direction of innovation in its area as well as in related technologies?

The last session included two presentations on the issue of standards. One by Jeffery Wheatley (British Telecom), 'Standards Issues in Telecommunications: A Prospect of the Industry', emphasized telecommunications aspects of the issue; the other, by Paul David (Stanford University), was entitled 'Some New Standards for the Economics of Standardization in the Information Age'. David examined the question of networks and interfaces and the ways in which levels of economic welfare may be affected by the manipulation of the 'interface' characteristics of products , i.e. those affecting the compatibility of components of existing and potential 'network technologies'. How might public policy interventions affect market resource allocation and influence the development and diffusion of emerging technologies?

David discussed the problems of formulating economically sensible standardization policies suitable for the era of the 'information revolution'. He considered a variety of direct and indirect governmental interventions to promote network compatibility, to achieve interchangeability among functionally equivalent components, and to foster systems integration by specifying standards for physical product characteristics, procedures, and other conventions. He argued that economists needed to adopt a non-standard way of thinking about how standards policies may affect the development of new technological systems.

David explored how a market economy may generate insufficient standardization, and also how the dynamics of a market economy may lock a technology into an inappropriate standard, such as the qwerty keyboard, he claimed. David examined the potential for government to play a role in setting standards and emphasized the importance of the historical aspects of standardization.