| dc.description.abstract | Legacy carriers in the U.S. airline industry have a long history of vigorously defending their most important hubs from low cost carrier expansion. Since 2005, the U.S. airline industry has undergone some of the most dramatic merger activity in its history, with five mergers between major carriers reducing the number of major carriers from eight to four. This merger activity has coincided with low cost carrier expansion into some hubs previously dominated by legacy carriers. This dissertation quantifies how mergers change the incentives of incumbent legacy carriers to accommodate new entry. A technical challenge in doing so lies in the well-known “curse of dimensionality” for modeling dynamic strategic competition, which is especially prohibitive in the context of network industries (including airlines). In the airline context, this curse is induced by the high-dimensionality of airline networks, since firms make simultaneous decisions about route structures, flight frequencies, and prices across thousands of markets. We solve this challenge by proposing a novel method for studying high-dimensional dynamic strategic competition which combines tools from machine learning, the econometrics of dynamic games, and approximate dynamic programming. Using this tool to analyze the Delta and Northwest merger and Southwest Airline’s entry patterns, we find evidence that Southwest was more likely to enter flight segments where, from Delta and Northwest’s perspective, the expected value of committing aircraft capacity, relative to other flight segments, fell the most post-merger. Outside of the airline context, we further illustrate this method by studying a dynamic spatial store placement game among big box retailers (including Walmart), extending the analysis of Holmes (2011). Finally, in an unrelated context, we systematically study the identification of the average treatment effect on the treated under the difference-in-differences design in the context of repeated cross-sectional data when post-treatment treatment status is unknown for the pre-treatment sample. We illustrate our approach by estimating the effect of the Americans with Disabilities Act of 1991 on employment outcomes of the disabled. | |
