Abstract
This paper provides a methodology for deciding which airports warrant grouping in multi-airport metropolitan areas. The methodology is based on the comparability of incremental competition effects from nearby airports on average fares at a metropolitan area’s primary airport. Results from a quarterly panel data set for the period 2003–2009 provide strong evidence that city-pairs, rather than airport-pairs, are the appropriate market definition for analyses of passenger air transportation involving many (but not all) large metropolitan areas. Based on the proposed method, we offer a recommended list of airports that should be grouped when creating city-pairs for the analysis of competition in the US domestic airline industry.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Author preferences tend to dictate the choice between city-pairs and airport-pairs, with the literature offering no systematic, formal method for choosing between the two approaches to market definition. Empirical papers using the airport-pair approach include Borenstein (1989, 1991), Brueckner et al. (1992), Brueckner and Spiller (1994), Brueckner and Whalen (2000), Brueckner (2003), Whalen (2007), and Armantier and Richard (2006, 2008). The city-pair approach is followed by Berry (1990), Werden et al. (1991), Evans and Kessides (1993, 1994), Bamberger et al. (2004), Boguslaksi et al. (2004), Berry et al. (2006), Peters (2006), Ito and Lee (2007) and Gayle (2008). Studies focusing on the fare impact of low-cost-carrier (LCC) competition (Morrison 2001; Goolsbee and Syverson 2008; Brueckner et al. 2013) take a hybrid approach that uses airport-pairs as the unit of observation while controlling for competition at adjacent airports within a metro area.
For example, at the time that the US Department of Justice conducted its review of the recent United/Continental merger, the merging parties were the only two carriers offering non-stop service between Newark Liberty International and San Francisco International airports, and carried over 95 % of the O&D passenger traveling between these airports. However, under the city-pair approach, there was an abundance of competition between the New York and San Francisco metro areas, with non-stop service that was offered by four additional carriers (American, Delta, JetBlue and Virgin America). Counting all of this service, the merging carriers’ O&D passenger share stood at only 37.6 %.
We also emphasize that our paper addresses only the question of geographic product market definition and that policy makers are often interested in other questions of market definition as they apply to the airline industry (for example, potential distinctions between leisure and business or “time-sensitive” passengers as well as non-stop versus connecting travel.) For a study of non-stop versus connecting passenger market definition, see Gayle and Wu (2011b).
This notion of competitive spillovers is closely related to the concept of market definition that is outlined by the US Department of Justice and Federal Trade Commission in their August 2010 Horizontal Merger Guidelines, which state: “In considering likely reactions of customers to price increases for the relevant product(s) imposed in a candidate geographic market, the Agencies consider any reasonably available and reliable evidence, including: how customers have shifted purchases in the past between different geographic locations in response to relative changes in price or other terms and conditions...” and “... evidence on whether sellers base business decisions on the prospect of customers switching between geographic locations in response to relative changes in price or other competitive variables...”.
For example, Phoenix’s Sky Harbor (PHX) and Mesa (AZA) airports are within relatively close proximity to one another (i.e., 33 driving miles) but there is insufficient service at AZA to perform our econometric analysis. A similar lack of data prohibits application of our method to other multiple-airport metropolitan areas, such as Orlando (MCO, SFB), Philadelphia (PHL, ILG), and Fort Myers (PGD, RSW).
We choose 1990 because it marks the start of the widespread expansion of LCCs such as Southwest Airlines, which traditionally relied heavily on serving “secondary airports” such as Chicago Midway (rather than O’Hare), Baltimore-Washington International (rather that Washington National or Dulles) and Oakland (rather than San Francisco). Thus, in many respects, 1990 marks the start of the airline market-definition debate [see Bennett and Craun (1993)]. By 2009 (the last year of data in our sample), each of the primary airports, with the exception of Miami and Washington, DC (where Fort Lauderdale (FLL) and BWI now have more domestic O&D passengers), still handled the most passengers.
We limited the fringe airports to those major airports that were less than 75 miles from either the primary airport or the city-center. Table 5 in the appendix shows the distances from each airport to the primary airport and to the city-center. Table 6 shows summary statistics for fares on routes from each primary airport.
Although connecting service represents another form of competition with the potential to discipline nonstop fares, we elected not to control for such competition because of potential endogeneity concerns. However, including as a covariate the percentage of passengers on the route using connecting service does not materially change any of our primary results.
All of the grouping recommendations contained in Sect. 3 are robust to the inclusion of routes that are less than 200 miles.
Our sample is non-directional, so that LAX-JFK, for example is considered the same route as JFK-LAX. Fare data is for roundtrip and one-way travel on non-stop tickets. Given the well-documented lack of consistency of the fare-class variable across carriers in the domestic DOT data, tickets from all fare-classes are included. We exclude from our sample the following types of itineraries: non-revenue itineraries, itineraries with directional fares including taxes and fees of less than $25 or that were coded in the DOT Survey data as “bulk” fares, and interline itineraries (i.e., those with more than one marketing carrier).
For the purposes of our study, we define the legacy carriers to be American, United, Delta, Northwest, US Airways, America West, Alaska, Continental, and Midwest. The LCCs are defined as Southwest, AirTran, JetBlue, Frontier, Spirit, Sun Country, Independence Air, Virgin America, ATA, SkyBus, Midway, National, and Allegiant. Although Midwest and Frontier were purchased by Republic Holdings in 2009 and began operating as a single (low cost) carrier starting in 2010, during the period of our analysis the two carriers operated as separate carriers.
For example, since Delta already serves Atlanta from New York’s LaGuardia airport, when counting the number of legacy carriers serving Atlanta from core airport JFK, Delta is not counted.
While this procedure is straightforward when the metro area at the other endpoint (call it \(M)\) contains a single airport, further considerations arise when \(M\) also contains multiple airports, which may be grouped. Note first that the previous description of how in and out competition is measured must be amended when the metro area \(M\) at the other endpoint contains grouped airports. Now, all airports within the group at \(M\) are considered to be the same destination, so that the number of in competitors of a given type (legacy or LCC) equals the number of distinct carriers that provide service between metro area \(N\)’s primary airport and any of the grouped airports at \(M\). Likewise, the measurement of out competition is also based on the grouping at \(M\), with the number of out competitors equaling the number of distinct carriers providing service between any of \(N\)’s out airports and any of the airports in \(M\)’s group. From this discussion, it should be clear that the grouping decision for any given metro area may depend on whether airports in other metro areas are grouped. Our goal is thus to find a set of mutually consistent airport groupings across metro areas, as is discussed more fully below. To circumvent this issue entirely, we also re-estimated our regressions while dropping as potential endpoints all airports that are contained in any multi-airport metro area. Although this restriction eliminates a large portion of the sample routes, the results are generally consistent with those reported below.
Carrier fixed effects are present in the data at the carrier-route-quarter-year level, but when the data are aggregated to the route-quarter-year level (averaging to generate the mean fare), the fixed effects become passenger share variables.
In this and the subsequent discussion, a “route” should be viewed as ending at either a single airport or at an airport group within a multiple-airport metro area, as appropriate (see footnote 14).
Even though the regressions include carrier, route, and time fixed effects, the absence of route-by-time fixed effects means that unobserved route-specific time trends that affect both entry and fares could be present, potentially biasing the competition coefficients. Two-stage least squares regressions using lagged competition variables as instruments were estimated to address this problem, but the regressions contained an unacceptable number of anomalies such as insignificant in coefficients coupled with significant out coefficients (counter-intuitively indicating that in competition has no effect while out competition does). The only changes in our groupings using the lagged variables as instruments were an expansion of the San Francisco and Miami groupings to include SJC and PBI, respectively. In any case, using a sophisticated model with endogenous entry, Gayle and Wu (2011) show that little bias results from treating the number of competing carriers as exogenous, as we have done.
An alternate approach that is less likely to produce grouping would be to reject the grouping hypothesis when equality of competitive effects is rejected for at least one carrier type (rather than for both types), with grouping requiring failure to reject for both types. As a practical matter, however, because legacy and LCC service in multiple-airport metropolitan areas is frequently bifurcated (with legacy carriers providing most or all of their service from one airport and LCCs providing most or all of their service from the other), the “both” test can only be applied to a handful of metropolitan areas (Washington, New York City, San Francisco, and Miami/Ft. Lauderdale), making it impractical as a general decision criterion. A variant of this approach would be to use the compound null hypothesis that the in and out coefficients are equal for legacies and for LCCs, in which case grouping would be rejected on the basis of a single test rather than two separate equality tests. This test is too stringent, however.
For example, the Dallas/Fort-Worth metropolitan area, there are two large airports (DFW and DAL), but the only two routes with legacy carrier service from DAL (DAL-IAH and DAL-MEM) are also served by the same carriers from DFW. Therefore there are no additional legacy carriers providing service from DAL, and leg_comps_out is equal to zero for all routes. In order to perform the legacy (or LCC) test, we require there to be at least 5 routes with legacy (LCC) in service and 5 routes with non-overlapping legacy (LCC) out service, over the course of the sample period.
In Houston, legacy and LCC service was largely bifurcated between the two main airports (Houston Intercontinental and Houston Hobby) during the period of our sample, with the result that neither the LCC nor the legacy test can be performed. In this case, we base our grouping decision on the magnitude of the lcc_comps_out coefficient, as is seen below.
Another possibility is that the second-stage tests reject any grouping, in which case the first- and second-stage tests give conflicting results, making the overall outcome inconclusive.
When there is more than one fringe airport, the second-stage test described above can also be performed.
A crucial simplification relative to our previous paper is the suppression of nonlinear competition effects. Whereas that paper allowed the fare effect of adding a second, third, or fourth competitor to be different, the current approach (which relies on a single competition coefficient) constrains the effect of an additional competitor to be constant. Unfortunately, attempting to relax this simplification creates complications, as follows. Allowing a simple form for nonlinearity, we re-estimated the model with linear and quadratic terms for each competition variable. The quadratic terms were sometimes significantly different from zero, indicating nonlinear competition effects. However, testing for competition spillovers becomes much less straightforward than in our simple framework. First, a starting point for evaluating the fare impact of additional competition must be chosen (say, a zero competition level). Then, the quadratic expression would be evaluated at a competition level of one, an expression that involves a linear combination of the linear and quadratic coefficients. Finally, the difference in the value of this linear combination between the in and out cases must be computed and tested for statistical significance to see whether the in and out effects are different. Our simple linear specification avoids these complications while offering a reasonable approximation to the fare effects of additional competition. Also in contrast to our previous paper, we cannot separate the competitive effect of Southwest from that of other LCCs since doing so would create an unmanageable third set of competition coefficients. Nor can we allow competition effects to depend on whether an endpoint airport is a hub or whether a route has potential competitors. In both cases, doing so would require the introduction of interaction terms (which would, for example, interact the in and out competition measures with an endpoint hub dummy). Rather than being homogeneous, competition effects would then be differentiated across routes, which would preclude a simple test for competition spillovers.
The larger out coefficient is clearly counterintuitive, and other similar anomalies appear occasionally in the remaining results. Our response is that, since we estimate a large number of coefficients, and since these coefficients are random variables, unexpected relationships are bound to emerge occasionally.
Moreover, if IAD is chosen as the primary airport instead of DCA, our method continues to find that all three Washington area airports should be grouped.
Ishii et al. (2009) found, for example, that “[b]usiness travelers find OAK and SJC to be similar substitutes to SFO, but leisure travelers find OAK to be a much better substitute, controlling for included airport characteristics” (p. 221). Moreover, Pels et al. (2003) found that “it appears that SFO is a substitute for both OAK and SJC, while OAK and SJC are less a substitute” (p. 76).
The groupings that are shown in Table 4 are “mutually consistent.” To understand this property, note that the grouping decision for any particular metro area may depend on the pattern of groupings across the remaining multiple-airport metro areas, a pattern that affects the number of endpoints that are served from the given metro area’s primary airport (see footnote 14). Each grouping in Table 4 is selected conditional on the groupings that are shown for the other metro areas, making the collection of groupings mutually consistent. It should be noted that, in principle, a mutually consistent set of groupings need not exist.
Table 3 Regression results (continued) Table 4 Airport groupings Although the grouping of two airports that are nearly 80 miles apart may seem counterintuitive, several surveys of Cincinnati area passengers confirm this finding. For example, one recent survey of 17,000 passengers at CVG airport found that only 23 % of passengers used CVG for “all or most” of their personal or leisure trips. Likewise, 82 % of business passenger responded that their firms encouraged them to use surrounding airports because of price. See http://www.cvgsurvey.com/. Likewise, our findings are also consistent with one US DOT study found that “Cincinnati’s high fares have motivated certain travelers to seek out lower fares at other airports in the region. One of the purposes of Delta’s rollout of Simplifares in Cincinnati was to stem the diversion of revenue and traffic to other cities, such as Louisville and Dayton, where fares were generally lower than those available from Cincinnati.” See Fare Restructuring in Cincinnati—Second Quarter 2005, Domestic Aviation Competition Issue Brief Number 28, US Department of Transportation, Office of Aviation and International Affairs, Aviation Analysis.
See, for example, Impacts of Continental Airlines Operations on the New York-New Jersey Regional Economy, NERA Economic Consulting, November 2009 and Public Comments of the Department of Justice on the Show Cause Order, Joint Application of Air Canada, The Austrian Group, British Midland Airways Ltd, Continental Airlines, Inc., Deutsche Lufthansa Ag, Polskie Linie Lotniecze Lot S.A., Scandinavian Airlines System, Swiss International Air Lines Ltd., Tap Air Portugal, United Air Lines, Inc. to Amend Order 2007-2-16 under 49 U.S.C. §§ 41308 and 41309 so as to Approve and Confer Antitrust Immunity, Docket OST-2008-0234, (June 26, 2009). In June 2010, Continental offered 400 daily flights to 143 destinations from EWR versus 154 daily flights to 29 destinations by US Airways from LGA, 169 daily flights to 95 destinations from JFK by Delta, and 150 daily flights to 53 destinations from JFK by JetBlue. One reason for the failure of our method to group EWR could be the so-called “perimeter rule” that limits operations at LGA to flights no longer than 1,500 miles, thereby limiting competition between EWR and LGA to short- and medium-haul routes.
In addition, the groupings shown in Table 4 are mutually consistent only when EWR is grouped with the other New York airports (see footnote 27). If either EWR or JFK is chosen as the New York City’s primary airport (rather than LGA), our method finds that the primary airport should be grouped with LGA, although the competitive spillovers between EWR and JFK do not reach the equality threshold.
Additionally, where the null hypothesis that the coefficients are equal cannot be rejected, we require that at least one of the tested coefficients be negative and significant. While this outcome never arose in the within-metro-area tests, the rule avoids grouping airports based on equality of insignificant coefficients.
As discussed earlier, there is currently not enough in and out service to perform our tests in large, multiple-airport metropolitan areas such as Phoenix and Orlando; this is a deficiency that new service could remedy.
Moreover, it is important to acknowledge that in some instances, certain policy decisions could even alter the extent to which two airports in a metropolitan region are substitutable. For example, eliminating the perimeter rule at LGA would increase substitutability between the three New York City airports by permitting greater competition on long-haul services between EWR, JFK, and LGA.
See http://www.visitingdc.com/airport/bwi-train-station.htm and http://www.caltrain.com/schedules/weekdaytimetable.html). A similar lack of rail service between Los Angeles and any of its core and fringe airports may also partly explain why our findings do not support the grouping of those airports.
Notably, our results do not support grouping in Boston, where no legacy carrier operates a hub. Likewise, even though United operates a hub at LAX, LCCs have traditionally had a large presence at that airport.
Although it would be possible to restrict the analysis to routes of less than (or greater than) a certain distance, doing so substantially reduces the number of routes for each hypothesis test, and in many instances, makes the tests unworkable. Nevertheless, when we split the sample into routes that are longer or shorter than 1000 miles, the grouping results (where the tests can be performed) are similar to those in Table 4, with the most notable difference being that LAX and LGB can be grouped for the sample of routes longer than 1,000 miles. This outcome presumably reflects the fact that, for the bulk of the period of our analysis, JetBlue provided long-haul service from LGB to multiple East Coast destinations.
References
Armantier, O., & Richard, O. (2006). Evidence on pricing from the continental airlines and northwest airlines code-chare agreement. In D. Lee (Ed.), Advances in airline economics (Vol. 1, pp. 91–108). Amsterdam: Elsevier.
Armantier, O., & Richard, O. (2008). Domestic airlines alliances and consumer welfare. Rand Journal of Economics, 39, 875–904.
Bamberger, G., Carlton, D., & Neumann, L. (2004). An empirical investigation of the competitive effects of domestic airline alliances. Journal of Law and Economics, 42, 195–222.
Bennett, R. D., & Craun, J. M. (1993). The airline deregulation effect continues: The southwest effect. U.S. Department of Transportation, Office of Aviation Analysis, May 1993.
Berry, S. T. (1990). Airport presence as product differentiation. American Economic Review, 80, 394–399.
Berry, S., Carnall, M., & Spiller, P. (2006). Airline hubs: Costs, markups and the implications of customer heterogeneity. In D. Lee (Ed.), Advances in airline economics (Vol. 1, pp. 183–213). Amsterdam: Elsevier.
Boguslaksi, C., Ito, H., & Lee, D. (2004). Entry patterns in the southwest airlines route system. Review of Industrial Organization, 25, 317–350.
Borenstein, S. (1989). Hubs and high fares: Dominance and market power in the U.S. airline industry. RAND Journal of Economics, 20, 344–365.
Borenstein, S. (1991). The dominant-firm advantage in multiproduct industries: Evidence from U.S. airlines. Quarterly Journal of Economics, 106, 1237–1266.
Brueckner, J. K. (2003). International airfares in the age of alliances: The effects of codesharing and antitrust immunity. Review of Economics and Statistics, 85, 105–118.
Brueckner, J. K., Dyer, N., & Spiller, P. T. (1992). Fare determination in airline hub and spoke networks. RAND Journal of Economics, 23, 309–333.
Brueckner, J. K., & Spiller, P. T. (1994). Economies of traffic density in the deregulated airline industry. Journal of Law and Economics, 37, 379–415.
Brueckner, J. K., & Whalen, W. T. (2000). The price effects of international airline alliances. Journal of Law and Economics, 43, 503–545.
Brueckner, J. K., Lee, D., & Singer, E. (2013). Airline competition and domestic U.S. airfares: A comprehensive reappraisal. Economics of Transportation, 2.
Evans, W. N., & Kessides, I. (1993). Localized market power in the U.S. airline industry. Review of Economics and Statistics, 75, 66–75.
Evans, W. N., & Kessides, I. (1994). Living by the ‘golden rule’: Multimarket contact in the U.S. airline industry. Quarterly Journal of Economics, 109, 341–366.
Gayle, P. (2008). An empirical analysis of the competitive effects of the delta/continental/northwest code-share alliance. Journal of Law and Economics, 51, 743–766.
Gayle, P., & Wu, C.-Y. (2011a). A re-examination of incumbents’ response to the threat of entry: Evidence from the airline industry. Kansas State University. : Unpublished paper.
Gayle, P., & Wu, C.-Y. (2011b). Are air travel markets segmented along the lines of nonstop versus intermediate-stop(s) products?. Kansas State University. : Unpublished paper.
Goolsbee, A., & Syverson, C. (2008). How do incumbents respond to the threat of entry? Evidence from major airlines. Quarterly Journal of Economics, 123, 1611–1633.
Harvey, G. (1986). Study of airport access mode choice. Journal of Transportation Engineering, 112, 525–535.
Harvey, G. (1987). Airport choice in a multiple airport region. Transportation Research Part A, 21, 439–449.
Hess, S., & Polak, J. W. (2006). Exploring the potential for cross-nesting structures in airport-choice analysis: A case-study of the Greater London area. Transportation Research Part E: Logistics and Transportation Review, 42, 63–81.
Ishii, J., Jun, S., & Van Dender, K. (2009). Air travel choices in multi-airport markets. Journal of Urban Economics, 65, 216–227.
Ito, H., & Lee, D. (2007). Domestic codesharing, alliances, and airfares in the U.S. airline industry. Journal of Law and Economics, 50, 355–380.
Marcucci, E., & Gatta, V. (2011). Regional airport choice: Consumer behaviour and policy implications. Journal of Transport Geography, 19, 70–84.
Morrison, S. A. (2001). Actual, adjacent, and potential competition: Estimating the full effect of southwest airlines. Journal of Transport Economics and Policy, 32, 239–256.
Pels, E., Nijkamp, P., & Rietveld, P. (2000). Airport and airline competition for passengers departing from a large metropolitan area. Journal of Urban Economics, 48, 29–45.
Pels, E., Nijkamp, P., & Rietveld, P. (2001). Airport choice in a multiple airport region: A case study for the San Francisco Bay Area. Regional Studies, 35, 1–9.
Pels, E., Nijkamp, P., & Rietveld, P. (2003). Access to and competition between airports: A case study for the San Francisco Bay Area. Transportation Research Part A, 37, 71–83.
Peters, C. (2006). Evaluating the performance of merger simulation: Evidence from the U.S. airline industry. Journal of Law and Economics, 49, 627–649.
Werden, G., Joskow, A., & Johnson, R. (1991). The effects of mergers on price and output: Two case studies from the airline industry. Managerial and Decision Economics, 12, 341–352.
Whalen, W. T. (2007). A panel data analysis of code-sharing, antitrust immunity, and open skies treaties in the international aviation market. Review of Industrial Organization, 30, 39–61.
Acknowledgments
The authors thank Lawrence White and two anonymous referees for helpful suggestions. The research reported in this paper was carried out with financial support from United Airlines. However, the views expressed in the paper are ours alone. The authors have also served as consultants to several other airlines.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Brueckner, J.K., Lee, D. & Singer, E. City-Pairs Versus Airport-Pairs: A Market-Definition Methodology for the Airline Industry. Rev Ind Organ 44, 1–25 (2014). https://doi.org/10.1007/s11151-012-9371-7
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11151-012-9371-7