What's the Difference? Measuring the Effect of Mergers in the Airline Industry

Abstract

We analyze the effect of four US airline mergers using three retrospective techniques: standard difference-in-differences regression, synthetic control, and nearest neighbor matching. In doing so, we compare the performance of the most recently developed merger-retrospective methodologies. Each method compares an outcome of interest, such as price, in a set of ``treated'' markets where a merger occurred to a set of ``control'' markets that are unaffected by the merger. The three methodologies, however, differ in how they select and weight control markets. We find that the three methods do not always align in the direction or statistical significance of the estimated merger effect. Furthermore, synthetic control and nearest neighbor matching are sensitive to the set of characteristics used to select control markets. Overall, we find that the American Airlines/USAir merger lowered prices, whereas the United/Continental merger increased prices. Even so, these findings are not robust across methodologies or specifications. We are unable to draw any consistent conclusions about the Southwest/AirTran or Delta/Northwest mergers.