Market Structure and Multiple Equilibria in Airline Markets

We provide a framework for inference in discrete games that involve multiple decision makers and use it to study airline market structure in the US. We make inferences on a "class of models" rather that looking for point identifying assumptions that pin down a unique model. Making inferences in the presence of multiple equilibria is complicated since the underlying econometric model is incomplete: for a given set of parameters, the model does not predict a unique distribution for the outcomes. Strategies to deal with multiplicity have involved a variety of assumptions to narrow down the class of models into a unique one that point identifies the parameters of interest. We allow for a more flexible model of entry, heterogeneity and player identities without making assumptions on equilibrium selection. This estimation strategy nests within it a class of models that obey the fundamental assumption that if a firm enters a market it expects nonnegative profits. The weak assumption on the class of models comes at the cost of partial identification of the profit functions. The estimated features of the profits function can then be used to test for particular equilibria and predict market structure. We study the airline industry because of its characteristic wide dispersion in the number and identity of firms serving comparable city-pair markets. We might observe the wide dispersion because demand for airline travel is non-linear; because entry costs depend on the identities of the other entrants; because some markets are more attractive to smaller airlines than others; or because large network airlines strategically accommodate the entry of smaller carriers. We use our new econometric method to investigate the determinants of dispersion in the number and identity of firms across airline markets, allowing for both observed and unobserved heterogeneity among firms, and allowing for multiple equilibria to exist.