A New View of the Market Structure--Performance Debate

ion typically considered in market structure-performance studies, data on cross sections of firms or industries are unlikely to be sufficiently 2 The question of data selection has been discussed fairly thoroughly by Weiss [44] and Demsetz [I 5] among others, although there appears to be little agreement on the direction of bias in the most commonly used samples. The question of exogeneity has been less thoroughly discussed, although in specific instances (e.g., Schmalensee [37], Martin [27]) it has been dealt with explicitly. This content downloaded from 157.55.39.35 on Sat, 03 Sep 2016 05:57:53 UTC All use subject to http://about.jstor.org/terms 400 JAMES L. BOTHWELL, THOMAS F. COOLEY AND THOMAS E. HALL informative to make such empirical analysis useful. It may, therefore, be more productive to return to industry specific analysis. There will undoubtedly be some who are unconvinced by this argument and for them we offer the observation that further empirical studies based on weak deductive foundations are likely to be convincing only to the extent that they account explicitly for the role of prior beliefs and specification uncertainty. II. DETERMINANTS OF PROFITABILITY The conventional market structure conduct-performance paradigm has guided vast amounts of empirical research in the field of industrial organization for the past three decades. Like empirical research in other areas of economics, the maintained hypotheses that are considered reflect to a large extent the prior beliefs of the researchers. The result is that there is always a suspicion that what is being engaged in is more advocacy than objective empirical analysis. In the market structure-performance literature, the most important, alternative prior beliefs (maintained hypotheses) appear to fall into six categories which we characterize as follows: I. Seller concentration explains profits. High rates of return are thought to be caused by monopoly power conferred by high degrees of seller concentration. This is the basic collusion specification of Bain [I]. 2. Seller concentration and entry barriers explain profits. High rates of return are thought to reflect monopoly power resulting from both high seller concentration and high barriers to entry. This prior comes from the traditional market structure-performance paradigm of industrial organization and accounts for the peristence of collusive gains. It is prevalent in the work of Mann [26], Weiss [45, 46], Comanor and Wilson [I I], Qualls [34 [1j35] and many other researchers. 3. Absolute or relative firm size explain profits. This prior reflects the Demsetz [i4], [I5] and Peltzman [133] notion that high rates of return result from superior firm efficiency or innovativeness, and not from oligopolistic collusion and joint profit maximization. This prior has also been adopted by others (e.g., Shepherd [39], Gale [I7]) who assume that high market shares indicate high barriers to entry. 4. Risk differentials explain profitability differences. This reflects the notion that high rates of return reflect neither monopoly power nor superior efficiency, but the payment of positive risk premiums to risk averse investors. It underlies much of the work of Stigler [4I], Fisher and Hall [i6], Cootner and Holland [I3], and many others. There are, however, considerable differences in beliefs about which measure of risk variability of return over time, financial leverage, or nondiversifiable risk is relevant. This content downloaded from 157.55.39.35 on Sat, 03 Sep 2016 05:57:53 UTC All use subject to http://about.jstor.org/terms THE MARKET STRUCTURE PERFORMANCE DEBATE 40I 5. Advertising intensity explains profits. This reflects the notion of Bloch [4] and others that differences in reported accounting profit rates result from differences in the amounts of omitted, intangible capital resulting from advertising expenditures. Because advertising is assumed by some to be a source of product differentiation, this prior also may reflect the belief that the extent of product differentiation is a major determinant of profitability (Comanor [i o], Schmalensee [13 7). 6. Growth explains profits. This reflects Brozen's [6], [7] disequilibrium hypothesis that profitability differences result not from monopoly power, but from adjustments in capacity lagging behind changes in demand. Given a particular maintained hypothesis about the most important determinant of profitability, there remains uncertainty about the rest of the specification of the model. If the results are not robust with respect to uncertainty about the conditioning set of variables then it cannot be concluded that the data are very informative about the hypothesis in question. The next section discusses the methodology used to determine whether the results are robust to such specification uncertainty. III. SPECIFICATION UNCERTAINTY The hypotheses discussed in the previous section all specify the primary determinants of profitability. In each case, however, the investigator must have in mind a set of possible conditioning variables that may or may not belong in the regression. Although uncertainty about the inclusion or exclusion of such variables is not often explicitly acknowledged, it is an inevitable and important aspect of virtually all empirical investigations. The primary concern for the problem at hand is the extent to which this specification uncertainty leads to uncertainty in the estimated parameters of primary interest. Consider the regression model