A joint CEPR workshop with the Wissenschaftszentrum Berlin für Sozialforschung (WZB) was held on 16/17 December on `The Empirical Analysis of Industrial Organization'. The workshop formed part of CEPR's research programme in `Market Structure, Industrial Organization and Competition Policy' funded by the European Commission's SPES programme, and addressed methodological obstacles to progress in empirical and sectoral studies. It was organized by Damien Neven (Université de Lausanne and CEPR) and Lars-Hendrik Röller (WZB and CEPR).

The first paper was on price adjustments which economists tend to associate with either fixed or variable transaction costs: fixed costs provide incentives to adjust prices infrequently and by large amounts, whereas variable costs provide incentives to adjust prices continuously. The importance of these costs should not be underestimated since adjustment costs associated with imperfect competition are sufficient to imply monetary non-neutrality in many macroeconomic models. In `Optimal Pricing with Costly Adjustment and Persistent Effects: Empirical Evidence', Margaret Slade (University of British Columbia and GREQAM, Marseilles) investigated empirically the existence, type and magnitude of such costs through a discrete-time dynamic programming model. The model differs from previous ones in the adjustment cost literature in several respects: it tests fixed and variable adjustment costs to determine their relative importance; model dynamics are not due to inflation or changes in the money supply, but result from the stocks of goodwill that a firm can accumulate or dissipate; the thresholds that prices must cross before a change is profitable, as well as the costs that firms incur when a change occurs, are estimated; and the model is both structural and dynamic. The data used in the model consist of weekly prices and sales of four brands of saltine crackers sold by four grocery chains in a small US town. The estimation revealed that the average cost per price change was $2.70 of which approximately 95% was fixed and 5% was variable. With such high costs, more than the weekly profit of a typical brand, managers cannot afford to fine tune their prices.

Lars-Hendrik Röller emphasized the value of the stock of goodwill in generating dynamics in the model, as well as the inadequacy of exogenous shocks. Paul Geroski (London Business School) asked if the price changes could not be explained by in-store promotions, and whether it was the retailer or the brand manager who was making the adjustment decision. Slade answered that advertising played a role but could not account for price variability since advertising only occurred 5% of the time while prices moved 20% of the time. She added that the identity of the decision-maker did not matter as neither can ignore the costs incurred by the other. Ernst Berndt (MIT) noted that the costs of adjustment should come down with advances in technology, such as wider use of bar codes.

In a paper written with Linda Bui, David Reiley and Glen Urban, `The Roles of Marketing, Product Quality and Price Competition in the Growth and Composition of the US Anti-Ulcer Drug Industry', Ernst Berndt examined the roles of prices, product quality, order of entry and various marketing efforts in this market. His study concentrated on the marketing of anti-ulcer drugs to physicians through retailing and medical journal advertising, distinguishing `industry-expanding' and `rivalrous' efforts. The empirical assessment revealed that the impact of all marketing expenditures on overall market size decreases as the number of products in the market increases. It also suggested that cumulative minutes of retailing and cumulative pages of medical journal advertising affect aggregate as well as individual sales. The results showed that the stock of industry-expanding marketing declines at a near zero rate while the stock of rivalrous marketing depreciates at an annual rate of 40%. Other factors with significant effects on sales of these drugs include price, order of entry into the market and quality attributes such as indications of FDA approval.

Len Waverman (INSEE and University of Toronto) focused on the observation that Zantac's sales were mostly due to early FDA approval for the GERD disease indication. He also suggested investigating pre-entry strategy as Tagamet had a six-year monopoly position before Zantac came into that market. Paul Geroski wondered why dosing frequencies did not affect market shares in a statistically significant way, because it would seem to be of considerable interest to the consumer. Berndt suggested that dosage frequencies should be modelled as a discrete variable and the lack of significance of this variable in his estimates may be due to an inadequate specification. Damien Neven suggested that the remaining life of patents should be taken into account for the time profile of prices. This may account for the otherwise puzzling observation that price increases are concomitant with entry.

In a paper with Robert Majure, `An Empirical Analysis of Bank Branching: Portugal 1989–91', Luis Cabral (Universidade Nova de Lisboa and CEPR) developed a simple oligopoly model to try to explain the phenomenal increase in the total number of bank branches in Portugal since the liberalization of this sector in the mid-1980s. His results showed that customer loyalty is a significant substitute for efficiency as the expected profit margins are higher for incumbents than for entrants, even though incumbents do not have an absolute advantage on &nbsppure &nbspefficiency &nbspgrounds. Because of loyalty, incumbent banks have a relative and absolute advantage in rural markets whereas entrants have a relative and absolute advantage in urban markets. Overall, however, no group has an absolute advantage over the other. The results also revealed that `number of branches' is a strategic substitute for some banks and a strategic complement for others. The pattern of strategic complementarity is not sufficient to explain &nbspthe large &nbspexpansion by incumbent banks, however.

Damien Neven warned against potential mis-specification arising from a focus on the number of branches as the strategic variable when competition simultaneously takes place in other dimensions such as credit competition. He also noted that the empirical results could not account for the rapid expansion of branches, and suggested following an alternative approach which would take more account of localized competition. Cabral acknowledged that the expansion could not be explained by his model, but he emphasized that it provides a rationale for the expansion of branches which is independent of entry. Margaret Slade proposed using a dynamic model rather than a static game to capture the behaviour of this sector more effectively. Paul Geroski suggested estimating the model over a longer sample, including a period without any increase in the number of branches to check whether the model could account for the rapid expansion. Ernst Berndt mentioned that part of that expansion was probably due to the fact that Portugal was relatively underbanked.

Empirical work analysing flows of entry into different markets, while presenting evidence that many new firms start up every year in most developed economies, seems to have reached the conclusions that new firms are small and compete with other small firms and that entrants exit before they can really become a threat to market leaders. In `The Survival of New Plants: Start-up Conditions and Post-entry Evolution' written with Pedro Portugal and Paulo Guimarães, José Mata (London Business School, Banco de Portugal, Universidade Nova de Lisboa and CEPR) examined the longevity of entrants and the process that leads new firms to exit after entry, using a dataset of new plants created during the 1980s in Portuguese manufacturing. The results revealed that size is an important determinant of the probability of survival and that current size seems to be a better predictor of failure than initial size: firms that have grown have a lower probability of failure. This observation is consistent with theories that emphasize post-entry learning as an important determinant of firm performance. After controlling for size differences, the results also showed that past growth matters for survival, suggesting a partial adjustment process for firm size. Lastly, the competitive environment provides some explanations for the longevity of new firms: the study suggests that entrants are more likely to live longer if they enter growing industries or industries with little entry activity. Paul Geroski commented on the relative importance of the death of new firms. He suggested that whether new entrants exit the market relatively quickly may not be really interesting. Whether failed firms embody new talents and ideas and whether these survive, are, in his opinion, questions of more interest.

In a paper written with Jean-Jacques Laffont, Alain Oustry and Michel Simioni, `Econometrics of Optimal Procurement Auctions', Quang Vuong (University of Southern California) developed a model to analyse both the way in which a public agency should arbitrage between the quality of a project and the bid price, and the bids that firms should submit to the public agency. In his model, implemented on auctions of software equipment managed by the French National Space Agency, firms bid prices based on the different qualities of their projects. The model postulates that the choice rule of the public agency optimizes social welfare as a function of the bid prices, the offered qualities, and the social value of a given project. The optimal bidding strategies of the auctioning firms can then be derived. The data collected for the estimation part of the model consist of 69 auctions of software equipment that occurred from 1984–93 where the number of bids varies between 2 and 5 (for a total of 203). The derived structural econometric model is a non-parametric framework with a discrete part, the agency's choice, and a continuous part, the firm's bid. It can be shown that the structural elements of the model, which consist of the cost function for realizing the project, the distribution of firms' cost characteristics, the social value of the project and the cost of public funds, are non-parametrically identified from observations on firms' bids, firms' offered qualities and the public agency's choice. An estimation of the social values of the projects, the firms' cost functions, and their hidden costs characteristics can be derived through a computationally simple multi-stage non-parametric estimation procedure.

After noting the importance of the method used in this analysis, John Panzer (Northwestern University) stressed that the model may be appropriate for France, but that the Laffont-Tirole incentive conditions would be unrealistic in the US because of the collusion that would inevitably occur in such procurement auctions. Damien Neven noted that collusion may be likely even in France given that a small number of firms (approximately 8) repeatedly bids in the 69 auctions.

In `Margins and Risk Premia in Bookmaking' Patrick Waldron (Trinity College Dublin) and John Fingleton (Trinity College Dublin and CEPR), attempted to explain the substantial differences between the margins apparently earned by bookmakers in Ireland and the UK. Using total per cent and overround per runner as accepted measures of bookmaking market performance, their study provided a theoretical explanation of both the level and variation of these measures as well as an empirical analysis of the behaviour of these variables in Ireland in 1993. They used a four equation system consisting of a demand equation for bets from bookmakers, a supply equation of bookmakers, another demand equation for bets from the tote (also known as parimutuel or pool), and a fourth equation describing the evenness of match. Empirical tests may be implemented to distinguish between three possible explanations of the differences between the Irish and the UK markets: that there might be anti-competitive behaviour in the Irish market, that the problem of insider trading might be greater in Ireland and that risk aversion could be more important for the Irish bookmakers. David Audretsch (WZB and CEPR) expressed his doubts about the ability of the model to capture the differences between margins in Ireland and the UK. Damien Neven noted that bookmakers have information that other market participants have not, and that the model should take account of strategic behaviour on the part of &nbspbookmakers in &nbspthe &nbsppresence &nbspof asymmetric information.

In a paper written with Sangeeta Kasturia and Jürgen Müller, `A Study of the Variable Cost Structure of the Deutsche Bundesbahn', Ronald Braeutigam (Northwestern University) characterized operating costs for the Deutsche Bundesbahn over the period 1958–90 by estimating a variable cost function, within an equilibrium of intermodal competition. The cost structure recognizes two outputs: the levels of freight and passenger services. The study was partly motivated by the prospect of a reorganization of the Deutsche Bundesbahn into three entities of &nbspwhich two, passenger and freight services, would be privatized. In empirical tests for the presence of cost complementarities and economies of scope, it revealed that there is no statistically significant evidence of either. In addition, track electrification has led to an increase in operating costs, indicating that while electrification &nbspmay &nbsphave been desirable on environmental grounds, it has actually increased total costs of the Bundesbahn. Technological change, other than that associated with electrification, has reduced variable costs at an annual rate of about 2% per year since 1958. Ionnes Kessides (World Bank) emphasized the importance of analysing economies of scope through different and very flexible cost functions and noted the scarcity of this type of study in the empirical industrial organization literature.

In `An Empirical Oligopoly Model of a Regulated Market' written with Jordi Jaumendreu, Consuelo Pazó (Universidad de Vigo) specified a theoretical and econometric framework to examine the behaviour of a three-firm oligopoly that controlled the nitrogenous fertiliser supply in Spain from 1976–88. The author explicitly modelled regulation in this industry since it was subject to price ceilings. Both aggregated and disaggregated versions of the model were developed to estimate the likelihood that observed market outcomes could be generated by different behavioural equilibrium concepts such as market sharing agreements in the frontier of profits, Cournot and Stackelberg. Restricted versions of each equilibrium were examined to determine the degree to which regulation constrained prices. The empirical estimation identified the regulation-constrained Stackelberg equilibrium as the most likely outcome. Welfare analysis also showed that given observed price-cost margins, the consumer surplus transferred to the firms and the deadweight welfare loss were significant. Maria Maher (Birkbeck College, London) stressed that the sample period may be too short to identify the subperiod where the regulation was binding. Margaret Slade expressed some doubts about the regulation constraint and suggested incorporating it into the model in an alternative way.