Asymmetric regulation and airport dominance in international aviation: evidence from the London-New York market

Southern Economic Journal, Oct, 2007 by Volodymyr Bilotkach

I ended up with the final sample of 3986 itineraries, a quarter of which are through itineraries. About 53% of tickets in the final sample represent travel to/from London; tickets to/ from Frankfurt constitute another 15%, and the remaining 32% are itineraries to/from Paris. Ninety percent of tickets in the final sample are associated with either CO (34%) or AA (56%). The remaining 10% are tickets of the second comparison carrier--Trans World Airlines. The average fare in my sample is $703.55, with a standard deviation of 522.29. The distribution of fares in the final sample is understandably asymmetric. In fact, only 9% of all itineraries are priced at or above $1000, with two thirds of tickets priced at or below $500.

Before proceeding with more careful analysis to control for airline and market-specific heterogeneities, I will present results of a simple "raw" data analysis. The purpose of this is twofold. First, I want to see whether the raw data point me toward supporting my research hypotheses. Second, this analysis both provides a clear first application of my identification strategy to data and helps me define the relevant variables to be included in regressions reported later. The summary statistic I am interested in is the average restricted economy class fare on selected routes. Results of the raw data analysis are presented in Table 3.

The following results are evident from Table 3. First, the three estimated airport dominance effects are rather close to each other, statistically significant, and equal to 27-29% of the average CO nonstop fare on a respective route (for example, the estimate of the airport dominance effect on New York-Frankfurt route--$209.78--is 28.9% of Continental's average New York-Frankfurt roundtrip economy class fare of $724.26). Second, the rough estimates of the difference in regulation effects are negative as expected but not statistically significant. One should, however, keep in mind that the identification strategy I employed might end up underestimating the difference in regulation effects. As noted above, one of the reasons such an effect may exist on the market in question is due to consumers' preference for Heathrow over Gatwick. In such a case, the term [[DELTA].sup.LONNYC.sub.Through] will include these effects. Thus, [[DELTA].sup.LONNYC.sub.Through] will be underestimated; hence, [[DELTA].sub.LONNYC] will be overestimated, and the difference in regulation effects will be biased toward zero. In the most extreme case where regulation effects are completely included into the through fares, [[DELTA].sup.LONNYC.sub.Through] will completely difference out the regulation effects, and [[DELTA].sub.LONNYC] will identify only the airport dominance effect. In fact, different values of [[DELTA].sub.Through] for different markets with the same comparison carrier (AA) suggest there are some factors not captured by the raw estimates reported in the above table (technically, [[DELTA].sub.Through] should only contain the difference in airline effects, assumed the same for all markets). Indeed, in obtaining the raw estimates reported above, I do not control for a number of important factors, such as distance, market size, and competition. This is the primary goal of the regression analysis that follows.