Antitrust Issues in International Comparisons of Market Structure1

The analysis and definition of markets, their structure, and concentration, especially for international comparisons, are complicated by a lack of adequate and comparable data. This is particularly so for multi-product and multinational firms. For instance, the U.S. Department of Commerce reports measures of industry concentration which do not embody either the control of subsidiary firms or the possible multinational nature of their ownership. Nor are these reported measures consistently based on sales data. Other countries produce similar reports, but these studies are generally not comparable to U.S. due, in part, to incompatible sector definitions. Few governmentsponsored studies provide firm-level detail or timely information. Also, given the widespread multinational nature of many larger firms, an international analysis of ownership and operations is necessary. This paper addresses the issues encountered in the construction of international market data from the existing financial reports, and provides methods for the comparison of measures of market concentration and industry diversity across countries. Using 1991 financial data, a firm level data set is constructed and used to compute comparable measures of market concentration and industry diversity in the food industries for the U.S. and European Community (EC). One innovation is to impute the distribution of sales of sub-code products by a firm based on simulations as well as nonparametric estimates of an existing data set. Introduction With the increasing globalization of the world’s economies, antitrust concerns for industries within a given country become increasingly complex. Traditional analysis of antitrust, market definition, and market power is focused on the “structure” of industries under scrutiny. One important component of the market power puzzle is the degree of concentration within an industry. Applied economists have understood for some time that measuring market power is not a one dimensional issue. Even before the advent of the so-called “New Empirical Industrial Organization” (henceforth NEIO) in the early 1980s, economists working on these matters understood well that one could not simply look at any one aspect of industrial behavior in a given industry, and draw meaningful conclusions about the level of competitiveness therein. Bresnahan (1989,1997) refers to the traditional empirical approach to analyzing competitiveness in a given market as the Structure, Conduct and Performance Paradigm (or the SCPP). That is, the traditional approach has been to define the relevant product and geographic markets, and then to examine structure by looking at the degree of concentration in the market and to look at pricing behavior to see if the firm(s) is (are) indeed pricing their outputs close to marginal cost. The implicit assumption in this body of work is that marginal cost is observable and measurable and that a reduced form analysis of structure and performance on cross-section data is sufficient, cf. Church and Ware (1997) for an initial critique of this approach. Church and Ware (2000, p. 239) note that under the SCP strategy, knowledge of market share is an important element in ascertaining the degree of market power within a given market. The NEIO approach emphasizes the fact that in general, economic marginal cost is not observable. In addition, each industry has its own nuances which distinguish it from others and a “conduct parameter” is an unknown to be estimated, not assumed in a cross-section model, cf. Bresnahan (1989, pp. 952-953). Bresnahan further notes that the NEIO approach focuses on the use of an econometric model for an individual industry, NOT on its reduced form and using data over time. The NEIO approach has been applied in practice in several applications. Ellison (1994) builds on work by Porter (1983) to show that demand for a given product (they focus on railroads) can be assumed to be log-linear in price and takes the form, log (Q) = α0 + α1 log (P) + α2 log (L) in this model α1 is the elasticity of own price demand, . The supply relationship can take the form, P (1 + Si wi / α1) = MC Where Si is the market share of the ith firm and wi is a “conduct” parameter. As in Church and Ware (2000, p. 441), let