Recent developments in the theory of international trade have suggested that the scope for interventionist trade policy may be greater in imperfectly competitive markets than is suggested by traditional analyses of international trade under perfect competition. Doubts have been expressed both about the empirical robustness of some of the theoretical models and about the likely magnitude of the effects of policy interventions. These questions can only be resolved by confronting theoretical models with observed data. CEPR and the National Bureau of Economic Research (NBER) have since 1986 carried out a joint programme of research on `Empirical Studies of Strategic Trade Policy' with this objective. The programme is funded by the Ford Foundation, with additional funding from the German Marshall Fund of the United States. Alasdair Smith (University of Sussex and CEPR) and Paul Krugman (MIT, CEPR and NBER) organized a workshop at the University of Sussex on July 8/9 to discuss work in progress in the research programme.
Two papers presented at the workshop were concerned with international competition in the market for commercial aircraft. In `Simulating Competition in the Market for Large Transport Aircraft', Gernot Klepper (Kiel Institute for World Economics and CEPR) focused on the role of economies of scope, that is, the effect on a firm's costs of the size of its product range. Klepper presented a simple model which he used to analyse the development and the production of large passenger aircraft. His model avoided many real world complexities, ignoring the length of the product cycle and competition between large and small aircraft, and supposing that decisions to purchase aircraft are made at only one point in time.
There are two firms in Klepper's model, each with a production technology characterized by significant fixed costs and marginal costs which fall as a result of learning-by-doing. Klepper lacked reliable estimates of the important parameters in the model. In particular, the extent of substitutability between aircraft and other inputs in the production of air transportation services is a key parameter in the analysis, but its value is difficult to estimate. Klepper assumed that each firm takes the other's sales levels as given when making its own sales decisions (Cournot behaviour). He found that the characteristics of the equilibrium under this assumption were very sensitive to changes in the parameters of the model. He then developed a version of the model in which firms produced more than one type of aircraft and benefited from economies of scope, and he found that such economies have important effects on the equilibrium of the model. Klepper concluded that further development of the model was required before it could be used to analyse the effects of government policy in the aircraft market.
Paul Krugman (MIT, CEPR and NBER) also focused on the same industry in `Problems in Modelling Competition in the Aircraft Industry', written jointly with Lael Brainard (MIT). Krugman argued that it is empirically inappropriate to treat the aircraft industry as characterized by small-group competition, huge development costs and a steep learning curve. Other features of the industry were also important, according to Krugman and Brainard: aircraft are long-lived capital goods, and there are long-term supplier-customer relationships between aircraft manufacturers and airlines.
The authors argued that, as a result, competition in the aircraft market might best be modelled by a variant of the `contestable market' theory. This hypothesis suggests that potential entry into the industry causes existing firms to set prices at a level determined primarily by the estimated average costs of potential alternative producers. Aircraft capital costs represent only around 20% of the total costs of running an airline and there is little scope for substitution between aircraft and other inputs entering into the production of air travel services. These factors, in conjunction with estimates of the overall demand for travel, suggest that the elasticity of demand for aircraft is low. Why then, in an industry with small numbers of firms and very inelastic demand, are prices not well above costs of production? Krugman and Brainard suggested that the lowest-cost firm in the market sets a price that is constrained by the average cost that the next-lowest-cost firm could have achieved if it were to have taken the market. The lowest-cost firm then sells all it can at that price. After the firm has captured the market it does not attempt to exploit its monopoly position by increasing prices: the long-term relationship between aircraft producers and airlines gives the firm an incentive to maintain its reputation. If this description of pricing behaviour were correct, there is no case to be made for government subsidies to encourage entry into the industry on the grounds of excess incumbents' profits, since any profits that can be made by such a policy are less than the required subsidy. There is a case for forcing down incumbents' prices in the interests of consumers, but that would best be done without incurring the costs of a new entrant.
Discussion of these two papers explored several aspects of the models used. Was it plausible to assume that an intrinsically dynamic model of production and competition could be `collapsed' into a static model in which pricing and production decisions were made at the start of the production process? Questions were also raised about the sensitivity of the models' behaviour to assumptions concerning their parameter values: whether economies of scope should affect both fixed and variable costs, whether con ventional wisdom on the size of learning effects is reliable, and whether demand elasticity is likely to change over time. Responding to questions about the different market structure assumptions in the two papers, Krugman argued that Klepper's Nash-Cournot assumption was inconsistent both with the apparently low demand elasticity and the price competition that seem to characterize the industry. Klepper, however, maintained that the role of market share forecasts in firms' plans and the fact that Airbus Industrie, at least, faced capacity constraints, provided evidence supporting his Cournot assumption, i.e. that firms take rivals' sales levels as given.
Richard Baldwin (Columbia University and NBER) examined the techniques used in recent work on the empirical evaluation of strategic trade policies in his paper, `On Taking the Calibration out of Calibration Studies'. Such evaluations, Baldwin noted, are inherently difficult for three reasons: there is no general agreement on how to model imperfect competition; important data on costs, prices and market shares at the firm level are often unavailable or unreliable; and industrial targeting policies can result in such dramatic changes in industry structure that standard empirical tools are of little use. The models used in empirical studies of trade policy are based on theoretical considerations, but there is often little information available concerning the values of the para- meters in these models. As a result, researchers `calibrate' the model by choosing parameter values such that the model reproduces the observed data in a base year. These parameter estimates, Baldwin remarked, are based on only a minimal amount of data and thus can be subjected neither to statistical analysis nor even to a satisfactory sensitivity analysis.
Baldwin discussed how his earlier study (with Paul Krugman) of the semiconductor industry could be modified to take account of these criticisms, although he stressed that his work was still at a preliminary stage. The industry is characterized by strong `learning' effects: the yield of usable semiconductor chips from a particular plant rises with the cumulative production of chips by the plant. Baldwin used data on the US market for semiconductors and on sales by individual firms to estimate the parameters both of the demand curve and of the producers' yield curves; these estimates differed substantially from the parameter values used in the earlier study. Baldwin suggested that the standard errors of the parameter values could be used to construct confidence intervals for the welfare results.
Discussants of Baldwin's paper agreed that in general one should use more information when it is available, but they worried about the difficulties that could result. For example, when parameters are estimated on the assumption that both Japanese and American producers find it profitable to produce, there is an inconsistency if the model estimates then imply that only one set of producers will in fact enter the market.
Richard Harris (Queen's University, Kingston) presented a paper written jointly with Victoria Kwakwa (University of Regina) on `The 1988 Canada-United States Free Trade Agreement: A Dynamic General Equilibrium Evaluation of the Transition Effects'. Harris and Kwakwa focused on the medium-term effects of removing trade barriers between the United States and Canada using a numerical general equilibrium model. Trade-induced unemployment can arise in their model, so that the adjustment costs of trade liberalization can be brought into the economic evaluation of free trade. As in Harris's earlier work on the evaluation of trade policy changes, the model used was one in which firms are imperfectly competitive and have economies of scale; but as a result of sluggish wage adjustment there is both frictional and natural unemployment in the labour market. To simulate the effects of the agreement, Harris and Kwakwa assumed that trade barriers would be gradually reduced to zero over a 10-year period and that the full effects would appear over a 20-year period. The simulations suggested that trade liberalization would cause employment to rise and lead to entry and exit of firms at a modest rate. Trade volumes rise, with Canadian imports rising more than exports (implying that some currency depreciation would be required to restore current account equilibrium). Free trade has little effect on labour reallocations, and it raises both real and nominal wages. The authors also carried out sensitivity analyses with respect to some of the key parameters of the model. Harris noted that their results were quite different from those obtained in the Michigan model of world trade, which is based on the assumption of competitive firm behaviour, so that it seemed that introducing imperfect competition has a dramatic effect on the evaluation of the effects of trade liberalization.
In `Industrial Organization and Product Quality: Evidence from South Korean and Taiwanese Exports', Dani Rodrik (Harvard University) examined the transition by industrializing countries from producing standardized, labour-intensive manufactures to more sophisticated, skill-intensive goods. Rodrik analysed the hypothesis that such a transition is more easily achieved when the domestic industry is highly concentrated. Firms cannot easily provide accurate signals to export markets of the true quality of their product, and perceived quality depends on other firms' quality as well as the true quality of the firm's own product, it is argued. This gives rise to an externality which discourages firms from investing in quality: improvements in quality may not generate major export sales unless other firms invest in improved quality as well. It is easier to overcome this externality in more concentrated industries.
The comparison of South Korea and Taiwan provided a test of this hypothesis since these two countries are similar in many respects but have radically different industrial structures, with Korean industry much more concentrated than in Taiwan. Rodrik obtained evidence on differences in product quality from data on the unit value of the two countries' exports to the United States. He compared 49 fairly disaggregated product classifications in which at least one of the two countries had exports to the US exceeding $100 million. In 30 out of the 49 classes, Korean unit values were higher, a statistically significant difference. Further, using Japanese unit values as a measure of product quality, Rodrik found that the difference between the Korean and the Taiwanese unit value premia increases with product quality. He concluded that the results have no normative implications but do suggest the desirability of further research on the links between concentration and industrial development.
Richard Harris observed that even at the disaggregated product level used in Rodrik's paper, there is a substantial problem of product heterogeneity within a single classification. Other participants noted differences between Korea and Taiwan apart from their degree of industrial concentration <196> in their distribution systems, in the pattern of foreign ownership of firms, in their financial markets, in the degree of government involvement in promotion of high technology sectors and in the trade restrictions they face in the US market. These differences might also explain observed differences in export unit values.
Klaus Zimmerman (Mannheim University) investigated the relation between domestic and export prices in `Relative Export Prices and Firm Size in Imperfect Markets', co-written with Lorenzo Pupillo (University of Pennsylvania). If domestic and foreign markets are segmented by transport costs, and if market power has no influence on price-cost margins, one would expect firms to charge higher prices abroad than in at home. Zimmerman and Pupillo tested this hypothesis using the results of a large-scale Italian survey of firms' pricing behaviour. This evidence indicated that prices were higher in the home market. Zimmerman interpreted this as indicating that market power affects margins and that firms face less elastic demand in foreign than in home markets. Further, increasing firm size and market concentration tend to be associated with lower relative export prices, while higher export shares are associated with higher relative export prices.

Paul Krugman commented that in a market for a homogeneous product with Cournot behaviour by firms, the price-cost margin should depend on demand elasticity and the firm's share, so it is not clear why firm size should be associated with margins. The effect of exchange rate changes on the results was also raised, but Zimmerman argued that all firms faced the same exchange rate so that exchange rate effects were unlikely to generate significant differences between firms. It was also noted that Zimmerman's analysis faced `simultaneity' problems: he had argued that market shares influenced the relationship between export and domestic prices, but shares in both home and foreign markets were clearly determined by firms' pricing behaviour.
The workshop concluded with two brief presentations of work in progress. David Ulph (University of Bristol and CEPR) and L Alan Winters (University of Wales and CEPR) discussed the implications of introducing inter-sectoral mobility of skilled labour into numerical models of imperfect competition, along the lines suggested in theoretical work by Dixit and by Grossman. Anthony Venables (University of Southampton and CEPR) discussed joint work with Michael Gasiorek (University of Sussex) and Alasdair Smith (University of Sussex and CEPR) on tariffs, subsidies and retaliation in a numerical model of imperfect competition. Preliminary results from a four-country model (in which the `countries' were the EC, the USA, Japan and the rest of the world), applied to two industries, indicated that while imperfect competition implied a welfare-increasing role for import tariffs and export subsidies, that role was greatly diminished in the presence of effective competition policy.