Industrial organization in the laboratory

Theories of industrial organization (IO) are tested in the laboratory more and more often. The example we consider throughout the paper is oligopoly theory, specifically, the Cournot and the Stackelberg model of duopolistic quantity competition with homogeneous products. These models have often been subjected to tests: participants were told that they represented firms and received profits according to their chosen quantities. However, testing a theory in the laboratory requires that the experimental design falls into the domain of the theory. The domain of mainstream IO is market behaviour of large (i.e., multi-agent) firms. The basic hypothesis is that these firms maximize profits. This hypothesis cannot be tested in experiments where single agents are told that they represent firms. For a test, it would be necessary to assume that multi-agent groups and single agents show the same behaviour. This assumption is inconsistent with all current theories of individual behaviour and has, moreover, been falsified in many experiments. It follows that many alleged experimental tests of IO models are irrelevant to mainstream IO. This raises the question of whether relevant laboratory tests are possible at all. The answer is positive if one considers theories of the firm that relate organizational behaviour to the internal structure of the firm, like Alchian and Demsetz’s (1972) contractual view of the firm.