GAME THEORY AND INDUSTRIAL ORGANIZATION: AN ABC

It has become a familiar observation in recent years that the literature on game-theoretic models in industrial organization faces a serious dilemma. The richness and flexibility of this class of models provide a framework within which we can “rationalize” a huge range of possible “equilibrium outcomes.” Whatever the phenomenon, we seem to have a model for it. Should we see this as a success, or as an embarrassment? Does this body of theory allow any outcome to be rationalized? After all, the content of a theory lies in the set of outcomes which it excludes. Judged on these terms, is the enterprise empty?>

The huge range of outcomes that can be rationalized can be traced to two features of these models. First, many of the models in this literature have multiple equilibria. Second, the appropriate specification of the model is rarely obvious. (Is competition to be à la Bertrand, or à la Cournot? Should entry be modeled as simultaneous, or sequential?) Sometimes it is possible, by referring to the observable features of some particular market, to decide in favor of one model specification over another. In other cases, however, the features that distinguish candidate models must be treated as unobservables, at least from the point of view of the modeler. Both these features tend to widen the set of outcomes that may be rationalized as equilibria (figure 4.1).