The impact of passenger mix on reported 'hub premiums' in the U.S. airline industry.

• Author: Lee, Darin

1. Introduction

   One of the most actively debated issues in the U.S. airline industry since its deregulation in 1978 has centered around the prices charged by network airlines for service to and from their hub airports. This topic, known as the "hub premium" debate, stems from the belief held by many travelers that they are being overcharged by network airlines on flights to and from their respective hubs. (1) In addition to much anecdotal evidence, numerous U.S. Government studies (e.g., U.S. Department of Transportation 1990; U.S. General Accounting Office 1990; U.S. General Accounting Office 1999) have found that average fares at concentrated hub airports tend to be higher--often substantially--than at other nonhub airports. For example, the U.S. General Accounting Office (1999) reported that average fares at one hub were 83% higher than the national average, and another recent study (U.S. Department of Transportation 2001) went so far as to refer to hubs as "pockets of pain."

   The existing literature has studied and attempted to quantify the hub premium using--for the most part--one of two approaches. One set of studies (e.g., Morrison and Winston 1995; Morrison and Winston 2000) analyzes a cross section of airports and relates airport concentration to average fares. These studies improve on initial studies by the U.S. General Accounting Office (1990) and the U.S. Department of Transportation (1990) by attempting to control for factors known to impact average fares, such as average distance, the proportion of connecting passengers, and frequent flyer tickets. A second set of studies (e.g., Borenstein 1989; Evans and Kessides 1993) disaggregates data at the carrier and market level, thus controlling for structural differences between markets, and attempts to distinguish between market and airport characteristics as sources of potential pricing power. (2) In general, this latter group of studies argues that high concentration by a single airline at an airport can lead to entry barriers in the form of frequent flyer programs, travel agency commission overrides, and long-term leases on gates and airport facilities, among others, thus dampening nonstop competition and allowing the hub airline to charge supracompetitive fares. Borenstein (1989) also suggests that ownership of computer reservation systems (CRSs) serves as an entry barrier by allowing large network airlines to bias the information presented to both travelers and travel agents in favor of their own service. (3)

   None of the mentioned studies, however, has directly controlled for the passenger mix (the proportion of leisure versus business travelers)--which is known to affect average fares--and thus, the estimation results in studies such as Borenstein (1989) and Evans and Kessides (1993) may suffer from omitted variable bias. Network airlines have long argued that average fares are higher in markets to and from their hubs compared to other markets within their networks because a greater proportion of passengers traveling to and from their hubs are business travelers, purchasing more flexible--and much more expensive--unrestricted tickets. Unrestricted tickets offer a number of attributes that make them both more attractive to customers as well as more costly for airlines to provide. For example, unrestricted tickets may be changed without any fees, are fully refundable, can be purchased at the last minute, and, depending on the airline and/or the traveler's frequent flyer status, may provide the traveler with a free first-class upgrade. Because unrestricted tickets can cost several times as much as restricted coach tickets, network airlines argue that even a few extra unrestricted passengers per flight can have a relatively large impact on average fares. For example, an Expedia.com search for restricted, roundtrip, nonstop coach class tickets between Boston and Los Angeles yielded fares of $281, $295, and $295 on Delta, American, and United, respectively. For unrestricted coach class tickets on the same flights, the fare was $2583.50 on the same three carriers. (4)

   Most recent hub premium studies acknowledge that passenger mix may impact average fares, so it is somewhat surprising that only one study (Gordon and Jenkins 2000) has explicitly attempted to control for this factor. (5) Using proprietary data from Northwest Airlines, the authors found that after controlling for passenger mix, distance, and the number of stops, Northwest's hub passengers receive a hub discount compared to Northwest's passengers traveling throughout the rest of its network. This study, however, only focuses on Northwest Airlines, and its methodology has been subject to criticism (for example, the authors treat passengers connecting from one of Northwest's regional partners as hub-originating passengers). Thus, after more than a decade of debate, the hub premium controversy is still largely unresolved, a point echoed in a recent study by the Department of Transportation's Volpe Center:

   In analyzing these questions, researchers have employed various data sources and measurement techniques. Their studies also cover different time periods and use varying methods to control for the factors that influence an airline's costs and travelers' demands for their services in order to isolate the degree to which airlines can set and maintain fares above competitive levels. None of these studies, however, has successfully isolated or controlled for differences in airlines' costs or in passengers' willingness to pay for different levels of service. As a result, the extent to which airlines are able to use "market power" to maintain high fares for trips to and from their hub airports remains controversial. (U.S. Department of Transportation 2000, p. 2.) The purpose of this paper is to investigate the impact of passenger mix on the hub premium using fare data that have been disaggregated at the booking class level. To the best of our knowledge, this study is the first of its kind to use publicly available data that have been disaggregated at the fare class level. (6) At the outset, it is important to note that there is no universally accepted definition of the hub premium. Some studies are interested in the degree to which average fares (pooling all airlines) differ at hubs versus other airports (e.g., Morrison and Winston 1995; Morrison and Winston 2000), whereas others focus on comparing the prices of a network carrier's hub markets versus the prices of all airlines in otherwise similar markets (e.g., Borenstein 1989; Evans and Kessides 1993). In this paper, we define the hub premium explicitly as the ability of a given network airline to charge passengers in the same fare category more per mile in markets to and from its hubs than on otherwise similar markets throughout the remainder of its network (i.e., those markets that connect via one of its hubs or are between two nonhub airports). The prices in markets that neither originate nor terminate at a hub are widely considered to be good competitive benchmarks because the overwhelming majority of these markets are "spoke-to-spoke" markets that do not offer nonstop service but provide competing one-stop service from numerous hubbing carriers. Thus, our approach differs from most other hub premium studies in that our primary focus is on comparing prices across different markets for a given airline. By focusing our analysis on the prices of a given airline, we can effectively control for the variations in cost and quality of service across different airlines. Even among the six "full service" airlines we consider in this paper, costs and quality of service can differ significantly. For example, American, Continental, and Delta all use Boeing 737-800 aircraft in two-class configurations. However, Delta and Continental configure their 737-800 aircraft with 154 and 155 seats, respectively, whereas American uses 134 seats, thus increasing unit costs and improving product quality. (7) Moreover, it is important to emphasize that because the primary focus of our analysis compares fares within a given carrier, our main results do not depend on whether or not airlines use the same definitions of fare codes for Department of Transportation reporting purposes, a concern that has often been expressed by researchers relying on the U.S. Department of Transportation (DOT) data. For comparative purposes, we also estimate pooled ordinary least squares (OLS), two-stage least-squares (2SLS) regressions as well as fixed effects models similar to those of Borenstein (1989) and Evans and Kessides (1993).

   In general, we find that much of the observed hub premiums can be explained by passenger mix. Indeed, controlling for passenger mix reduced the average hub premium at major U.S. hubs (i.e., those hub airports where over 50% of passengers make connections) from 19.5% to 12.2%. Likewise, we find evidence that some carriers are successful at extracting additional hub premiums from first, business, and unrestricted coach (premium) passengers. In general, our findings are consistent with Ramsey pricing: leisure travelers have much more price-elastic demand than business travelers, and thus, the less convenient and lower quality connecting service offered by competing carriers is likely to have a greater disciplining effect on restricted coach fares than on premium fares for service to and from hubs. (8) Results from pooled estimation also show that indirectly controlling for passenger mix by using market fixed effects significantly reduces estimated hub premiums.

   Although passenger mix is an important element in understanding the hub premium debate, we stress that our analysis does not account for a number of other quality of service factors--and their associated time savings--that likely impact average prices at hubs such as greater flight frequency or preferences for nonstop service. Moreover, our analysis does not consider the effect of frequent flyer...