| This dissertation addresses several related empirical topics, using evidence from the U.S. airline industry in the late 1980's. In the first chapter, “The Price Effects of Mergers,” I summarize the changes in airfares after the largest airline mergers of the period. I find significant price increases on routes where both merging firms competed prior to the merger, with more ambiguous effects on other routes. In the second chapter, “Demand and Conduct in the U.S. Airline Industry,” I estimate an empirical model of demand and supply, to identify alternative sources of market power. The results suggest that the demand curve facing an individual firm is significantly more elastic than found previously, implying that unilateral market power may have been overstated. There is mixed evidence about the importance of capacity pre-commitment, while collusive conduct seems to be quite limited. In the third chapter, “Evaluating the Performance of Merger Simulation,” I use the model to predict post-merger prices for each airline merger, and compare these predictions with actual post-merger prices. Combining post-merger information with merger simulations allows us to decompose actual price changes into their component effects. The results suggest that standard simulation methods, which measure the effect of ownership transfer on pricing incentives, can account for a large component of the post-merger price change, but should not be expected to account for all of it. Changes in marginal cost or firm conduct generally account for most of the remainder, while post-merger entry and changes in demand-side variables have a relatively small effect. In the fourth chapter, “Estimation of a Model of Entry in an Airline Network,” I use observed entry and exit decisions to estimate a model of airline profits, explicitly taking into account the network structure of the industry and the interdependence of entry and exit across different segments. The empirical strategy is to treat the wave of airline mergers as an exogenous source of variation in firms' networks. The performance of the estimator is decidedly mixed, but it appears to yield more reasonable results than a naive estimator which ignores the endogeneity issue entirely. |
