TOM KING Executive Vice President - External Relations British Airways Plc Rue Wiertz 50 B-1050 Brussels, Belgium

Transcription

1 BEFORE THE DEPARTMENT OF TRANSPORTATION WASHINGTON, D.C Joint Application of: : : AMERICAN AIRLINES, INC. : BRITISH AIRWAYS PLC : FINNAIR OYJ : IBERIA LÍNEAS AÉREAS DE ESPAÑA, S.A. : ROYAL JORDANIAN AIRLINES : : under 49 U.S.C and for : approval of and antitrust immunity for : alliance agreements : DOT-OST [PUBLIC] JOINT APPLICANTS MOTION FOR LEAVE TO FILE AND SUPPLEMENTAL COMMENTS Communications with respect to this document should be sent to: For American Airlines: For British Airways: For Iberia: WILLIAM K. RIS, JR. Senior Vice President Government Affairs American Airlines, Inc th Street, Suite 600 Washington, DC TOM KING Executive Vice President - External Relations British Airways Plc Rue Wiertz 50 B-1050 Brussels, Belgium VIRASB VAHIDI Senior Vice President Planning American Airlines, Inc. P.O. Box , MD 5628 DFW Airport, TX KENJI C. HASHIMOTO Vice President Strategic Alliances American Airlines, Inc. P.O. Box , MD 5635 DFW Airport, TX CARL B. NELSON, JR. Associate General Counsel American Airlines, Inc th Street, Suite 600 Washington, DC R. BRUCE WARK Associate General Counsel JEFFREY A. OGAR Senior Attorney American Airlines, Inc. P.O. Box , MD 5675 DFW Airport, TX (817) jeff.ogar@aa.com PAUL C. JASINSKI General Counsel, Americas JAMES B. BLANEY Senior Counsel, Americas British Airways Plc Astoria Boulevard Jackson Heights, NY (347) paul.jasinski@ba.com DARYL A. LIBOW Sullivan & Cromwell LLP 1701 Pennsylvania Ave., N.W. Washington, DC (202) libowd@sullcrom.com DON H. HAINBACH Garofalo Goerlich Hainbach PC 1200 New Hampshire Ave., N.W. Washington, DC (202) dhainbach@ggh-airlaw.com ELVIRA HERRERO Senior Vice President - Int l Relations & Alliances Iberia Líneas Aéreas de España, S.A. Velazquez, Madrid, Spain WILLIAM KARAS CAROL GOSAIN Steptoe & Johnson LLP 1330 Connecticut Ave., N.W. Washington, DC (202) wkaras@steptoe.com For Finnair: SAMI SARELIUS Vice President & General Counsel Finnair Oyj Tietotie 11 A FI-01053, Finland +358 (9) sami.sarelius@finnair.fi For Royal Jordanian: MICHAEL J. HOLLAND Condon & Forsyth LLP 7 Times Square New York, NY (212) mholland@condonlaw.com September 8, 2009

2 TABLE OF CONTENTS INTRODUCTION... 2 I. DOJ S TWO FARE STUDIES ARE METHODOLOGICALLY FLAWED AND REACH INACCURATE AND UNRELIABLE RESULTS... 4 A. Immunized Alliances Will Not Raise Fares On Nonstop Overlaps Immunized Alliances Do Not Cut Hub-To-Hub Capacity DOJ s Fare Study Of Nonstop Routes Asked The Wrong Question... 7 B. Immunized Alliances Deliver Lower Fares On Connecting Itineraries Assuming There Is Inter-Alliance Competition... 9 II. CARVE-OUTS WOULD DESTROY THE JOINT BUSINESS AGREEMENT S METAL NEUTRALITY AND REDUCE CONSUMER BENEFITS A. No Basis Exists For Imposing Carve-Out Remedies In This Proceeding DOJ s Flawed Fare Study Does Not Justify Carve-Outs There Is No Basis For The Department To Abandon Its Established Precedents In SkyTeam II And Star II B. Carve-Outs Would Result In Inefficient Yield Management C. Carve-Outs Would Create Other Competitive Harm And Reduce Consumer Benefits D. If Imposed, Carve-Outs Should Exclude Joint Venture Traffic III. ABSENT ANTITRUST IMMUNITY, THERE WOULD BE NO PROSPECT OF ONEWORLD DELIVERING AN IMPROVED TRANSATLANTIC NETWORK PRODUCT TO CONSUMERS A. Inter-Alliance Competition B. Domestic Competition C. Scope of Immunity CONCLUSION... 24

3 BEFORE THE DEPARTMENT OF TRANSPORTATION WASHINGTON, D.C Joint Application of: : : AMERICAN AIRLINES, INC. : BRITISH AIRWAYS PLC : FINNAIR OYJ : IBERIA LÍNEAS AÉREAS DE ESPAÑA, S.A. : ROYAL JORDANIAN AIRLINES : : under 49 U.S.C and for : approval of and antitrust immunity for : alliance agreements : DOT-OST [PUBLIC] JOINT APPLICANTS MOTION FOR LEAVE TO FILE AND SUPPLEMENTAL COMMENTS The Joint Applicants hereby move for leave to file the following supplemental comments for the Department s consideration. These comments are narrowly tailored to address specific issues raised by the U.S. Department of Justice in its June 26, 2009 comments ( DOJ Comments ) in the Star II docket (DOT-OST ). The Joint Applicants hope that addressing these issues proactively will enable the Department to expedite its issuance of a Show Cause Order and approval of global antitrust immunity for the Joint Applicants. Indeed, with Continental s announcement that it will formally join Star on October 27, each day that passes makes it even more critical that the proposed oneworld alliance be approved and immunized within the statutory October 31 deadline. Recognizing that other interested parties, such as DOJ, may want to respond to these comments, the Joint Applicants propose a period for responses limited to the specific issues that our supplemental comments address. This comment period should be consistent with the Department s precedent in Star II to avoid unnecessary delay in this proceeding (Notice Establishing Supplemental Comment Period, DOT-OST , 6/26/09).

4 - 2 INTRODUCTION In Star II, DOJ suggested that there is no need to approve further transatlantic alliances because the United States has already reaped the full benefits of the U.S.-EU Open Skies Agreement (DOJ Comments, pp ). Yet it would be naïve to assume that the United Kingdom and Spain two of the countries most directly impacted by liberalization (and whose support was critical to the adoption of the agreement) - would find value in a regulatory policy that denies their consumers the benefits of fully integrated alliance hub operations. In arguing for carve-outs from the Department s grant of immunity, DOJ asserted in Star II that oneworld is an effective competitor in the current environment because its two largest members American and British Airways today function effectively without immunity between each other (Id. at 34). Such a statement ignores reality, as oneworld has fallen further and further behind the Star/SkyTeam alliances over the past decade. It would be poor public policy to deny consumers the benefits of network competition through London and Madrid, while at the same time allowing Star and SkyTeam to expand their positions in Paris, Frankfurt, Amsterdam and other airports across Europe. PUBLIC VERSION

5 - 3 While DOJ claims that achieving balance in the market success of differing alliances is not a legitimate goal of sound competition policy (Id.), the Department s mandate is far broader, and specifically includes strengthening the competitive position of air carriers to ensure at least equality with foreign air carriers, including the attainment of the opportunity for air carriers to maintain and increase their profitability in foreign air transportation (49 U.S.C (e)). In any case, it is not obvious how withholding antitrust immunity from one alliance, in effect leaving a duopoly for the remaining alliances, advances competition policy. To be clear, oneworld will not remain viable as an alliance if its transatlantic members are denied the equal opportunity to compete with Star and SkyTeam on a network basis. The following comments provide further evidence that: (a) DOJ s methodology for analyzing nonstop and connecting transatlantic fares is fundamentally flawed and asked the wrong questions; (b) carve-outs would destroy the metal neutrality of the Joint Business Agreement and reduce significantly the consumer benefits the JBA would otherwise create; and (c) antitrust immunity is a necessary pre-condition to deeper and expanded cooperation among oneworld s transatlantic airlines and to the consumer benefits that will arise out of enhanced inter-alliance competition. PUBLIC VERSION

6 - 4 I. DOJ S TWO FARE STUDIES ARE METHODOLOGICALLY FLAWED AND REACH INACCURATE AND UNRELIABLE RESULTS A. Immunized Alliances Will Not Raise Fares On Nonstop Overlaps 1. Immunized Alliances Do Not Cut Hub-To-Hub Capacity DOJ claims that its evidence shows that a reduction in the number of competing airlines offering nonstop transatlantic flights may result in large, statistically significant price increases (DOJ Comments, p. 47). But capacity not simply the number of competitors is the primary determinant of price in the airline industry. Network carriers make route and capacity decisions based upon network considerations, which dictate that they increase capacity on hub-to-hub routes to better compete for flow traffic (Casey Decl., Exhibit 2, 5). Higher fares would only result if a reduction in the number of competitors is accompanied by a reduction in capacity, a variable for which DOJ did not control. History has shown that immunized alliances increase capacity on hubto-hub routes, even if there is no other competition on the route. As American demonstrated in Star II, there is a direct correlation between the amount of integration in an alliance and the ratio of frequencies/seats to the size of the local market: PUBLIC VERSION

7 - 5 Ratio Of Nonstop Flights/Seats To O&D Bookings, Hub-To-Hub U.S.-EU Routes Source: OAG, MIDT (Year Ended May 2009) City-Pair BPDEW 1 Seats Ratio Flights Ratio NW/KL Joint Venture AMS-DTT 33 1,192 36:1 4 1:8 AMS-MEM :1 1 1:8 AMS-MSP :1 3 1:11 AF/DL ATI Alliance ATL-PAR :1 4 1:14 CVG-PAR :1 1 1:20 AA/BA/IB Non-Immunized Alliance DFW-LON :1 4 1:27 MAD-MIA :1 2 1:48 Within the Northwest/KLM joint venture, three extremely thin local hub-to-hub routes (with a combined 74 local O&D bookings per day each way) support eight daily flights. In fact, Northwest (now Delta) and KLM operate twice as many daily flights and seats between Amsterdam and Detroit as American and Iberia operate between Miami and Madrid despite the fact that the DTT-AMS local market is one-third the size of MIA-MAD. This is because the highly integrated Northwest/KLM alliance requires capacity on this route to accommodate flow traffic, leading to increased capacity despite the fact that the carriers face no nonstop competition on the route. 1 Bookings Per Day Each Way. PUBLIC VERSION

8 - 6 Absent a reduction in capacity, the Joint Applicants would have no ability or incentive to raise local fares. [REDACTED] (Joint Applicants Response to Order , p. 27). To the extent DOJ s fare study found any significant difference in fares, these likely resulted from differences in capacity rather than the number of competitors. In addition, airlines have an extremely limited ability to fine tune seat capacity, as an increase of just a single daily wide-body frequency on a route can add nearly 200,000 seats on the route each year. For example, American and British Airways currently operate 3-4 daily frequencies between Dallas/Fort Worth and Heathrow, with an average total of 782 daily seats. The addition of even a single daily frequency operated by American s smallest wide-body jet (Boeing ) would add 219 seats or 28% of capacity. The idea that an alliance such as oneworld, having been granted antitrust immunity, would exercise market power on local customers which only comprise a fraction of total onboard traffic is unrealistic in practice and not supported by the empirical data. PUBLIC VERSION

9 DOJ s Fare Study Of Nonstop Routes Asked The Wrong Question DOJ s study in Star II was not a study of the effect of alliances; rather, it was a study of the effect of any change in the number of distinct carriers serving a route. DOJ simply assumes without justification that granting antitrust immunity on a route has the same effect on fares as the exit of a competitor s capacity from the market. However, there are fundamental economic distinctions between these situations, and the effect of antitrust immunity is quite different from the effect of a carrier exiting the market. In fact, the expected outcome of the JBA is to create a strong incentive to expand capacity on hub-to-hub routes, and the added capacity will naturally put downward pressure on fares. With this distinction in mind, Compass/Lexecon conducted a fare study that isolates the effect of antitrust immunity on fares, and found no effect (Exhibit 1). Before equating antitrust immunity to exit, DOJ s Star II study of nonstop fares had significant methodological shortcomings. DOJ only used one quarter of data 3Q 2008 even though many years of data were available (Exhibit 1, 18). Thus, while DOJ compared Atlanta-Rome (which had one carrier in that quarter) with Atlanta-Frankfurt (which had two carriers), it is difficult to fully control for all differences between the two routes (Id.). For example, the cities have different levels of demand and different distances. Compass using a panel dataset from Q through Q was able to control for these differences by using route fixed effects (Id.). This makes the Compass methodology far stronger than that used by DOJ. PUBLIC VERSION

10 - 8 Compass also performed the analysis using additional standard controls to ensure that its findings were not simply the product of insufficient changes in the number of competitors on particular routes (particularly when measuring the impact of antitrust immunity) (Id. at 21). In this analysis, Compass, like DOJ, controlled for route distances, but improved upon the simple linear measure used by DOJ (Id.). Airlines recognize that the relationship between distance and cost is non-linear, and Compass used a five-part spline to address this (Id.). Compass also added controls that DOJ simply excluded, such as a control for the percentage of traffic on each route served by different carriers in order to control for the fact that different carriers serving different routes may lead to differences in fares (Id.). By separating the effect of antitrust immunity from the effect of actual reductions in the number of carriers serving a route, Compass tested its hypothesis that the effect of antitrust immunity differs from the effect of a competitor exiting a route. The answer is clear across a range of specifications. The measured effect of antitrust immunity on nonstop fares is almost exactly zero and is statistically insignificant (Id. at 20, 22). This demonstrates that (a) antitrust immunity delivers consumer benefits without a cost on the nonstop overlaps; and (b) there is no justification for any remedies to be imposed on any of the nonstop overlap routes in this proceeding. PUBLIC VERSION

11 - 9 Additionally, DOJ s findings with regard to price effects from reducing the number of competitors on a route were overstated. Applying route-fixed effects, Compass demonstrated that the price effects on 2-to-1 routes were cut in half and rendered statistically insignificant (Id. at 20). On 3-to-2 routes, DOJ s original findings were themselves statistically insignificant (and the only 2-to-1 route in this proceeding is Dallas/Fort Worth-London), 2 and remained so after Compass applied routefixed effects (Id.). Using additional data and controls, the effect on 2- to-1 routes was just 3.6%, and on 3-to-2 routes it was just 1.1% (both results statistically insignificant) (Id. at 22). In sum, Compass has demonstrated that DOJ s assumptions were incorrect, its study of nonstop routes was flawed, and there is no evidence that antitrust immunity leads to increased fares on hub-to-hub overlap routes. B. Immunized Alliances Deliver Lower Fares On Connecting Itineraries Assuming There Is Inter-Alliance Competition DOJ also claimed that its study in the Star II docket showed that immunized alliance fares are 2.1% higher than online fares and 3.6% higher than non-immunized alliance fares (DOJ Comments, p. 51). This finding directly contradicts the peer-reviewed Whalen (2007) paper a study performed by one of DOJ s well-published economists in the airline industry. Moreover, DOJ s study failed to generate a statistically significant finding with regard to non-immunized alliance fares (i.e., simple codesharing). It is hard to see how a comparison of immunized alliance fares to a statistically insignificant finding about codeshare fares can itself become significant particularly when the conclusion contradicts a decade of academic literature on the subject. 2 Miami-Madrid is now a 3-to-2 route, as Air Europa (an associate member of SkyTeam) has announced that it will be offering five weekly flights on the route effective March 19, 2010 (Air Europa Press Release, 7/28/09). PUBLIC VERSION

12 - 10 Compass was asked to review this study as well, and they confirmed that DOJ s study fails to include a number of controls that are typically used in the academic literature (See, e.g., Whalen (2007)). Compass added the following standard controls that DOJ appears to have omitted from its work without explanation (Id. at 28): Income and population of the U.S. and EU cities making up the endpoints of the O&D city-pair; Carrier-fixed effects to account for the difference in fares charged by different marketing carriers; Year-fixed effects to account for average changes in fares over time; Market concentration for online/alliance fares and separately for interline fares; and An indicator for routes between countries with U.S. open skies agreements. By adding these standard controls, Compass found that ATI alliance fares were on average 6.0% lower than interline fares, and 3.4% lower than non-immunized codeshare fares and all results were statistically significant (Id.). While the absolute numbers differ somewhat from Whalen (2007), the findings are consistent antitrust immunity has historically resulted in lower fares. One potential explanation for the difference between the Whalen and the more recent Compass numbers is the significant reduction in inter-alliance competition that has taken place over the past decade as oneworld s share of traffic has declined. 3 3 This is consistent with The Brattle Group s findings in the SkyTeam I docket (DOT-OST ), in which it concluded that the Department s earlier findings of consumer benefits from immunized alliances had been reversed by the lack of three competing alliance networks. PUBLIC VERSION

13 - 11 II. CARVE-OUTS WOULD DESTROY THE JOINT BUSINESS AGREEMENT S METAL NEUTRALITY AND REDUCE CONSUMER BENEFITS A. No Basis Exists For Imposing Carve-Out Remedies In This Proceeding 1. DOJ s Flawed Fare Study Does Not Justify Carve-Outs In the past, the Department has imposed carve-outs on antitrust immunity for nonstop overlap routes where there was limited prospect for new entry given the characteristics of the route. However, these carve-outs were typically limited in nature and, most importantly, applied to alliances that had not truly integrated their networks into a metal-neutral joint venture. DOJ which has never provided any evidence that carve-outs preserve local O&D competition - argued for their imposition in Star II based on an analysis of coach fares on international routes indicat[ing] that the narrow carve-out language is insufficient to protect competition... because decreasing the number of nonstop competitors increases all nonstop coach prices (DOJ Comments, p. 40). Given the serious methodological flaws in DOJ s study and the facts set forth above, there is no basis to suggest that carve-out remedies should be considered in this proceeding. PUBLIC VERSION

14 - 12 Moreover, the Department now has two well reasoned precedents of removing carve-outs from integrated metal-neutral joint ventures. The Department discussed the issue of carve-outs in the context of a true joint venture in SkyTeam II, holding that carve-outs could undermine the effectiveness of the 4-way JV and jeopardize the public benefits (Order , p. 10). In Star II, the Department followed this precedent, holding that the broad public benefits of metal-neutral cooperation on these key, trunk routes would go unrealized without immunity, [and] carve outs inhibit the realization of efficiencies and thereby the consumer benefits resulting from those efficiencies (Order , p. 20). The Department did not carve out cooperation between participants in the Star joint venture on overlapping transatlantic routes in its final grant of immunity (Id.). Choosing to do so in this proceeding would be fundamentally unfair to the Joint Applicants. 2. There Is No Basis For The Department To Abandon Its Established Precedents In SkyTeam II And Star II American, British Airways and Iberia following Department precedent relied upon the established legal regime in structuring their JBA (Consolidated Edison Co. v. FERC, 315 F.3d 316, 323 (D.C. Cir. 2003); see also DOT Regulatory Information, The Department should not abandon its recent precedents in adjudicating this proceeding the third major alliance case presented in just the last two years (Id.). Moreover, the European Commission has consistently rejected carve-out remedies in airline alliance and merger cases; if the Department were to impose such remedies here it would also be creating inconsistent outcomes, contrary to the principles of convergence called for under the U.S.-EU Open Skies Agreement. PUBLIC VERSION

15 - 13 B. Carve-Outs Would Result In Inefficient Yield Management As American demonstrated in Star II, carve-outs destroy metal neutrality. While DOJ claims that [i]n analyzing airline matters, the relevant markets are no larger than city pairs (DOJ Comments, p. 17), this ignores the network nature of commercial aviation and the supply-side reality that airlines measure profitability by flight segment, not O&D city pair (Casey Decl., Exhibit 2, 9). Each flight segment can be sold as a part of hundreds or thousands of different city-pair itineraries (Id.). A carve-out in a metal-neutral joint venture would inherently disrupt the airlines ability to efficiently allocate seats on the transatlantic segments to the consumers who value them most. This is precisely why the Department has not imposed carve-outs on metal-neutral joint ventures. Nonetheless, DOJ claims that carve-outs would not lead to diminished efficiencies or consumer value because Star members have long operated under carve-outs imposed as a part of prior immunity grants (DOJ Comments, p. 39). Yet DOJ provides no evidence to support its assumption that just because Star operated with carve-outs, they were not disruptive or inefficient. 4 Moreover, Star was not a fully integrated alliance during this period, and DOJ fails to recognize the quantum leap that alliances have made from 2005 (when the Department rejected SkyTeam s application for expanded antitrust immunity based on the lack of a fully integrated joint venture) to today. As the Department held in SkyTeam I, the consumer benefits that are perhaps directly attributable to antitrust immunity appear to be dependent on the successful implementation of an economic benefit sharing agreement among the alliance partners (Order , p. 37). 4 DOJ also apparently and incorrectly believes that oneworld is a fully effective competitor to Star and SkyTeam even in the absence of antitrust immunity (DOJ Comments, p. 34). Evidence of mere survival should not be mistaken for evidence of success. PUBLIC VERSION

16 - 14 Carve-outs are antithetical to a joint venture dependent on metal neutrality and revenue sharing. In a metal-neutral joint venture, carriers share the revenues from all passengers on a transatlantic segment both local and flow. When yield managing the Dallas/Fort Worth-London route, for example, American and British Airways might value three O&D passengers as follows: 5 O&D Fare Operating Carrier Revenue Rank AUS-LHR $1,000 AA: $400/BA: $500 #1/#1 DFW-CDG $900 AA: $360/BA: $450 #2/#2 DFW-LHR $800 AA: $320/BA: $400 #3/#3 In the example above, the three carriers would have the same economically efficient yield management incentives. However, a local traffic carve-out would radically transform the equation: O&D Fare Operating Carrier Revenue Rank AUS-LHR $1,000 AA: $400/BA: $500 #2/#2 DFW-CDG $900 AA: $360/BA: $450 #3/#3 DFW-LHR $800 AA: $800/BA: $800 #1/#1 In this second example, both American and British Airways individually stand to gain more revenue from a local customer than their share of a highervalue flow customer. This is because the airlines get all the revenue from a carved-out passenger, but only their proportional JBA share of the flow traffic. The result is to create a perverse incentive for each party to carry as much local traffic as possible, funneling connecting traffic to the partner (while still collecting the revenue share). Incentivizing each carrier in such a manner would destroy the efficiencies inherent in the proposed joint venture by eliminating metal neutrality and discouraging the carriage of flow traffic. The JBA could not operate in such an environment. 5 For simplicity s sake, the tables assume no proration of behind/beyond revenue, and a simple 50%-40%-10% revenue share among the parties. While this does not reflect the actual working of the JBA, the conclusions remain valid. PUBLIC VERSION

17 - 15 Additionally, since carve-outs in a metal-neutral joint venture would create an artificial incentive to carry local traffic before flow traffic, the relative flow yield for JBA traffic would have to increase if it were to supplant local traffic (Id. at 22-23). This would leave oneworld completely uncompetitive with Star and SkyTeam on these flow O&D city-pairs, reducing the consumer benefits generated by the proposed alliance (Id.). C. Carve-Outs Would Create Other Competitive Harm And Reduce Consumer Benefits Carve-outs would have a negative effect on other elements of the metal-neutral joint venture. When making decisions about scheduling or capacity changes, airlines assess the combined contribution of local and flow passengers. An efficient joint venture must be able to plan in the same manner. Thus, the expected network contribution from a hub-to-hub transatlantic flight will depend on measuring demand from local and flow passengers across the network. [REDACTED] Limiting the ability to fully integrate the networks would make it less likely that the parties would add capacity, thus restricting oneworld s ability to maximize consumer value (Id. at 28). PUBLIC VERSION

18 - 16 Carve-outs would also limit the alliance s ability to re-time flights to optimize connectivity. For example, before Swiss was acquired by Lufthansa, it had a joint venture on JFK-Zurich with American (Id. at 19). The partners decided it was in the alliance s combined interest to change American s JFK departure time from the 1700 hour to the 2100 hour (resulting in a 1030 arrival instead of 0700), even though was the ideal time slot for local passengers (Id. at 21). The change was made because the 2100 departure would increase connecting options for flow traffic over Zurich (Id.). Moreover, the shift offered local passengers more choices (1800 and 2100) (Id.). However, if the local market had been carved out, neither carrier would have had an incentive to move away from the optimal local departure time a classic example of Hotelling s Law (See Harold Hotelling, Stability in Competition, Economic Journal, vol. 39, no. 153 (March 1929), 41-57) (describing the fact that while it would be more socially beneficial if two shops on a street separated themselves and moved to one quarter of the way along the street from each end, neither shop would be willing to do this independently). PUBLIC VERSION

19 - 17 Prior to American s re-timing, JFK-ZRH had 83,825 annual O&D bookings. After the shift, that jumped 13.5% to 95,144 bookings (MIDT, Years Ending October 2004/2005). As a result of the shift, American lost 11.5 points of O&D booking share, dropping from 40.2% to 28.7%. However, Swiss recaptured some of that share, as did other carriers. More importantly, the alliance was carrying more flow traffic, leading to higher load factors and RASM, while yield remained relatively constant. Thus this joint venture both created consumer benefits and maximized network revenue. This example shows that collaboration on a nonstop overlap can stimulate local demand by spreading departures across the day and improve connectivity for flow traffic. The Joint Applicants initial analysis (which was extremely conservative, as the parties have not yet been able to create a fully optimized joint schedule) shows that even minor schedule changes would generate over 350,000 hours of annual time savings for local customers as a result of spreading out departures more efficiently across the day, as well as nearly 200,000 hours per year in time saved by connecting passengers. Using the Department s standard valuation of passengers time, this comes to nearly $20 million in consumer benefits (See GRA, Inc., Economic Values For FAA Investment And Regulatory Decisions, 10/3/07, p. 16 and Bureau of Labor Statistics, Employment Cost Index, from which a present value of $35.70 per hour is derived). Carve-outs would destroy that significant consumer benefit. PUBLIC VERSION

20 - 18 Carve-outs would also limit the parties ability to close the transatlantic frequent flyer gap that prevents American s and British Airways customers from enjoying the same benefits offered by Star and SkyTeam. [REDACTED] In fact, American receives approximately [REDACTED] requests per year for mileage credit on ineligible British Airways flights due to the transatlantic gap. While it is impossible to accurately quantify this consumer benefit, it is obvious that consumers demand frequent flyer reciprocity and that the only way that the American/British Airways transatlantic gap will be closed is by approving the JBA without carve-outs. At the same time, Compass measured the effects of carve-outs on fares and service quality on nonstop overlap routes. Holding fares fixed, Compass found that consumers prefer full antitrust immunized service to that including a carved-out segment. The value of this benefit to consumers is more than $60, and at the same time fares on connecting itineraries with a carved-out segment are 2.7% higher than fares on those which are fully integrated (Exhibit 1, 35). In other words, carve-outs cause consumers to get fewer benefits at a higher fare a total harm of at least $55 million per year if imposed on the proposed alliance (Id.). It is clear that carveouts do more harm than good, and should be abandoned as a remedy in antitrust immunity cases involving a metal-neutral joint venture. PUBLIC VERSION

21 - 19 D. If Imposed, Carve-Outs Should Exclude Joint Venture Traffic The final Star II order appropriately determined that carve-outs should not apply to the parties A++ joint venture (Order , Appendix A, p. 1). To the extent that the Department feels compelled to impose carve-outs on any route, it should follow that precedent here. III. ABSENT ANTITRUST IMMUNITY, THERE WOULD BE NO PROSPECT OF ONEWORLD DELIVERING AN IMPROVED TRANSATLANTIC NETWORK PRODUCT TO CONSUMERS A. Inter-Alliance Competition DOJ claimed in Star II that alliance efficiencies can be created by no more than... a special prorate agreement (DOJ Comments, p. 35). Yet the Joint Applicants have conclusively demonstrated by their market behavior over the past decade that this is incorrect. While American and British Airways have had a limited alliance since 1997 (including a very liberal special prorate agreement), they have hardly created the type of integration enjoyed by the Star and SkyTeam alliances customers: Codesharing: Unlike most other alliances, American and British Airways do not codeshare on transatlantic routes. Nor do they codeshare on many behind/beyond routes. Absent metal neutrality, each carrier s individual interest in carrying traffic on transatlantic segments has limited its willingness to codeshare on the other. Nor do the two carriers codeshare on major spoke cities, such as Milwaukee, New Orleans, Geneva and Venice. Thus while DOJ claims that codesharing and special prorate agreements alone can lead to lower fares (DOJ Comments, p. 36 n.98), the lack of antitrust immunity has prevented American and British Airways from fully integrating their networks (Casey Decl., Exhibit 2, 15). Frequent Flyer Benefits: American and British Airways do not have fully reciprocal frequent flyer programs. American s AAdvantage members do not earn miles on British Airways transatlantic flights, and vice versa for BA s Executive Club members. American (which has more FFP members) is unwilling to open its program to British Airways for fear that it will lose revenue, while British Airways fears that its customers would migrate to AAdvantage if they could earn AAdvantage miles on British Airways transatlantic flights (Id. at 14). PUBLIC VERSION

22 - 20 Fare Combinability: While other alliances allow fare combinability which creates more inventory for consumers at lower fares American and British Airways do not. Again, the two carriers see no value in fare combinability if the result is simply diversion of revenue from each other. As a result, Star and SkyTeam are able to continue taking traffic away from both carriers (Id. at 16). Joint Corporate Dealing: DOJ claimed in Star II that [f]ew, if any, corporate travel managers we interviewed during our investigation of this application and in the course of other airline investigations have stated a desire for increased interalliance competition (DOJ Comments, p. 34). This claim is directly contradicted by the evidence in this proceeding, as the Joint Application is supported by over 500 corporations and small businesses as well as the Business Travel Coalition and the American Small Business Travel Association. BTC stated that in the absence of comprehensive network competition, consumer benefits from more efficient alliances cannot fully materialize. An immunized oneworld would level the playing field and considerably strengthen network competition at a critical time for consumers as the global airline industry is consolidating rapidly. oneworld cannot be a player in joint corporate dealing absent antitrust immunity (Casey Decl., Exhibit 2, 16). Despite DOJ s unsubstantiated assertion that a special prorate agreement is all the Joint Applicants need, 6 none of the foregoing problems has been resolved in the more than a decade American and British Airways have been in a marketing alliance. In fact, American and British Airways have not expanded their behind/beyond codesharing by a single route in a number of years (Id. at 15). The explanation is simple even though the result would be a more compelling product offering, the absence of antitrust immunity has kept the airlines from resolving the diversion problem and functioning effectively as an integrated competitor to Star and SkyTeam. Each carrier s economic incentive to steer flow traffic toward its own flights creates strong disincentives to fully integrated and efficient cooperation that are far too strong to be remedied by a special prorate agreement. 6 DOJ s comments provided no specific suggestions on how this result could be achieved, casting doubt on the factual basis upon which DOJ makes its claim given the failure of alliances to integrate absent ATI. PUBLIC VERSION

23 - 21 Moreover, a special prorate agreement does nothing to create non-price related efficiencies, such as more convenient scheduling, increased flight frequency, new service and other types of network expansion. Without the Joint Applicants sharing revenues through a metal-neutral joint venture, there would be little economic incentive to coordinate on these areas, as there would be little incentive to make joint investments. Moreover, concerns over frivolous litigation being filed by competitors (such as Virgin Atlantic) would make any implementation unlikely. As a result, AA/BA interline fares are 10%-15% higher than American s online fares for U.S. point-of-sale customers (O&D Survey, 3Q05-3Q08), and the product quality is lower. This explains why Star and SkyTeam achieve far more share of U.S.-EU traffic on connecting O&D city-pairs. In Germany, for example, Star carries the most traffic on O&Ds where connecting service is necessary ( flow O&Ds ) 49%. This is to be expected, since Star has hubs in Frankfurt and Munich. SkyTeam claims 39% of the traffic, despite the fact that it does not have a hub in Germany. oneworld is a distant third with just 10%. This disparity exists because SkyTeam has antitrust immunity to coordinate flow traffic over Paris and Amsterdam, while oneworld lacks the ability to compete on equal terms over Heathrow and Madrid. The same is true in France SkyTeam carries 56% of flow O&D traffic by virtue of its Paris hub and Star carries 26% (despite having to back-haul from Germany), compared to just 17% for oneworld (Exhibit 4, p. 11). PUBLIC VERSION

24 - 22 Incredibly, even in the UK where oneworld has a hub at Heathrow oneworld carries just 24% of the traffic on flow O&Ds, compared to 46% for SkyTeam and 26% for Star. In Spain, oneworld s 33% share falls well behind SkyTeam s 41%. Thus even in its own home countries oneworld is unable to compete on equal terms for flow traffic (Id.). This, among other reasons, is why noted industry experts (including Randy Bennett, who has assisted the Joint Applicants in this proceeding) have concluded that the Department s antitrust authority provides a means whereby airlines can overcome legal limitations in order to implement international alliances that demonstrably benefit consumers and enhance efficiency (Exhibit 3, p. 7). B. Domestic Competition The Department also needs to consider the role that international alliances play in preserving domestic competition. Delta now the world s largest airline has a fully integrated joint venture with Air France/KLM (currently Europe s largest airline). United and Continental the third and fourth largest U.S. network carriers will have a fully integrated joint venture with Lufthansa (soon to be Europe s largest airline) and Air Canada. This gives these carriers a distinct competitive advantage in operating a domestic network something American would be denied absent approval and immunity for the proposed oneworld alliance on equal terms with Star and SkyTeam. American s ability to expand internationally and thus support a robust domestic network is dependent upon alliance development (Id. at 30-31). PUBLIC VERSION

25 - 23 C. Scope of Immunity In Star II, the Department rejected DOJ's argument that the grant of immunity should be limited to specifically identified markets, finding instead that the public would benefit from the parties' global cooperation and the competition that such cooperation would create among alliances. The Department found that, "granting immunity beyond transatlantic markets will enhance the ability of immunized Star carriers to cooperate globally outside of the joint venture and will assist [them] in their efforts to formulate joint ventures in other regions of their combined networks, thereby promoting greater service benefits to consumers" (Order , p. 22). The current proceeding, because it involves only one U.S. airline, does not raise the same initial concerns as Star II on routes beyond the transatlantic. Nevertheless, the Joint Applicants wish to clarify that by seeking approval of and antitrust immunity for their multilateral, Alliance Coordination Agreement (Joint Application, Exhibit JA-1), they have sought a global grant of immunity as well. The Joint Applicants expect that as oneworld grows, and as the Department's efforts to obtain Open Skies agreements with liberal seventh freedom traffic rights succeeds, there may be opportunity for the parties to expand their cooperation in ways that will benefit U.S. consumers. Immunizing the oneworld parties now, on a global basis will facilitate future market entry and enhance inter-alliance competition. PUBLIC VERSION

26 - 24 CONCLUSION The Department s examination of an international alliance must look at the broader networks offered by its members. In this case, the Joint Applicants have presented the Department with more evidence than any other alliance demonstrating the existence of public benefits and a lack of harm to competition. No other alliance has provided internal fare data. No other alliance has performed specific analyses of consumer benefits. No consumer groups have opposed this alliance, and the only competitor objecting is Virgin Atlantic which as a strong point-to-point competitor on long-haul London routes has a self-interested motive to keep oneworld weak. oneworld in its current form simply cannot compete effectively with Star and SkyTeam because the lack of antitrust immunity puts the Joint Applicants at a severe regulatory disadvantage. Consumers lose out every day that this unfair, inequitable and unsustainable situation persists. PUBLIC VERSION

27 - 25 Accordingly, we request that the Department move quickly to issue an Order to Show Cause approving the proposed oneworld alliance and the American/British Airways/Iberia Joint Business Agreement. Respectfully submitted, CARL B. NELSON, JR. PAUL C. JASINSKI Associate General Counsel General Counsel, Americas R. BRUCE WARK JAMES B. BLANEY Associate General Counsel Senior Counsel, Americas JEFFREY A. OGAR British Airways plc Senior Attorney American Airlines, Inc. DARYL A. LIBOW Sullivan & Cromwell LLP ROGER W. FONES DON H. HAINBACH Morrison & Foerster LLP Garafalo Goerlich Hainbach PC ANDREW B. STEINBERG Attorneys for British Airways Jones Day Attorneys for American Airlines WILLIAM KARAS CAROL GOSAIN Steptoe & Johnson LLP Attorneys for Iberia HENRIK ARLE Deputy Chief Executive Officer SAMI SARELIUS Vice President & General Counsel Finnair Oyj MICHAEL J. HOLLAND Condon & Forsyth LLP Attorneys for Royal Jordanian September 8, 2009 PUBLIC VERSION

28 EXHIBIT 1

29 Competitive Effects of Airline Antitrust Immunity I. Qualifications A. Robert Willig 1. I am Professor of Economics and Public Affairs at Princeton University where I have held a joint appointment in the Economics Department and at the Woodrow Wilson School of Public and International Affairs for 30 years. I have authored some 75 articles in the economics literature and a book on competition and the theory of industrial market structure. I am co-editor of the Handbook of Industrial Organization, which summarizes the state of economic thinking on the structure of industries and the nature of competition among firms. 2. I have extensive experience analyzing economic issues arising under the law. While on leave from Princeton, I served as the Deputy Assistant Attorney General in the Antitrust Division of the United States Department of Justice from 1989 to 1991, and in that capacity served as the Division s Chief Economist. I have also served as a consultant to the Federal Trade Commission, the United States Department of Justice, The World Bank, OECD, the InterAmerican Development Bank, and many private parties on competition, regulation and other policy issues. In addition, I have appeared as an expert witness before Congress, federal and state courts, federal administrative agencies, and state public utility commissions on subjects involving competition, regulation, intellectual property rights, and antitrust. My work has involved many different industries and sectors, including major projects for the Department of Justice and for private parties involving airline mergers and alliances, airport slot issues, airline pricing and capacity deployment, and the operations of computer reservation systems. B. Mark Israel 3. I am a Senior Vice President at Compass Lexecon and a Visiting Associate Professor of Management and Strategy at the Kellogg School of Management at Northwestern University. From , I served as a full time member of the Kellogg faculty. I received my Ph.D. in economics from Stanford University in At Kellogg and Stanford, I have taught graduate level courses in business strategy and economics. In my academic research, I specialize in the economics of industrial organization, which is the study of individual markets and includes the study of antitrust and 1

30 regulatory issues, as well as the economics of information and insurance markets. My research has been published in leading economics journals including the American Economic Review and the Rand Journal of Economics. 5. I have worked in consulting at Compass Lexecon since 2005, where I have applied theoretical and empirical methods to the analysis of mergers and related antitrust issues, intellectual property, class certification, and damages calculations, in a range of industries including airlines, consumer products, financial markets, mining, payment cards, pharmaceuticals, publishing, and various high technology industries. C. Bryan Keating 6. I am a Senior Economist at Compass Lexecon, where I have worked since I received my Ph.D. in economics from Stanford University in At Compass Lexecon, I have conducted economic and econometric analysis in support of litigation, arbitration/settlement discussions, and efforts to secure regulatory approval for mergers. I have substantial experience designing and implementing complex econometric models using large-scale databases, especially in industries that involve differentiated products. I have analyzed issues relating to market definition, competitive effects, welfare analysis and merger simulation in a wide variety of industries including airlines, computer software and hardware, consumer products, health care, payment cards, sports, and telecommunications. II. Overview of Empirical Findings 7. As part of its June 26, 2009 comments on the Department of Transportation s Show Cause Order regarding the Star Alliance, the Department of Justice (DOJ) included two results from econometric studies of airline fares: For nonstop routes, DOJ used a cross-sectional comparison of average fares on different routes in the 3 rd quarter of 2008 to argue that a reduction in the number of competing airlines offering nonstop transatlantic flights may result in large, statistically significant price increases. 1 For connecting routes, DOJ compared ticket prices for different types of service (using 3rd quarter data for from connecting behind routes) to argue that, on 1 Comments of the Department of Justice on the Show Cause Order, Public Version, Docket OST , June 26, 2009, p. 47 ( DOJ Comments ). 2

31 average, on given routes, fares charged by alliances with antitrust immunity ( ATI ) are higher than non-immunized fares. 2 DOJ used these econometric conclusions as evidence to support its position that granting immunity would risk significant competitive harm in certain markets 3 and that granting immunity may not be necessary to achieve benefits to consumers In this affidavit, we present results from a series of analyses, which together demonstrate that DOJ s econometric findings should not be relied upon for the formulation and support of policy towards airline alliances. We start by noting that DOJ s study of non-stop fares is not, in fact, a study of the effect of alliances, but rather is a study of the effect of any change in the number of distinct carriers serving a route. In particular, DOJ simply assumes, without justification, that granting antitrust immunity to two carriers on a given route has the same effect on fares as an actual reduction in the number of carriers, such as would occur following exit by a carrier. However, there are fundamental economic distinctions between these situations, meaning that the effect of ATI may quite possibly be significantly different from the effect of exit. In particular, as explained in the affidavit of Don Casey, Vice President for Revenue Management at American Airlines, one expected result of joint decision making by carriers in an immune alliance is to create strong incentives (due to network effects) to expand capacity on those transatlantic routes that serve an important role in carrying connecting traffic over the combined network (such as the main routes at issue in this matter). Such capacity expansions naturally put downward pressure on fares, an effect not present for situations in which a carrier exits a route. 9. To address the relevant question for alliance policy whether or not granting antitrust immunity generates higher fares on transatlantic non-stop overlaps we specify and implement an econometric analysis that controls for the total number of carriers on a route (whether in an alliance or not) and then asks whether fares are higher when antitrust immunity is granted to two of those carriers. We show, across a range of model 2 DOJ Comments, p DOJ Comments, p DOJ Comments, p On this point, note that the DOJ empirical evidence only evaluated relative fares across different routes and different types of service. Even taken at face value, it provides no evidence about the improved service quality and associated consumer benefits enabled by ATI. In contrast, our results, presented below, directly quantify the substantial consumer benefits created by permitting ATI service. 3

32 specifications, that there is no evidence that antitrust immunity leads to any increase in fares on overlap routes. 10. In addition to separating the effect of ATI from the effect of actual reductions in the number of distinct carriers serving a route, our econometric analysis also improves on DOJ s specification by extending the data beyond DOJ s needlessly-limited study period (a single quarter, Q3 2008) and including sufficient (and standard) controls for the myriad factors that lead to different fares for different routes. With these additional data and controls in place, we re-examine the effect of reductions in the number of carriers serving a route (as distinct from the effect of granting ATI to two of those carriers). We show that the measured effect of going from two-to-one carriers on a route is substantially smaller than found by DOJ and statistically insignificant in all specifications. 5 For a reduction from three to two carriers, we find that the inference of higher fares is not robust across alternative specifications, 6 with the estimated effect ranging from statistically indistinguishable from DOJ s result to small and statistically insignificant (as DOJ s own result was). 11. With regard to the effect of antitrust immunity on connecting traffic, we first show that DOJ s finding of higher fares for ATI service holds only when one fails to include sufficient (and standard) controls to account for the various factors that lead to different fares for different types of service. We then go further to examine additional issues that the DOJ did not consider, but that are highly relevant to the policy questions addressed by the DOJ, including the consumer benefits from better service quality via ATI (rather than interline or code-share) service. In particular, for connecting routes: We show that following the professional literature by adding additional controls to the DOJ econometric analysis restores the policy-important result that has generally been 5 An impact or effect (in this case, the effect of changing the number of competitors on a route on average fares on the route) is said to be statistically insignificant if there is not enough information from the data analysis to conclude with confidence that the effect is not zero. Throughout this Affidavit, we rely on the standard 95% confidence level. So we say an effect is statistically insignificant if we cannot conclude with 95% confidence based on standard statistical tests that determine the level of statistical confidence that the effect is not zero. If we can conclude with at least 95% confidence that the effect is not zero, it is said to be statistically significant. 6 A statistical result is said to be robust if it continues to hold under a range of reasonably minor modifications to the base specification. So, for example, if a result continues to hold for a somewhat different time period or when a slightly different set of variables are controlled for, we say it is robust. Otherwise, if such changes affect the substance of the result, we say it is not robust. 4

33 found in the literature ATI itineraries have lower fares than code-share (or interline) itineraries. 7 We present an econometric demand analysis that demonstrates that, holding all else (including fares) equal, consumers prefer ATI service to code-share (or interline) service, with the extent of this preference equivalent to the preference consumers have for itineraries with substantially reduced fares. 12. Our econometric analyses also indicate that carving-out particular segments from immunity diminishes or eliminates the benefits from ATI for those connecting itineraries that flow over a carved-out segment. In particular, we show that otherwise ATI itineraries including a carved-out segment have higher fares than ATI itineraries with no such carve out. In addition, we show that the extent of consumer benefits from ATI service is substantially weakened on itineraries involving carved-out segments. These results showing that carve-outs substantially diminish the significant benefits to consumers from ATI take on particular importance in the light of our findings on the lack of effect of ATI on non-stop fares. Since the data do not show any impact on non-stop fares from granting antitrust immunity, there is no support in the empirical evidence for carving routes out from such immunity for the purported purpose of holding fares down with thereby preserved competition. Instead, the empirical evidence indicates that carve-outs are not only unnecessary to hold fares down, but are also harmful to consumers by vitiating the fare and quality-of-service benefits that connecting passengers receive from full ATI. We conclude the affidavit by quantifying the harm that would occur to travelers on those connecting itineraries that use the four oneworld overlap segments that currently have three or fewer independent carriers (the type of routes that DOJ focused on in its comments on the Star alliance) if those routes were to be carved out from immunity. 8 III. The Effects of Antitrust Immunity on Fares on Non-Stop Routes 13. Using data from the Airline Origin and Destination Survey (DB1B) and the Official Airline Guide (OAG) from the third quarter of 2008, DOJ conducted a regression analysis of the effect of the number of competitors on non-stop fares for 65 transatlantic routes involving hubs. Of particular note, DOJ treated two carriers with antitrust immunity on a particular route as though they were only one competitor, thereby assuming that ATI has the same 7 See, e.g., W. Tom Whalen (2007), A panel data analysis of code-sharing, antitrust immunity, and open skies treaties in international aviation markets, Review of Industrial Organization, 30: These routes are DFW-LHR, MIA-LHR, BOS-MAD, and MIA-MAD. 5

34 effect as exit from the route. With this methodology, DOJ found that reducing the number of nonstop competitors on a route from two to one increased average fares by a statistically significant 15 percent. It also found that reducing the number of competitors from three to two increased fares by a statistically insignificant 6.6 percent. DOJ used this result to support the inference that granting ATI may lead to higher fares on non-stop overlap routes Based on the analyses presented on the following pages, we conclude that DOJ s study does not, in fact, measure the effect of ATI as distinguished from actual changes in the number of carriers on a route on non-stop fares. And we reject the notion that DOJ s findings can in any way support an inference that ATI leads to higher non-stop fares on overlap routes. An important implication of this conclusion is that, because ATI has no significant effect on nonstop fares, there is no basis to support carve-outs from immunity as a means to reduce nonstop fares. 10 In addition, our analysis makes use of additional quarters of data and more complete standard controls for differences among routes. This improved specification demonstrates that DOJ s findings on the fare effects from reducing the number of non-stop competitors on a route are not robust. 15. Before turning to our own analysis, we attempt to replicate the DOJ methodology and results. While we do not have full information on all steps taken by DOJ, we have been able to use what is essentially the DOJ methodology (based on the information in DOJ s comments) to find results that are fairly close to those found by DOJ. 11 In particular, as seen in Table 1, with this methodology we find a fare effect of 10.4 percent when the number of competitors 9 Details in this paragraph are drawn from DOJ Comments, pp We have separately tested and confirmed the result that carve-outs do not significantly reduce nonstop fares. 11 We attempt to construct the data in the same way described in the DOJ Comments and to apply the same filters. In particular, we focus on the same routes that DOJ listed in the attachment to the DOJ Comments. However, because DOJ listed Detroit Frankfurt twice, our analysis includes only 64 routes. We also mimic DOJ s approach on the following dimensions. We focus on non-directional city pairs and split round-trips into one way tickets. For round-trip tickets, we divide the fare by two to get one-way fares. We focus on coach-class tickets and filter out routes with fares less than $50. We count a carrier as operating on a route if it operates at least 60 flights on the route in a quarter. We treat immunized members of the same alliance as a single carrier, unless the route has been carved out from the immunity for the alliance. As a result of our attempt to match DOJ s data processing methodology, for the 3 rd quarter of 2008 we closely match DOJ s carrier counts. We have 17 routes with a single carrier whereas DOJ has 18. We have 27 routes with two carriers whereas DOJ has 28. We have 14 routes with three carriers whereas DOJ has 13. We both have four routes with four carriers. We have one route with five carriers whereas DOJ has zero. Finally, we have one route with six carriers, whereas DOJ has two. 6

35 declines from two to one, 12 and a fare effect of 5.7 percent (statistically insignificant as in DOJ s own results) when the number of competitors declines from three to two. Table 1: DOJ Methodology Price Effects from Loss of Nonstop Competition on Transatlantic Routes The Dependant Variable is ln(average Fare) Explanatory Variables Estimate Standard Error Estimated price effect in percentages >= >= % Number of Nonstop Competitors >= % >= % 5 or more * ln(mileage on the Route) * Mean of City Populations in the Route Constant Term Data Route Time Period 2008q3 Year-Quarter Fixed Effects No Carrier Fixed Effects No Route Fixed Effects No Notes: Robust standard errors computed * indicates statistical significance at the 5% level Percentage effects calculated as exp(-coef)-1 We allow for a linear effect of adding an additional carrier beyond five. 16. To address the issues with DOJ s approach that are introduced above, we implement a new econometric specification, which makes three basic changes to DOJ s specification. First, we do not impose on the model the assumption that ATI has the same effect as actual reductions in the numbers of distinct carriers on a route. Instead, we allow the data to determine whether this is true, by controlling for the total number of carriers on the route (whether in an immune alliance or not) and then including an indicator (or dummy ) variable for routes on which two of the carriers are in an immune alliance. In this way, the coefficient on the dummy variable directly measures the effect (if any) of ATI on fares. 12 Unlike DOJ, we find that this result falls just short of the standard for statistical significance (based on a 95% confidence level). 7

36 17. Second, in order to measure the effect of ATI on just those routes about which DOJ has expressed concern those served by three or fewer carriers we only include route-quarters with one, two, or three non-stop competitors in our analysis Finally, the estimation approach employed by DOJ relies on cross-sectional variation across routes in the number of carriers, and the associated variation in average fares, during a single quarter. For example, this method relies on a comparison of the average fare on Atlanta Rome (which had one carrier in the 3 rd quarter of 2008) with the average fare on Atlanta Frankfurt (which had two carriers in the 3 rd quarter of 2008), as well as many other similar cross-route fare comparisons. A potential problem with this approach is that it may be difficult to fully control for all other important differences across routes, as would be required to separate the effect of the number of competing carriers on fares from other economic factors affecting fares. To account for these differences, we implement a standard solution using route fixed effects, to control for all differences across routes, which means that we base our identification of fare impacts on changes in the number of competitors on each route over time. 14 In order to implement this method, we extend the single-quarter dataset employed by DOJ to form a panel dataset covering 2005:Q1 to 2008:Q Table 2 reports the carrier counts for each route and quarter in the data along with the total number of changes (between one and two carriers, or between two and three carriers) on each route. These changes and any associated changes in fares are the source of variation used to identify the effect of changing the number of competitors on fares. Table 3 lists all examples of route-quarters with three or fewer competitors on which two of the competitors are in an immune alliance. The presence of such routes (and changes in the immunity relationships on given routes over time) enables us to measure the effect of ATI on fares after controlling for the number of carriers on the route. 13 We define our data by the combination of route and quarter (route-quarter). So, for example, we would keep an observation for Atlanta Amsterdam in the 4 th quarter of 2008 because we observe two carriers operating on that route in that quarter. 14 In our specification, we allow the route fixed-effects to vary by quarter, to allow for any routespecific seasonal effects. We have also estimated the model with constant route fixed effects, which do not depend on the quarter. None of our substantive conclusions are changed if we use this alternative specification. 8

37 Table 2: Carrier-Counts by Route and Quarter Within Quarter (Year on Year) Total Changes Changes (2:1, Route 2005Q1 2005Q2 2005Q3 2005Q4 2006Q1 2006Q2 2006Q3 2006Q4 2007Q1 2007Q2 2007Q3 2007Q4 2008Q1 2008Q2 2008Q3 2008Q4 (2:1, 3:2) 3:2) Atlanta-Amsterdam Atlanta-Frankfurt Atlanta-London Atlanta-Paris Atlanta-Rome Chicago-Amsterdam Chicago-Dublin Chicago-Frankfurt Chicago-London Chicago-Manchester Chicago-Munich Chicago-Paris Chicago-Rome Cincinnati-Amsterdam Cincinnati-Paris Cincinnati-Rome Denver-London Detroit-Amsterdam Detroit-Frankfurt Detroit-London Detroit-Paris Houston-Amsterdam Houston-London Houston-Paris Los Angeles-Frankfurt Los Angeles-London Memphis-Amsterdam Miami-London Miami-Madrid Miami-Paris Minneapolis-Amsterdam Minneapolis-Paris New York-Amsterdam New York-Athens New York-Barcelona New York-Berlin New York-Brussels New York-Budapest New York-Copenhagen New York-Dublin New York-Edinburgh New York-Frankfurt New York-Lisbon New York-London New York-Madrid New York-Manchester New York-Milan New York-Paris New York-Rome New York-Shannon New York-Stockholm New York-Zurich Philadelphia-Frankfurt Philadelphia-London Philadelphia-Paris Salt Lake City-Paris San Francisco-Frankfurt San Francisco-London Washington-Amsterdam Washington-Frankfurt Washington-London Washington-Munich Washington-Paris Washington-Zurich Total Changes

38 Table 3: ATI Alliance Partners on Routes with 2-3 Carriers Route Time Period Carriers ATI Atlanta-Amsterdam 2008q2-2008q4 AF/KL DL Skyteam (AF/KL, DL) Chicago-Munich 2005q3-2008q4 LH UA Star (LH, UA) Detroit-Amsterdam 2008q2-2008q4 AF/KL NW Skyteam (AF/KL, NW) Detroit-Paris 2008q2-2008q3 AF/KL NW Skyteam (AF/KL, NW) Los Angeles-Frankfurt 2008q1-2008q3 LH UA Star (LH, UA) New York-Milan 2005q1-2008q1 AZ CO DL Skyteam (AZ, DL) New York-Rome 2005q1-2008q1 AZ CO DL Skyteam (AZ, DL) New York-Zurich 2005q1-2006q4 AA CO LX Oneworld (AA, LX) San Francisco-Frankfurt 2005q1-2008q4 LH UA Star (LH, UA) Washington-Amsterdam 2008q4 AF/KL NW UA Skyteam (AF/KL, NW) Washington-Munich 2005q3-2008q3 LH UA Star (LH, UA) 20. Results from our improved specification implementing these three changes to DOJ s analysis are presented in Table The main conclusion is clear. The measured effect of ATI on non-stop fares is almost exactly zero and clearly statistically insignificant. In addition, the measured effect of reducing the actual number of carriers (whether or not they are in an immune alliance) from two to one is a statistically insignificant 5.5% (as opposed to 10.4% for changes from two-to-one carriers in our version of DOJ s model). The measured effect of going from 3 to 2 carriers is 7.7%, which is not statistically distinguishable from the 5.7% three-to-two effect in our implementation of DOJ s model. 15 As indicated in the table, we also include year and quarter fixed effects to account for time-series changes in average fares that are common across routes. In addition, we include carrier fixed effects in the form of the share of each carrier on the route to account for changes in fares that are due to changes in the identity of the competitors on a route, rather than the competitive effects associated with changing the number of carriers. 10

39 Table 4: ATI Effects on Non-Stop Fares (with Route-Quarter Fixed Effects) Price Effects from Loss of Nonstop Competition on Transatlantic Routes The Dependant Variable is ln(average Fare) Explanatory Variables Estimate Standard Error Estimated price effect in percentages >= Number of Nonstop Competitors >= % >= * 7.7% ATI % Constant Term * Data Route Time Period All Qtrs Year-Quarter Fixed Effects Yes Carrier Fixed Effects Yes Route-Quarter Fixed Effects Yes Notes: Robust standard errors computed * indicates statistical significance at the 5% level Percentage effects calculated as exp(-coef) While the inclusion of route fixed-effects most fully controls for differences among routes, a potential concern with this approach might be that there are not sufficient changes in the presence of ATI competitors on particular routes to precisely measure the effect of ATI on non-stop fares. To address this potential concern, we have estimated an alternative regression specification in which we continue to use the more inclusive dataset (from Q Q4 2008) but do not include route fixed effects. 16 Instead, we include additional standard controls to account for differences among routes. For example, we agree with DOJ that it is important to control for the route distance, as this is certainly related to fares. However, to allow for a more flexible relationship between distance and fares than permitted by a simple linear or logarithmic specification, we specify the relationship between average fares and 16 Without route fixed effects, the measured fare effects are based not only on fare changes associated with changes in the market structure on a given route over time, but also on fare differences across routes with different number of carriers. Given that the two methods rely on distinct sources of variation, it is not surprising that they produce somewhat different results for the fare effect of reducing the number of carriers on the route. Critically, however, results from either specification confirm the lack of any effect of ATI on average fares on a route. 11

40 distance as a five-part spline. 17 In addition, the simple fact that there are different carriers serving different routes may lead to differences in fares. So, as in the specification used for Table 4, we control for the percentage of traffic on each route served by different carriers As seen in Table 5, this alternative specification which is closer to DOJ s original specification in that it makes use of cross-sectional variation across routes, while improving that specification by using a more inclusive dataset and a more appropriate set of controls than those used by DOJ bolsters our findings. Once again, the measured effect of ATI is almost exactly zero and clearly statistically insignificant. In this case, the measured effect of an actual reduction from two to one carriers (as distinct from granting antitrust immunity to the carriers) falls to 3.6%, while the measured effect of going from three-to-two carriers falls to 1.1%, with both results statistically insignificant. 17 The spline allows the effect of distance on fare to enter the estimation in a piecewise linear fashion rather than linearly. For example, it allows the effect of changing distance from 500 to 550 miles to differ from the effect of changing distance from 2000 to 2050 miles. Our substantive conclusions are unchanged if we include the log of distance in place of the spline. 18 As in Table 4, we also include year and quarter fixed effects to control for changes in fares over time that are common across different routes. 12

41 Table 5: ATI Effects on Non-Stop Fares (without Route-Quarter Fixed Effects) Price Effects from Loss of Nonstop Competition on Transatlantic Routes The Dependant Variable is ln(average Fare) Explanatory Variables Estimate Standard Error Estimated price effect in percentages >= Number of Nonstop Competitors >= % >= % ATI % Mean of City Populations in the Route Mileage on the Route - Spline * Mileage on the Route - Spline * Mileage on the Route - Spline * Mileage on the Route - Spline Mileage on the Route - Spline Constant Term * Data Route Time Period All Qtrs Year-Quarter Fixed Effects Yes Carrier Fixed Effects Yes Route Fixed Effects No Notes: Robust standard errors computed * indicates statistical significance at the 5% level Percentage effects calculated as exp(-coef) In sum, then, our results do not support DOJ s assumption that granting antitrust immunity has the same effect on fares as an actual reduction in the number of carriers. Instead, the results clearly demonstrate the lack of any measurable effect of antitrust immunity on nonstop fares. A closely related corollary result is that imposing carve-outs on some or all of these overlap routes will have little or no beneficial effect on fares on these routes. In addition, we find no statistically significant fare effect from reducing the actual number of carriers on a route from two to one (as distinct from granting antitrust immunity to the carriers). For reductions from three to two carriers, we show that DOJ s findings on fare effects are not robust across specifications. 13

42 IV. The Effects of Service Type on Fares on Connecting Routes 24. The economics literature contains a number of articles that examine the question of whether provision of antitrust immunity will reduce fares on connecting routes by mitigating the double marginalization problem associated with interline (or code-share) service. 19 The double marginalization problem occurs because, absent an ability to cooperate on prices, carriers have an incentive to maximize profits on their segment(s) of an interline itinerary, taking as given the fare charged by the other carrier(s) on the itinerary. Such separate pricing generally leads to higher combined prices than the carriers would agree to if they could jointly set prices to maximize overall profits on the route. 25. The literature has generally concluded that ATI helps to solve this double marginalization problem by enabling harmonized carrier decisions on pricing and capacity deployment on complementary segments, and that, as a result, fares on ATI itineraries are generally lower than fares for comparable code-share (or interline) service. However, the literature relies for its empirical analyses on older data, generally from the 1990s. In its comments, DOJ presented an analysis of fares by service-type using more recent data, from the 3 rd quarter for 2005 through Based on this study, DOJ concluded that immunized alliance fares are higher than non-immunized ones It is crucial to note that the DOJ study not only uses more recent data, it also employs a different specification than that found in previous papers (such as Whalen, 2007) published in the peer-reviewed professional literature. For example, DOJ does not include many of the control variables that Whalen employed. In this section, we demonstrate that more recent data is not the reason for DOJ s conclusion that, contrary to the previous literature, immunized fares are higher than non-immunized ones. Instead, the reason for DOJ s results is its use of a specification that fails to control for the same factors as have been controlled for in the relevant academic literature. When we utilize the same more recent data employed by DOJ, and control for the same factors as in the literature, we find confirmation of the literature s results that ATI service lowers fares compared to code-sharing and interline service. 19 See, e.g., Jan Brueckner and W. Tom Whalen (2000), The Price Effects of International Alliances, Journal of Law and Economics, 43, ; Jan Brueckner (2003), International Airfares in the Age of Alliances, Review of Economics and Statistics, 85, ; W. Tom Whalen (2007), A panel data analysis of code-sharing, antitrust immunity, and open skies treaties in international aviation markets, Review of Industrial Organization, 30: DOJ Comments, p

43 27. DOJ found that code-share fares are 1.5 percent below online fares (although the difference is not statistically significant), and ATI fares are 2.1 percent higher than online fares (and the difference is statistically significant). 21 While we do not know the precise details of DOJ s methodology, Table 6 shows that we are able to generate similar results using the same data period and (what we believe to be) the same set of controls We focus on DOJ s results based on coach-class tickets in Table 2 of the DOJ Comments. DOJ s results based on all tickets are similar. 22 It is unclear whether DOJ included carrier fixed-effects in its analysis. Table 2 refers to carrier fixed-effects, but they are not mentioned in the text of the DOJ Comments. In our analysis we only find higher ATI fares if these carrier fixed-effects are excluded. We define carrier fixed effects on the basis of the marketing carriers for each itinerary. We did not try to exactly replicate DOJ s dataset, but rather followed Whalen (2007). This creates some differences in the structure of the data. For example, DOJ focuses on round-trip travel whereas we break itineraries into one-way components. In addition, DOJ treats each ticket in the data as a separate observation whereas we aggregate to the route-carrier-service type level. We follow both DOJ and Whalen (2007) in focusing on behind routes that include at least one domestic segment in order to ensure that we have complete data for the route. We also exclude any routes that include non-stop travel. We limit our data to itineraries with no more than six coupons. We also focus on coach ( X ) class itineraries and exclude itineraries with non-credible fares, round-trip fares less than $100 and round-trip fares greater than or equal to $9999. We match DOJ s definitions of online, ATI, code-share, and interline. In particular, we define an itinerary as online if all operating and marketing carriers are the same after assigning regional carriers to their parents. We define an itinerary as ATI if it is not online and all operating and marketing carriers are immunized members of the same alliance. We define an itinerary as code-share if it is not online or ATI and all operating and marketing carriers are non-immunized members of the same alliance. We classify the remaining itineraries as interline. We also limit our analysis to routes with more than 50 passengers in a quarter. Our substantive conclusions are not changed if we include all the routes in our analysis. 15

44 Table 6: DOJ Methodology Price Effects of Service Type on Transatlantic Routes The Dependant Variable is ln(average Fare) Explanatory Variables Estimate Standard Error Estimated Price Differentials (Relative to Online) Online Service Type ATI * 4.7% Codeshare * -2.4% Interline * 8.4% Coupons * Mileage on the Route * Constant Term * Data Route-Carrier-Service Type Time Period 3rd Qtr, All Years Year Fixed Effects No Carrier Fixed Effects No Route Fixed Effects Yes Notes: Robust standard errors computed * indicates statistical significance at the 5% level Percentage effects calculated as exp(coef) To investigate the reason for DOJ s finding higher measured fares on ATI (relative to codeshare) itineraries, we add additional controls, following those used in Whalen (2007). In particular, we include the income and population of the US and EU endpoints, an HHI for online and alliance service and a separate HHI for interline service, an indicator for routes between Open Sky signatories, carrier fixed-effects, and year fixed-effects. 23 Table 7 shows that when we include these additional controls the result that has generally been found in the literature is restored. In particular, we find that ATI fares are, on average 6.0 percent lower than interline fares and 3.4 percent lower than code-share fares, with both differences statistically significant. 23 Like Whalen (2007), we use MSA-level data from the Census Bureau and the Bureau of Economic Analysis on income and population for U.S. cities and country-level data from Eurostat for European cities. We also follow Whalen (2007) in treating immunized members of the same alliance as a single carrier for the purposes of computing HHI, while dividing passengers equally amongst non-immunized alliance members. 16

45 Table 7: Panel Data with Whalen Controls Price Effects of Service Type on Transatlantic Routes The Dependant Variable is ln(average Fare) Explanatory Variables Estimate Standard Error Estimated Price Differentials (Relative to Online) Online Service Type ATI * 3.2% Codeshare * 6.8% Interline * 9.5% Coupons * Mileage on the Route One-way Flight * HHI OA HHI Interline Open Sky * Population of US City Income of US City Population of EU City Income of EU City * Constant Term * Data Route-Carrier-Service Type Time Period 3rd Qtr, All Years Year Fixed Effects Yes Carrier Fixed Effects Yes Route Fixed Effects Yes Notes: Robust standard errors computed * indicates statistical significance at the 5% level Coupons, mileage, and oneway are computed as passenger-weighted averages for each route-carrier Percentage effects calculated as exp(coef) These results indicate that more recent data actually confirms the results from the literature as long as the same approach controlling for the same factors is used. As expected, ATI reduces fares relative to code-share or interline service, likely because it (at least partially) mitigates the double marginalization problem by allowing harmonization of carriers pricing and capacity deployment decisions on complementary segments. Furthermore, code-share service, which does not provide the same ability to coordinate on pricing, does not provide the same benefits. 17

46 V. The Effects of Carve-outs on Fares and Quality on Connecting Routes 30. In the remainder of this affidavit, we examine issues that DOJ did not address in the econometric results presented in its comments. These include the consumer benefits from better service quality via ATI (rather than interline or code-share) service and the extent to which the benefits from ATI service are lessened or eliminated when an otherwise ATI itinerary includes a carved-out segment. 31. To focus our findings on a policy issue of particular interest whether certain non-stop overlaps should be carved out from immunity we organize our discussion primarily around the effects of carving out particular segments on the otherwise beneficial impacts of immunity. We reach several conclusions: As seen above, on nonstop routes, granting antitrust immunity has no significant effect on fares, implying that carve-outs will also have no significant effect on fares. On connecting routes, the fares on an otherwise ATI itinerary including a carved-out segment are higher than fares on itineraries with full ATI service. Holding fares fixed, consumers prefer full ATI service, not only to code-share and interline service, but also to otherwise ATI service including a carved-out segment. The difference in consumer preference for full ATI service rather than service using a carved-out segment is equivalent to more than $60, which indicates that the quality of full ATI service is substantially higher. By raising fares and reducing quality on connecting itineraries (flown by American Airlines, British Airways, or Iberia Airlines), imposing carve-outs from immunity on the four oneworld overlap routes with three or fewer distinct carriers (the type of routes on which DOJ has focused attention) would generate harm to consumers equivalent to at least $55 million annually. Thus, the data indicate that imposing carve-outs will impose substantial harm on consumers (relative to full ATI) while producing no offsetting benefits to consumers. A. Fares on full ATI versus ATI carve-out itineraries 32. We start our analysis by explicitly splitting ATI service into two types connecting service that includes a segment with a carve-out and connecting service that does not include any such segment. In this way, we define five types of service online, ATI, ATI carve-out, 18

47 code-share, and interline. 24 We then re-run the fare analysis (presented in Table 7 and described in paragraph 29-30) for these five types of service. Table 8 shows that not only does full ATI service have significantly lower fares than code-share or interline service, it also has significantly lower fares than otherwise ATI service including a carved-out segment. Hence, this is evidence that carve-outs significantly increase fares on connecting routes relative to full ATI service. Accordingly, imposition of carve-outs has a real down-side for consumers. Moreover, there is no evidence of a countervailing upside, since the data show no significant ATI-related fare increases on non-stop routes that might otherwise be thought to suggest imposition of carve-outs as a remedy. 24 We observe four carved-out segments in our data: 1) DL/AF on Atlanta Paris; 2) DL/AF on Cincinnati Paris; 3) UA/LH on Chicago Frankfurt; and 4) UA/LH on Washington, DC Frankfurt. 19

48 Table 8: Allowing for a Carve-Out Effect on Connecting Routes Price Effects of Service Type on Transatlantic Routes The Dependant Variable is ln(average Fare) Explanatory Variables Estimate Standard Error Estimated Price Differentials (Relative to ATI carve-out) Online * -5.2% ATI * -2.7% Service Type ATI carve-out Codeshare * 1.3% Interline * 3.9% Coupons * Mileage on the Route One-way Flight * HHI OA HHI Interline Open Sky * Population of US City Income of US City Population of EU City Income of EU City * Constant Term * Data Route-Carrier-Service Type Time Period 3rd Qtr, All Years Year Fixed Effects Yes Carrier Fixed Effects Yes Route Fixed Effects Yes Notes: Robust standard errors computed * indicates statistical significance at the 5% level Coupons, mileage, and oneway are computed as passenger-weighted averages for each route-carrier Percentage effects calculated as exp(coef)-1 B. Consumer benefits from full ATI service versus ATI carve-out service 33. Next, we examine the relative quality levels of the different service types to determine whether ATI service appears to be of higher quality, thus generating consumer benefits beyond just lower fares. To do so, we estimate a nested-logit demand model to explain shares of each carrier s itineraries on each route. 25 The model specifies the probability that a 25 See Steven Berry (1994), Estimating Discrete-Choice Models of Product Differentiation, RAND Journal of Economics, 25:4, In our model, the option not to fly comprises one nest, with flights by all carriers in a second nest. 20

49 consumer chooses a particular itinerary option (or not to fly at all) as a function of the fare, 26 total itinerary distance, number of coupons, service type (online, interline, ATI, ATI carveout, or code-share), and fixed effects for year, route, and carrier. We estimate the model using the same dataset as used for the fare regressions described above. 34. Table 9 presents results from this model. 27 We find that, for a given fare, consumers most prefer the characteristics associated with flying online, followed by ATI, then ATI carve-out, then code-share, and then interline. 28 This ranking is consistent with our expectations. Moreover, the specification of the model allows us to monetize the magnitude of the additional benefit consumers receive all else (including fares) equal due to the evidently superior quality offered by particular types of service. 29 In particular, the results shown in Table 8 imply that consumers would be willing to pay an extra $175 (28 percent of average fares in the data) to fly an ATI itinerary relative to a code-share itinerary. 30 Put differently, the characteristics of ATI service provide $175 worth of additional value to consumers relative to code-share service. However, the benefits to flying an ATI itinerary diminish by $67 (about one-third of the value of ATI relative to code-share service) down to a gap of only $108 relative to code-share on otherwise ATI itineraries including a carved-out segment In estimation, we instrument for fares (and the nested logit term) using fuel prices as a costshifter. We interact the fuel price with the carrier fixed effects, to allow for differential effects of higher fuel prices on different carriers, due to different equipment and other differences. 27 When we run the demand model on all routes, including thinly-travelled routes, we find the same pattern of consumer benefits, but greater magnitudes. 28 Because we lack data on the underlying characteristics of the different service types, we cannot determine which characteristics are driving consumers preferences. However, factors such as scheduling convenience, network breadth, frequent flier programs and airport amenities and ease of connections are all likely to affect consumers decisions about which itinerary to choose. 29 In particular, the ratio of the difference in the coefficients for two service types to the coefficient on fare measures the amount by which fares would need to decrease in order to provide the same level of benefit to consumers as is gained by switching from one service type to another. In that sense, it provides a dollar-equivalent for the consumer benefits from the improved service quality. 30 All our fare results are based on a single direction of flight. So for a round trip, the dollar values would be doubled. 31 The results also imply that consumers would pay an additional $235 (37 percent) to fly online rather than ATI. 21

50 Table 9: Nested Logit Demand Model Demand Estimation The Dependant Variable is ln(share_j/share_0) Explanatory Variables Estimate Standard Error Fare * σ (Nesting Parameter) * Online * ATI * Service Type ATI carve-out Codeshare * Interline * Mileage on the Route * Coupons * Constant Term * Data Route-Carrier-Service Type Time Period 3rd Qtr, All Years Year Fixed Effects Yes Carrier Fixed Effects Yes Route Fixed Effects Yes Notes: Robust standard errors computed * indicates statistical significance at the 5% level C. Carving-out from antitrust immunity the four oneworld overlap routes with three or fewer carriers would impose harm on airline consumers equivalent to at least $55 million dollars annually. 35. The results presented here demonstrate that, relative to full ATI service, passengers on otherwise ATI itineraries that make use of carved-out segments are harmed both due to higher fares and lower service-quality. 32 To measure the potential magnitude of these harms, we calculate the annual harm that would be incurred by those passengers flying connecting itineraries making use of one of the four transatlantic oneworld overlaps with three or fewer competitors (the types of routes on which DOJ focused in discussing the Star alliance) if 32 We have also run the fare analysis at the route level instead of at the level of individual itineraries. Our results were consistent with the evidence presented above, that carving out segments from antitrust immunity leads to higher fares, relative to full ATI, on average among all itineraries on the route, although the difference fell slightly below the standard for statistical significance. 22

51 these segments were to be carved out from antitrust immunity. 33 In particular, we apply the 2.7 percent difference in fares between full ATI and ATI with carve-out service reported in Table 8 and the $67 value to an average passenger of the difference in quality reported in Table 9 to oneworld passengers who used BOS-LHR, DFW-LHR, MIA-LHR, or MAD-MIA as part of a connecting itinerary in As seen below, the potential total annual harm that carve-outs on these segments would impose on oneworld passengers flowing over them amounts to more than $55 million. Table 10: Harms from Carve-Outs on Connecting Itineraries Segment Avg. Fare Oneworld Flow Passengers Harm from Higher Fares Harm from Lower Quality Total Harm BOS-LHR $ ,473 $3,083,003 $12,451,974 $15,534,976 DFW-LHR $ ,237 $453,897 $1,425,774 $1,879,671 MIA-LHR $ ,624 $4,277,666 $17,295,926 $21,573,593 MAD-MIA $ ,828 $3,898,372 $12,475,807 $16,374,179 Total 650,162 $11,712,938 $43,649,481 $55,362,419 Notes Average coach-class fares computed for "behind" itineraries flowing over route for 2008 Passengers computed from MIDT data and include oneworld flow passengers in 2008 VI. Conclusion 36. Our analysis demonstrates that the results presented by DOJ do not provide a reliable source of support or direction for government policy toward alliances. DOJ did not actually address the relevant question the effect of ATI on non-stop fares. In addition, its results contradict the existing literature and are not robust to the inclusion of more data and additional controls because the econometric analyses do not employ the best available methodologies. When the relevant question is analyzed and better methodologies employed, the DOJ conclusions are essentially nullified and reversed. 37. In contrast to DOJ s conclusions, our results provide strong evidence that granting ATI is beneficial to consumers and that carving routes out from immunity is harmful to consumers. In particular, there is no evidence that ATI leads to higher fares on non-stop overlap routes. 33 To make this calculation, we apply the costs from carve-out presented in Tables 8 and 9 to passengers who used these segments as part of connecting travel during While it is possible that after immunity (and potential carve-outs), traffic patterns could change, we use the historical data as a simple summary of the potential magnitude of harms. 34 We compute the weighted average coach class fare for all itineraries flowing over each of the four segments using data from the DB1B. Since the DB1B does not contain a full count of passengers in particular, it lacks data on many beyond routes we obtain passenger counts from the MIDT, which reports data separately for flow traffic. The MIDT data do not include fares. 23

52 Therefore, there is also no evidence that carving out non-stop segments would lead to lower fares on those routes. In contrast, however, there is evidence that carving segments out from antitrust immunity substantially mitigates the significant reduction in fares and enhancement in service quality that is otherwise associated with ATI (rather than code-share or interline) service. 24

53 EXHIBIT 2

54 BEFORE THE DEPARTMENT OF TRANSPORTATION WASHINGTON, D.C Joint Application of: : : AMERICAN AIRLINES, INC. : BRITISH AIRWAYS PLC : FINNAIR OYJ : IBERIA LÍNEAS AÉREAS DE ESPAÑA, S.A. : ROYAL JORDANIAN AIRLINES : : under 49 U.S.C and for : approval of and antitrust immunity for : alliance agreements : DOT-OST [PUBLIC] DECLARATION OF DON CASEY 1. My name is Don Casey, and I am currently Vice President for Revenue Management at American Airlines ( AA ). I have been in the airline industry for 25 years and have held a number of positions at AA, including director level jobs in AA s International Planning and Capacity Planning departments. Based on this experience, I have personal knowledge as to how AA makes capacity decisions and how AA prices and yield manages that capacity after it is in the market. I also have personal knowledge of the recent history of existing and proposed cooperation between the parties, AA, British Airways ( BA ), and Iberia ( IB ), and the terms of the Joint Business Agreement ( JBA ). 2. I have reviewed the comments submitted by the U.S. Department of Justice ( DOJ ) in the recently completed proceeding granting Continental Airlines request for immunity within the Star Alliance ( Star II, DOT-OST ). I have also reviewed the final order issued by the Department in Star II. In its comments, DOJ argued that any immunity order should carve out coordination on service provided to local passengers on certain concentrated routes, such as New York to Copenhagen. DOJ contended that

55 - 2 fares paid by local passengers would be expected to increase on routes where immunity would reduce the number of non-stop competitors from 3-to-2 or from 2-to-1. DOJ further argued that carriers operating under an immunized relationship tend to charge connecting passengers higher fares than carriers operating under a non-immunized codeshare relationship. 3. I understand that, at the request of the Joint Applicants, economists from Compass/Lexecon have analyzed DOJ s findings and have conducted their own empirical studies, the results of which rebut many of DOJ s conclusions. Among other findings, Compass/Lexecon concluded that a reduction in the number of non-stop competitors on a route caused by a grant of immunity does not have the same fare effects as a competitor s outright exit from a route. As I explain below, that conclusion is reasonable to me since carriers in an immunized relationship tend to increase capacity on the routes that connect their networks, while the exit of a competitor from a route reduces capacity. Compass/Lexecon also found, contrary to DOJ s conclusions, that immunized alliances tend to offer fares that are lower than non-immunized alliances as well as offer a product that is preferred by consumers. These findings are also not surprising to me and indeed are consistent with my understanding of how airlines make capacity and pricing decisions. In this declaration, I further intend to explain how carve outs within the context of the proposed JBA would create inefficiencies, distort competition, and discourage the parties from further deepening and extending their cooperation in both new and existing markets and in ways that would benefit both local and connecting passengers. PUBLIC VERSION

56 In summary, as I explain in detail below: (i) on a transatlantic hub to hub route, such as DFW to London, immunizing the proposed JBA would not lead to higher fares for local passengers, which is consistent with the Compass/Lexecon results; (ii) the lack of an agreement to share all transatlantic revenue has prevented the applicants from creating additional consumer benefits; (iii) in November 2004, AA entered into a revenue sharing agreement with Swiss Airlines on a single route, JFK to Zurich, which led to improved scheduling options, more connecting opportunities, and more passengers carried. I expect that the proposed JBA among AA, BA, and IB would create these same efficiencies, but on a much larger scale; (iv) carving out local passengers from shared revenue would greatly diminish, if not destroy, the efficiencies inherent in the proposed JBA. In the context of the JBA, carve outs would not preserve existing competition on the local routes, but would instead introduce perverse incentives that would undermine the very cooperation that is necessary for the joint venture to succeed. Carve outs would also make it even more unlikely that AA and BA would agree to create new consumer benefits, such as improving schedule options, permitting fare combinability, or allowing FFP earn-and-burn on the overlap routes. In the end, I believe that carve outs would result in harming all passengers, including the local passengers that they are intended to protect. PUBLIC VERSION

57 - 4 The Proposed JBA Should Not Lead To Higher Prices For Local Passengers 5. A few important characteristics of the airline industry explain why a carrier can have a large share of local O&D passengers but still lack pricing power. First and foremost, pricing and capacity are directly related. A competitor that earns a high share of local passengers because it creates a large supply of seats is no threat to consumers or competition. An increase in supply will always reduce prices especially in the airline industry where unsold inventory spoils once the aircraft leaves the gate. Unlike other industries, airlines cannot store excess supply into inventory and wait for improved market conditions. They must take the best available price for each seat or forego all revenue from those unsold seats. Thus, market share alone does not create pricing power. Rather, market share must be accompanied by conditions and incentives that would encourage airlines to contract supply. Because network carriers make capacity decisions based on network considerations and not simply on the conditions on the local route the incentives are to increase capacity on the trunk routes that feed the broader network, even if the airline has already achieved a very high share of local passengers. 6. Any changes to flying will not only affect service for the nonstop passengers but will also impact service for passengers traveling on thousands of connecting itineraries. In AA s case, flights between DFW and London are not only a means to serve local passengers, they are inputs to service on thousands of less densely traveled city-pairs that can only be economically served by us on a connecting basis. The profitability of any individual flight depends on conditions in all of the city-pairs that are PUBLIC VERSION

58 - 5 affected by that additional flying. Indeed, in some instances, AA will operate a route that, standing alone, is not profitable but otherwise helps AA generate revenue in other parts of its network. Said differently, AA asks one key question does it expect that a new flight will generate enough revenue within its network and from a mix of local and connecting passengers to cover its costs and earn a reasonable profit. 7. The airline industry is also characterized by capacity decisions that are made months in advance and are inherently lumpy. Said another way, capacity changes come in increments of a large number of seats, particularly on long-haul international routes that use larger aircraft. A network carrier operating transatlantic service cannot fine-tune capacity additions, but instead must make decisions on how to deploy wide-body aircraft that have from 200 to 400 seats. Consider, for example, Dallas/Fort Worth to London, where AA and BA operate 27 weekly frequencies with far more seats than local passengers. Thus, the majority of passengers on one of these flights purchase a connecting itinerary. If AA were to constrict capacity in order to gain pricing power against local passengers, it would simultaneously diminish the quality of its product and the inventory that it has available to sell to the far more numerous connecting passengers. A network carrier cannot have it both ways if it reduces flying on a route, it must withdraw large chunks of capacity that serve not only as inventory for local passengers but also as trunk capacity for connecting passengers on PUBLIC VERSION

59 - 6 many more routes. Given that margins in the airline industry are notoriously thin (and all too often negative), the loss of even one or two passengers can determine whether a flight generates a loss or profit. Shrinking critical trunk capacity in order to earn supra-competitive fares from a relatively small subset of passengers is simply not a viable longterm strategy for network carriers. A carrier like AA must offer service that is attractive to both local and connecting passengers. 8. History in this business has shown that network carriers tend to add much more capacity on hub to hub routes than is necessary to serve the local market. Similarly, as illustrated in the Joint Applicants Supplemental Comments, transatlantic alliances have consistently increased the capacity that connects their U.S. and EU hubs. Importantly, the KLM/Northwest relationship which is the only truly metal neutral joint venture currently in the market offers more non-stop seats per local passenger than any other alliance relationship. In the current proceeding, the Joint Applicants are seeking to replicate this experience and will have the same incentives to maximize the capacity on their primary trunk routes, such as DFW to LHR or MIA to LHR and MAD. 9. Because market conditions vary among the many different city pairs that can be served by seats offered on any given flight, AA prices and yield manages its seats based on individual O&Ds. [REDACTED] Instead, we seek to earn the most revenue from each flight by selling seats to the most profitable mix of local and connecting passengers. Thus, the yield that AA expects to earn from selling to connecting passengers directly affects how it prices and sells to the local passenger. If we were to close inventory PUBLIC VERSION

60 - 7 to local traffic hoping to drive up the price paid by that local traffic - - those seats would either spoil or have to be sold to lower yielding connecting passengers. Thus, any attempt to raise local fares could have a negative impact on overall profitability from that flying which is what we measure and what really matters. For example, if the highest yielding nonstop passenger would contribute $800 in revenue, but the highest yielding connecting passenger would contribute only $700, we will sell the seat to the local passenger. If AA withheld that seat from non-stop passengers it would be foregoing the additional $100 in incremental revenue that a local passenger would be willing to pay for that same seat. 10. Local traffic is usually higher yielding than connecting traffic regardless of the number of non-stop competitors serving the route. This difference in yield is explained by a combination of product superiority, internal revenue proration practices, and the circuity of connecting routings. Non-stop carriers offer a product that is superior and in greater demand than a connecting product serving the same O&D. Thus, it is not surprising that AA can sell more of its higher yielding fares on the local route than it can sell with a connecting product. Indeed, many of AA s connecting passengers are traveling on itineraries where one or more of its competitors offer non-stop service which means that, in those instances, AA will capture fewer higher yielding passengers. Importantly, as additional capacity is added to capture more overall traffic, more seats must be sold and the difference in yield between local and connecting traffic encourages us to sell more seats to the local passengers. AA consistently favors the traffic that creates the best return for its overall network, without regard to whether those passengers are local or connecting. Thus, passengers of all types benefit from increases in capacity. PUBLIC VERSION

61 In sum, network considerations, not conditions in the market for non-stop passengers, dictate how network carriers deploy their resources. A network carrier cannot grow by constraining capacity on the trunk routes that create its very best connecting opportunities. The capacity that creates those connecting opportunities comes in the form of large blocks of seats available for sale to both local and connecting passengers. Indeed, as a carrier adds more flights to a route it can capitalize on economies of density to lower its costs and become an even less expensive and more efficient competitor. The Lack of a Metal Neutral Joint Venture Has Prevented AA and BA from Further Improving oneworld s Transatlantic Product Offerings 12. The lack of antitrust immunity has placed oneworld at a clear competitive disadvantage as compared to the other much larger and widely immunized international alliances. oneworld lags far behind the other alliances both in size and in the degree of integration. AA and BA have opened up less of their behind and beyond networks, have not allowed combinable fares, have not offered consolidated corporate and agency contracts, and are the only alliance that does not offer reciprocal FFP benefits on all transatlantic flying. 13. AA and BA both recognize that they could offer a more competitive product by deepening and expanding their relationship. Any such effort would have two effects. On the one hand, it would make oneworld a more effective competitor against the larger Star and SkyTeam alliances. On the other hand, however, enhanced cooperation would lead to a transfer of value between the two parties. Antitrust concerns and the lack of revenue sharing have prevented the parties from exchanging the commercially sensitive information that would be necessary to identify, measure, and remedy these disproportionate impacts. PUBLIC VERSION

62 In my prior capacity as a Director in AA s international planning department from October 2004 to August 2009, I had numerous discussions with BA about finding ways to improve the oneworld product. The inability to share commercially sensitive information and the lack of a mechanism to share proportionately in the value created repeatedly led the parties to abandon initiatives to improve our joint product. For example, AA and BA know that the members of our frequent flyer programs are frustrated and confused by inconsistent FFP benefits. However, closing the FFP gap would alter how consumers choose between AA and BA services, and neither AA nor BA is willing to close our FFP gap absent a joint venture that evenly distributes the effects of any such change. 15. In yet another example, the parties have not implemented any new codesharing for a number of years. In prior discussions, AA sought to add other destinations beyond London, such as BA s India flights, but BA resisted those additions because it feared this would divert traffic to AA s transatlantic flights that otherwise would have flown on BA s transatlantic flights. In turn, AA has withheld some important U.S. domestic destinations, such as New Orleans, from BA s codeshare network. Even when the parties tried to identify new opportunities that minimized the risk of diversion, efforts to expand the scope of cooperation have faltered. AA considered putting its code on BA s Philadelphia to London flight, but absent input into AA s corporate pricing policies, BA was unwilling to expose its inventory on these flights to AA corporate deals. PUBLIC VERSION

63 The parties have also discussed fare combinability, but those discussions never progressed for many of the same reasons. Efforts to improve our corporate products have likewise proven futile without a revenue sharing agreement for all transatlantic passengers. In 2005 and 2006, the parties attempted to develop a new joint product for corporations in the Benelux countries. However, several factors, including the lack of any integration between the sales forces and, lacking immunity, an inability to discuss and disclose corporate selling strategies, rendered these efforts largely useless. 17. AA and BA have been in the oneworld alliance since 1999, and despite our best efforts we have been unable to implement changes that we know would improve our alliance product. Absent the grant of antitrust immunity and the implementation of the JBA, I do not believe that AA and BA will substantially improve the oneworld product offering. The JBA will not succeed without transparency and a predictable balance of benefits which are achievable only with the same immunity already provided to the other two alliances. 18. oneworld s inability to offer an alliance product that is comparable to Star and SkyTeam is reflected in a study conducted for the parties by RBB. I do not intend to repeat the conclusions from that study other than to note its principal finding that the other alliances have made meaningful in-roads in competing in oneworld s home markets. oneworld, in contrast, has not achieved similar results in Star s and SkyTeam s home PUBLIC VERSION

64 - 11 markets. As further evidenced by data the parties supplied in our application, there are now thousands of transatlantic city pairs, flown by millions of passengers, where Star and/or SkyTeam have shares in excess of 80%. Under these circumstances, I believe that the failure of oneworld to become a comparably integrated alliance would be a huge loss for transatlantic aviation. The Swiss Example of a Metal Neutral Joint Venture 19. In 2004, AA and Swiss enjoyed an immunized relationship but were operating competing flights between JFK and Zurich. Both carriers operated a single flight a day which departed within an hour of each other. This scheduling was the preferred time of departure for local passengers. 20. Both parties recognized that they could improve their joint product by spreading the flights out to create more scheduling options for the local passengers and more destinations and opportunities for the connecting passengers. However, moving one of the flights from the time preferred by the local passenger would make that flight less profitable. Thus, absent revenue sharing for all passenger types, a unilateral change would have rendered one carrier better off and the other carrier worse off and thus would never have been made. 21. Having immunity, the parties entered into a revenue sharing agreement for that route, and AA retimed its JFK to Zurich flight to arrive three and one half hours later from a 7 am to a 10:30 am arrival. This retimed flight gave AA s and Swiss s passengers access to more schedule options and a broader range of connecting opportunities. After the change, AA and Swiss carried more passengers, load factors and RASM increased, and importantly overall local traffic increased with revenue sharing. Exhibit A graphically illustrates this effect. The left side of page 1 shows the PUBLIC VERSION

65 - 12 connecting opportunities that were available for the passengers traveling from JFK beyond Zurich before the change. The right side shows the increase in connecting opportunities for these same passengers after the change. The second page shows the connecting opportunities at Zurich going in the other direction both before and after the change. I believe this experience is typical in the airline industry and that the as-yet untapped benefits from optimizing the combined networks of the JBA carriers would be exponentially larger than the benefits from moving this single AA flight between JFK and Zurich. Carve Outs Would Destroy Metal Neutrality, Discourage Cooperation, And Harm Consumers 22. Carving out revenue and cooperation on certain of the JBA overlap routes would retain and even amplify the barriers that have prevented the parties from deepening and expanding their cooperation. Carve outs would eliminate the transparency and ability to balance benefits that are necessary to create the truly integrated and efficient joint venture envisioned in the JBA. 23. Carve outs would not merely preserve the existing competition between the parties for local passengers, but would instead create perverse new incentives that would greatly diminish cooperation. The Compass/Lexecon results show that carve outs have not only been ineffective, but there is evidence that they have led to higher fares. For reasons explained below, I believe these results reflect the difficulties that aligned carriers are likely to encounter in agreeing on capacity decisions and coordinating traffic over carved out routes. With carve outs in effect, carriers are less likely to agree on when and how to expand capacity. Before turning to those points, however, I want to explain why carve outs would be even more pernicious in the context of the proposed JBA here. PUBLIC VERSION

66 Today, in the absence of the JBA, AA and BA share no revenue on any of their transatlantic flying. If AA loses a local passenger to BA, BA gets 100% of that revenue. AA will seek to sell its unsold seat to the next best available passenger either local or connect and it retains 100% of that revenue. In many instances, AA will successfully recapture a large percentage ( [REDACTED] ) of the revenue lost to BA for that local passenger by carrying a connecting passenger. Implementation of the JBA with carve outs for local passengers would dramatically alter this competition in a perverse and inefficient manner. Under these circumstances, when AA loses a local passenger to BA, BA continues to keep 100% of the revenue from the local passenger, but AA will no longer be able to recapture a large portion of this lost revenue by carrying a connecting passenger. Instead, under the anticipated JBA revenue sharing formula, AA would only be able to retain []% of the revenue from a connecting passenger. If the fare for the connecting passenger is only []% of the fare for the local passenger, AA would recapture only []% of the []%, or []%, of the revenue lost from BA s sale to the local passenger. This disparate treatment of local and connecting revenue creates an artificial dynamic that would dim any hope that the parties would expand capacity on these carved out routes or offer new consumer benefits, such as FFP reciprocity or fare combinability. PUBLIC VERSION

67 If carve outs are imposed, new and vastly different economics for connecting versus local passengers would put tens of millions of dollars at stake between the parties. Each will try to find ways to fill their partner s aircraft with the shared connecting traffic and protect its own inventory for the now relatively much more lucrative local passengers. As explained in AA s Supplemental Comments, the parties would no longer open inventory to the local and connecting passengers based on which passenger values the seat the most, but instead those decisions would be based primarily on whether the passenger is local or connecting. This suboptimal yield management will likely depress load factors and constrain capacity in thousands of connecting markets now dominated by Star and SkyTeam. 26. The disparate treatment of local versus connecting revenue under the JBA with transatlantic carve outs would create other perverse incentives, some of which could lead to the collapse of effective cooperation even on connecting routings. For example, sum of local pricing is rarely the best alternative for connecting passengers. But in an environment in which all parties have an incentive to accommodate local traffic even if that traffic earns a yield that is merely a fraction of what a connecting passenger would bring sum of local pricing could become an attractive way to capture revenue that the parties would otherwise have to share. If, for example, BA offered a DFW based corporation an artificially low local fare (because that price is still better than splitting revenue even from higher yielding connecting passengers), that corporation would now have an incentive to use that fare not only for travel to London, but for travel to other points in Europe by also buying a beyond ticket from London. In this scenario, none of the revenue is captured by the joint venture. Of course, since the parties could not share information PUBLIC VERSION

68 - 15 on local pricing, there is no way for the parties to measure, understand, or balance the inequities within the joint venture caused by these distorted incentives. Such distortions would ultimately lead the parties to limit cooperation, make suboptimal capacity and yield management decisions, and in the worst case could undermine the long term viability of the joint venture. 27. In addition to these severe distortions, carve outs impose other inefficiencies, both in the form of diminished incentives to cooperate and burdensome administration costs. For example, schedule optimization like what AA undertook on the single JFK-Zurich route would not occur on any of the carved out overlaps. Neither party would willingly sacrifice its position by moving to a time less attractive for local passengers, especially given the amplified importance of local passengers in a carved out JBA. The result would be continued inefficient wingtip flying, fewer schedule options for local passengers, and fewer connections with longer wait times for connecting passengers. 28. Similarly, with local revenue carved out, it would be more difficult for the parties to agree on whether capacity should be added, and if so by which carrier. As explained earlier, capacity decisions are network based decisions that need to account for all passenger types. Adding a new frequency to Dallas/Ft. Worth, for example, would have implications for the yield and mix on all pre-existing frequencies on that route, as well as other routes where the parties are cooperating. The timing of that frequency will determine whether it is likely to attract local traffic from existing flights or will largely serve connecting passengers, some of whom could have been served by other routings. These questions will have enhanced implications for the parties. Once these types of concerns drive the parties to have different agendas, it will be PUBLIC VERSION

69 - 16 difficult to achieve consensus on the question of whether a carved out route could profitably sustain the addition of another 250 daily seats, and if so, whether the effects would be shared equitably within the joint venture. Thus, I believe that carve outs would lead to less capacity on these routes. I would again note that the metal neutral joint venture between NW and KLM created the greatest capacity additions relative to the size of the local market. It is my opinion that those capacity additions were a direct result of KLM and NW being free to share all relevant information and revenue in order to make the most informed and efficient capacity decisions. 29. Carve outs would also continue to discourage the parties from implementing other consumer benefits for local passengers. As mentioned, AA and BA have not been able to agree on reciprocal FFP benefits on overlap routes, nor have they allowed fare combinability. But, the potential for an inequitable transfer of value between the parties would be even greater in a carved out JBA than it is today. Thus, if key routes are carved out, I believe the possibility that the parties would offer these benefits is remote, at best. 30. Similarly, under the JBA the parties had hoped to begin offering comprehensive and easy to administer corporate contracts. Corporate customers have made it clear to us that the oneworld product is uncompetitive with the more integrated contracts offered by Star and SkyTeam. Carving out a few routes would create more complexity and administrative burden for both the parties and their customers, without any meaningful benefit. PUBLIC VERSION

70 Finally, given the level of integration that the parties had hoped to achieve under the JBA, carve outs would introduce substantial administrative costs and needlessly expose the parties to meritless antitrust claims. The parties are putting large portions of their businesses at stake in this joint venture, and we are convinced that deep and meaningful integration across the companies is necessary for its success. Thus, the parties are not contemplating a few touch points between the separate companies, but rather robust sharing of information and data bases involving large numbers of employees. Carving out a few markets would require each airline to duplicate often in real time -- how it maintains, collects, and uses information one for its own internal purposes, and another which is largely identical, except for information on a few carved out routes. 32. For one example, the parties intend to place revenue and pricing personnel in each other s pricing and yield management departments. We envision that those employees would have access to the same information as their AA or BA counterparts. With carve outs, however, the companies would have to keep and maintain separate data bases to eliminate access to information on carved out routes. 33. Moreover, even if the parties endured this unnecessary burden and expense, carve outs could expose the parties and their employees to the risk of meritless, but costly, litigation. If, for example, the parties were to discuss implementing a new surcharge on U.S. to EU traffic, those discussions could not include the carved out routes. Yet the same economics that led the parties to agree that such a surcharge made sense in all the other U.S. to EU routes would likely also apply to the carve out routes, and it would not be unusual for the parties to unilaterally reach the same PUBLIC VERSION

71 - 18 conclusion on the carved out routes. But once this entirely common and predictable parallel pricing occurs on the carved out routes, the parties could be subject to a claim that any such decision was the product of an agreement. The thousands of touch points between the companies, and the discussions and agreements that occur as to non-carve out routes, become easy targets as plaintiffs attempt to construct a case from wholly innocent and, indeed, pro-competitive conduct. 34. It is certainly true that AA has successfully coped with carve outs in other immunity orders. This proposed JBA, however, is materially different from those prior experiences both in terms of scope and integration. AA has never implemented a metal neutral joint venture with full revenue sharing. For the reasons I have already explained, in this new context, carve outs are particularly damaging to the efficiencies and coordination contemplated in the JBA. In sum, I believe that carve outs would lead to less capacity growth, less efficient decision making, more costs, and fewer consumer benefits. The Department s most recent Star order properly distinguished between cooperating on a route that is part of a metal neutral joint venture and cooperation in the absence of any such agreement. In this matter, equity and efficiency compel the same result. A decision that saddles oneworld the smallest alliance by every relevant measure with government imposed inefficiencies on the very routes that are most important to feeding the parties combined networks would not only be unfair, but would be a severe disservice to transatlantic consumers. September 8, 2009 DON CASEY PUBLIC VERSION

72 JFK ZRH to Beyond Connection Windows Before and After AA / LX Revenue Pooling Arrangement October 2004 November :00 AA Arv 06:00 NUE LCY 07:00 08:00 LX Arv NUE 07:00 08:00 LX Arv LUX GVA MAD, MXP DUB, STR 09:00 10:00 11:00 12:00 13:00 14:00 AA Connection Windo ow LX Connect tion Window AA A / LX Connection n Window LUX GVA CPH, MAN, STR ATH, BUD, MAD, OTP CAI, DME, IST, NBO, SSG, WAW BEG, BRU, TLV FCO, GVA, NCE BOM, CDG, DXB, VIE MXP LCY 09:00 10:00 11:00 12:00 13:00 14:00 AA Arv AA A Connection Window LX Connect tion Window AA / LX Connectio on Window CAI, LIS, NBO, PMI, SSG ATH, DME, OTP,TLVTLV BRU BOM, CDG, GVA NCE, VIE LCY GVA BHX, CPH, MAN BEG, MXP, WAW HAJ, LHR BCN, DUS, FCO, FRA AMS, HAM, STR BRU, IST, MUC, TXL CDG, PRG DXB, NUE, RUH, VIE AGP GVA NRT

73 Beyond to ZRH JFK Connection Windows Before and After AA / LX Revenue Pooling Arrangement 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 AA Connection Window AA Dep LX Connection Window LX Dep October 2004 November 2004 AA / LX Co onnection Windo ow NUE DUS, FRA GVA, MXP, VIE BRU HAJ, HAM, PRG ATH, CDG, DME, FCO MUC AMS, NCE, TLV, WAW CPH, STR LHR LHR MXP GVA CDG, LCY DUS, MUC FRA, TXL AMS, VIE FCO, HAJ, HAM, LHR BCN, BRU, NUE 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 AA Connect tion Window AA Dep LX Conne ection Window LX Dep AA / LX Conn nection Window HKG, JNB, TIP BKK, SSG BOM, DXB, RUH, SKG NBO BSL, STR GVA FRA GVA MXP NUE HAM GVA, VIE DUS BRU, NCE ATH, CDG, DME, FCO MUC AMS, LHR, TLV STR MXP HAJ, MUC DUS, GVA, PRG LCY AMS, BUD, FRA, TXL HAM, VIE BHX, BRU, CDG, FCO, NUE BCN CPH, LHR, LUX, STR, WAW MAN GVA MXP

74 EXHIBIT 3

75 INTERNATIONAL AIRLINE ALLIANCE DEVELOPMENT Network Airlines' Mainline Domestic and Atlantic ASMs Expressed as Index Numbers (March 31, 1994 = 100) ASMs Expressed as Index Numbers (March 31, 1994 = 100) Network Airlines' Mainline Atlantic ASMs Network Airlines' Mainline Domestic ASMs 0 M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D Months Ended Necessary for Network Airlines and Consumers 8,000 Operating Profit or Loss ($millions) Network Airlines, Domestic Operations 6,000 4,000 Operating Profit or Loss ($millions) 2, ,000-4,000-6,000-8,000-10,000-12,000 MJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJ SDMJSDMJSDMJSDMJSDMJSDMJS Months Ended By Randy Bennett Patrick Murphy Jack Schmidt July 2009 Gerchick-Murphy Associates LLC

76 2 Table of Contents Description Page Executive Summary: Introduction.. 3 Network Airlines Domestic Plight 4 International Expansion and the Importance of Alliances 5 The Need for Antitrust Immunity.. 6 Antitrust Immunity Creates Substantial Consumer Benefits 8 Conclusion... 8 Introduction 10 Section I Network Airlines Domestic Problems Continue: Overview.. 13 Revenue Effects of Low Cost Airline Expansion The Cost Gap With Low Cost Airlines. 23 Section II Alliance Integration: Overview.. 26 The Role of Networks in International Service Antitrust Immunity ABOUT THE AUTHORS AND THE STUDY: Randy Bennett, Patrick Murphy and Jack Schmidt are former U.S. government officials who collectively provided more than 100 years of economic analysis to the U.S. Department of Transportation, and previously to the U.S. Civil Aeronautics Board. During their extensive tenures, and since leaving government, the authors have been closely involved in the major economic policy issues affecting the U.S. airline industry both before and after airline deregulation in Gerchick-Murphy Associates assisted in the management of the project. The analyses and conclusions are those of the authors. The authors used DOT techniques and publicly-available data. Gerchick-Murphy Associates LLC

77 3 Executive Summary Introduction INTERNATIONAL AIRLINE ALLIANCE DEVELOPMENT Necessary for Network Airlines and Consumers EXECUTIVE SUMMARY For the past three decades the heart of the U.S. network airlines has been the domestic hub-and-spoke systems that developed soon after deregulation in For the future, however, it is clear that any lasting and significant improvement in network airline financial health and consumer benefits will largely be the result of continued international expansion made possible by the development of immunized international alliances. Immunized alliances provide airlines the necessary economic incentives to integrate their business processes in order to optimize their alliance networks. This integration stimulates traffic growth through improved products, services and prices to consumers. The network airline sector of the U.S. airline industry is at a crossroads. The network airlines position of dominance in the domestic market that enabled years of highly profitable growth during the 1990 s is gone and will not reemerge. Instead, network airlines are in the process of a fundamental change in their overall operating structures. Domestic operations are sharply declining while international operations are sharply increasing. The differences between the network airlines domestic and international financial performance and growth trends are stark. Their ongoing efforts to rebalance the relative sizes of the domestic and international components of their systems are neither modest nor temporary; rather, this process is necessarily accelerating. Between 2000 and 2008, network airlines domestic mainline ASM capacity was reduced by an unprecedented 28 percent, and this implosion will exceed one third once the latest cuts are in place for Even the capacity of the regional affiliate feeder carriers, which had increased since 2000 as network airlines had used them to replace mainline operations, began to decline starting in mid Regional airline ASM capacity for the fourth quarter of 2008 was down by 10 percent over the previous year. Since the mid-1990 s, when international alliance development began to accelerate, the network airlines have steadily and profitably grown their international operations, except for a brief period following the terrorist attacks of September 11, which was also compounded by SARS and the initial stages of the Iraq war. Between 2003 and 2008, when domestic capacity cuts not only continued but sharply accelerated, network airline international ASM growth Gerchick-Murphy Associates LLC

78 4 Executive Summary increased by 41 percent. This international growth was highly dependent upon alliance development. Network airlines had already had two decades of independently extending the reach of their domestic networks, mostly by serving larger foreign destinations as spoke points from their domestic hubs. Alliance development enabled carriers to extend further the reach of their networks to thousands of new destinations. The purpose of this study is to examine the reasons why continued immunized international alliance development is necessary for the large U.S. network airlines, as well as the consumers and communities they serve. For almost two decades international alliances have played an important role in the U.S. network airlines ability to successfully expand and compete abroad. In recent years, however, the economic significance of alliances for airlines and the competitive significance for consumers have markedly, and suddenly, changed. International alliances were initially used as a tool for extending the reach of domestic networks. Now they have become essential for the economic survival of the network airline industry. Network Airlines Domestic Plight: During this decade the U.S. domestic airline industry has gone through the most dramatic change in its competitive structure since deregulation more than 30 years ago. A combination of circumstances has favored the low cost business model, and as a result the large network airlines have been replaced by the low cost airlines as the competitively dominant domestic airline sector. The low cost airlines, not the large network airlines, now drive change in both price and capacity growth. Since 2000, while network airline mainline domestic capacity has been reduced by nearly one third, low cost airline capacity has more than doubled. This low cost airline expansion is particularly problematic for the network airlines because of a large gap in cost structures that now exceeds 50 percent, and has been increasing for the past two years. Almost 80 percent of network airline domestic passengers, and more than 70 percent of their domestic revenues, are now in markets with a low cost competitor. These percentages have surged since 2000, when they were 55 percent and 44 percent, respectively. Because the low cost airlines can profitably serve the large price-sensitive demand sector, the low cost business model, unlike the network airline business model, does not rely upon extreme market segmentation and charging significantly higher prices for time sensitive demand. As a result, where low cost airlines compete, their presence significantly disrupts the ability of network airlines to charge higher prices as a means of compensating for their higher costs of operation. The estimated impact on network airlines is a loss of $11.2 billion annually, compared with annual passenger revenue of less than $45 billion. Gerchick-Murphy Associates LLC

79 5 Executive Summary The conclusion at the outset that the network airlines will not return to positions of dominance in the domestic market is largely based on the fact that all of the tendencies that have put the network airlines in competitive jeopardy in the domestic market are, from the network airline perspective, continuing to worsen. The cost gap is continuing to increase, the low cost airline market share continues to increase, and high end demand continues to gradually decline. In some major markets, low cost airlines have recently begun to compete with each other. In these circumstances, network carriers have been forced to abandon markets they have served for decades. There is nothing to suggest that the domestic competitive environment is going to become more favorable for the network airlines. The effect of all this on network airlines is perhaps best revealed by looking at their financial results. Between 2000 and the third quarter of 2008, they earned a domestic profit in just six of the 31 quarters, and suffered operating losses of $30 billion. As would be expected, the network airlines have acted aggressively to shore up their domestic competitive positions: they have achieved large reductions in their cost structures, in some cases through the bankruptcy process; they have handed off a substantial proportion of their domestic operations to regional partners; and they have reduced their own capacity both to eliminate financially stressed operations and to reduce capacity available for price-sensitive traffic in an effort to achieve higher average fares. But the cumulative effects of these and other steps have not been enough for the network airlines to reestablish domestic profits. International Expansion and the Importance of Alliances As the network airlines began to lose the domestic battle to low cost airlines, they had to weigh two alternative responses. One was sweeping domestic consolidation, which would have reduced the number of networks that are competing for the same, ever shrinking, domestic high-fare demand. The other was international cooperation, in the form of alliances, as a means of expanding the reach of their networks. This increases revenue flows to support domestic networks, thereby increasing system viability. If done on a broad scale, domestic consolidation would have reduced network capacity at home and around the globe. Conversely, international cooperation via alliances expanded network reach and, subsequently, increased international network capacity while minimizing the need for domestic capacity reductions. While some domestic consolidation may still have to be pursued, the successful development of international alliances has curbed the perceived need for consolidation at home, and has enabled the network airlines to remain strong global competitors despite their domestic problems. Gerchick-Murphy Associates LLC

80 6 Executive Summary International expansion via alliance development has been positive for all airline constituencies and is consistent with consumer needs, airline desires to fill those needs, and good public policy. The international growth strategy has provided consumers with more capacity, more choice, and more competition. Small communities in the U.S. that rely heavily on network airlines have seen their service strengthened, and employment has benefited from strong international growth. The service benefits of international alliances are especially apparent in longerdistance intercontinental markets, where the vast majority of city-pair markets can only be served on a connecting basis. Indeed, only a tiny fraction of either city-pair markets or cities can sustain nonstop intercontinental service. Passengers currently travel in more than 40,000 city-pair markets across the Atlantic, yet only 220 of those markets are served nonstop, and only one third of those markets receive more than a daily flight. The connecting markets, and virtually all city-pair markets that now support multiple frequencies, can only be served effectively and efficiently by network systems. But most individual airlines lack sufficient networks to provide such service. Restrictive cabotage and ownership laws prevent U.S. airlines from using economic tools that could improve their ability to serve international markets. U.S. airlines cannot operate inside foreign countries because of cabotage restrictions. They cannot set up subsidiaries in the EU to operate networks of feeder flights to support their inter-continental US-EU services. US carriers cannot invest in existing EU airlines and gain control over them because of ownership limitations. Nor can US carriers partner with financial institutions to set up new airlines in the EU unless they are prepared to cede control of the venture to the financial partners. Apart from legal and infrastructure constraints, no airline can efficiently provide service with its own aircraft and crew to most of the international destinations its customers require. As a consequence, the only feasible way to offer quality service to the vast majority of markets around the globe is to create alliances by linking networks already in place on the various continents. The Need for Antitrust Immunity Although they have been successful in creating online-like connecting services that have significantly stimulated demand, alliances are, at best, imperfect business models. As is true for independent airlines, in order to successfully serve customers, alliances must develop goals, along with policies and strategies to achieve those goals. The ability of alliances to successfully accomplish this, however, is made far more difficult by virtue of the fact that all decisions must be joint. Decision-making is even more severely hampered by the antitrust risks of jointly collaborating on the most essential components of an airline business Gerchick-Murphy Associates LLC

81 model such as pricing, scheduling, revenue management, and capacity decisions. 7 Executive Summary DOT s antitrust authority provides a means whereby airlines can overcome legal limitations in order to implement international alliances that demonstrably benefit consumers and enhance efficiency. In the absence of DOT s antitrust authority, the airline industry could well decide to take somewhat greater antitrust risks, but it is unlikely that they would implement the broad-based alliances that have proven to be so valuable, or find it worth the price of private antitrust lawsuits that could result. DOT has used its authority to grant antitrust immunity to develop an international alliance policy that has resulted in network growth that has benefited both consumers and airlines, not just in international markets, but in domestic markets as well. The potential risks of legislative action to change this process are clear; the potential gains of doing so are not, particularly in the absence of evidence that the grant of antitrust immunity has resulted in anticompetitive behavior by immunized alliances. Antitrust immunity is important not just to overcome the chilling effect of concerns of private antitrust challenges, but also because it sparks a dynamic growth process. Immunity enables alliance partners to create economic benefit sharing arrangements in which the alliance partners create standard business processes for key business functions that airlines utilize to distinguish themselves from their competitors: planning, scheduling, pricing, and managing capacity throughout the network. Optimizing these functions for passengers flowing through the alliance network creates significant consumer benefits such as lower prices, more service and increased convenience. Only by implementing an economic benefit sharing agreement the core of which is often a revenue sharing agreement are carriers incentivized to make network planning, selling, and managing decisions in the interest of the alliance rather than in the interest of their own individual businesses. In order to better ensure that these merger-like efficiencies are obtained and passed on to consumers, DOT has recently focused on the need for large alliances involving major network carriers to include a revenue sharing agreement which makes the carrier s marketing efforts metal neutral that is to say indifferent as to the alliance carrier on which the passenger chooses to travel. Requiring a revenue sharing agreement for large alliances seeking antitrust immunity provides additional financial incentives for alliance partners to focus on jointly competing with other market participants on a broad scale throughout the network. As the alliance succeeds in jointly creating the necessary policies and procedures for successful integration, the groundwork is developed for new services that are now viable only because both risks and rewards are shared between alliance partners who are now equally incentivized to support the new services. As these new services are integrated with the broader network Gerchick-Murphy Associates LLC

82 8 Executive Summary product, additional passengers are attracted to the alliance. This leads to further capacity increases not only on pipeline hub-to-hub routes, but also on additional intercontinental routes from alliance partners respective hubs, and in their respective homeland networks. Antitrust Immunity Creates Substantial Consumer Benefits Evidence shows that antitrust immunity is pro-consumer. Not only do consumers in thousands of connecting markets have access to better service as a result of better network integration, in many markets they enjoy competitive access to multiple alliance products. The phenomenal traffic growth during the past five years (54 percent) in the transatlantic market, which has benefited the most from ongoing alliance development, supports a conclusion that alliances are price competitive. This robust growth has been fueled by alliance development that has enabled the airlines to tap a large pool of demand that historically has been poorly served. Furthermore, an analysis of fare data shows that fare increases have been restrained by alliance development, and fares have not kept pace with fuel cost increases. Indeed, the continuing rapid traffic growth in international markets where alliances operate supports the view that alliance pricing practices are proconsumer. Furthermore, in the transatlantic market where the US EU open skies agreement opened the door in 2008 to new competition, one can reasonably expect that airline prices will be adequately protected by new competition. In non US-EU markets the answer to any pricing concerns that might arise by airlines closely cooperating in alliances is not constraining the alliance benefits that consumers appreciate, but encouraging more competition in the marketplace. Conclusion Given the drastic changes in the domestic airline sector during this decade, the fundamental question is whether the network airlines can remain viable. That success depends on the interaction of their fast growing international systems with their stressed domestic systems. The successful growth of their international operations is dependent on the revenue flows generated by their domestic networks. The international operations generate revenues that flow through and support the domestic networks. The network airlines domestic predicament that continues to result in a major retreat shows that continued successful expansion in international markets as part of the ongoing rebalancing of the domestic and international components of their system of networks is the only course of action they can take to stabilize overall viability, short of full scale domestic consolidation. Such expansion will be made possible by the development of international alliances that benefit from antitrust immunity, and will become even more important as the competitive Gerchick-Murphy Associates LLC

83 9 Executive Summary environment becomes more difficult for U.S. network airlines. Anything that clouds or inhibits the network airlines ability to participate successfully in immunized international alliances could produce very disappointing results for the carriers, their employees, the hundreds of domestic communities that low cost airlines still do not serve, and, ultimately, the country as a whole. Gerchick-Murphy Associates LLC

84 10 Introduction INTRODUCTION This study does not rely upon data for any individual airline but instead is based primarily upon data for two groups of airlines: the large network airlines and the low cost airlines. 1 The need of the network airlines for continued international growth through the increased use of immunized international alliances is the primary theme of the study, but the impact of low cost airline growth in the domestic market is what has driven the network airlines to shift their focus from domestic markets to international markets. International alliance growth is no longer just one means of stabilizing overall network strength; it is the only way to accomplish that necessity. For nearly two decades following the development of network systems of service soon after domestic deregulation, the network airlines were able to successfully pursue the natural incentives of the network business model for expansion. Although revenue-side advantages of size provide the incentive for any successful airline to grow, growth is particularly important for hub-based systems of service. This is due to the built in interdependence between the various service components of the network. Every discrete part of the network supports, and in turn is supported by, revenue flows generated by the rest of the network. Increasing network size results in increased network flows throughout the network. Network systems of service are dynamic business models. The greater reliance on network flows to support service on virtually every specific segment of service reduces the risk of failure because the network has the flexibility of tapping into a wide array of potential revenue alternatives to find the combination that best supports the service. For example, when entry by a price competitor results in the loss of local revenue, the network has the option of rerouting traffic flows to change the mix of flow revenues in order to compensate for lost local revenue. Expansion of a network system of service is thus both easier and less risky because new service does not have to depend solely (or even extensively) on local revenues, but instead is supported by revenue flows from throughout the rest of the network. This revenue support typically is substantial, and on any given segment of service the flow revenues are likely to be much greater than the local revenues. Network expansion also increases revenue flow opportunities for the rest of the network from any new point of service; from this it is obvious that a healthy network is a growing network. Conversely, the concept of shrinking to profitability is a contradiction for a network system, and efforts to do so are evidence of very difficult times. The creation of geographically broad network systems of service in the domestic market following deregulation provided the network airlines a solid base to build 1 The large network airlines are American, Continental, Delta, Northwest, United, and US Airways. The low cost airlines are ATA Airlines, AirTran, Frontier, JetBlue, Southwest, and Spirit, the airlines that DOT has identified as low cost airlines. ATA Airlines no longer operates, but did for most of the period encompassed by this study. Frontier has entered bankruptcy but continues to operate. Gerchick-Murphy Associates LLC

85 11 Introduction upon. This allowed them to extend their reach into international markets as regulatory constraints were gradually lifted, and enabled them to better access historically underserved international markets. Early on, international destinations tended to be served as spokes to the network airlines domestic networks, but in the 1990s, these carriers began to use various types of alliances with foreign airlines to expand further their international reach. By experiencing growth throughout their systems, they had seemingly endless opportunities for network expansion both at home and abroad. With a selffeeding cycle of growth that increased overall network strength, which in turn led to more growth, the success of the network model seemed assured. Indeed, for a time, this led to such dominance by the network airlines that some industry observers were concerned about the ability of other competitors to adequately discipline price in the domestic market. But immediately after 2000, this seemingly endless cycle of growth abruptly vanished. The combination of a two-year spike in operating costs that was punctuated by a massive decline in high-fare revenues that occurred virtually overnight led to a nearly six-year period of losses that accumulated to over $30 billion. Despite large capacity reductions that were undertaken in an effort to stop the financial hemorrhaging, the network airlines continued to suffer extensive losses and were forced to turn to shrinking to profitability, despite the inherent conflict of that strategy with network economics. During this period, low cost airlines were able to rapidly and profitably expand due to the continued abundance of price sensitive demand and their low cost structures. It is the widespread presence of low cost airline competition that compels the network airlines to pursue international expansion as the means of improving network revenue flows and stabilizing network strength. The network airlines seem to have done all they could to address their domestic financial problems, but they could not change the competitive landscape. Robust international growth has been the one piece of good news for the network airlines at a time of historic capacity reductions and continued losses in the domestic market. Absent their international success, it is difficult to contemplate how different the large U.S. network airline sector would be today. However, although increased international growth is a promising prospect for stabilizing their system of networks, it is totally dependent upon the airlines ability to continue to better integrate their international alliances. Immunity from the antitrust laws is necessary because it gives carriers the freedom to optimize their ability to act as one with their airline partners in tens of thousands of international markets where alliances are the only meaningful service option. The study starts with an examination of changes in the domestic airline structure and why that has been so problematic for the network airlines. The study then turns to an examination of the very different network airline results in international markets and why antitrust immunity is essential to the large network airline business model. Gerchick-Murphy Associates LLC

86 12 Section I Network Airlines Domestic Problems Continue Overview Revenue Effects of Low Cost Airline Expansion The Cost Gap with Low Cost Airlines Gerchick-Murphy Associates LLC

87 13 Network Airlines Domestic Problems Continue NETWORK AIRLINES DOMESTIC PROBLEMS CONTINUE Overview Although the primary theme of this study is the necessity of the network airlines to continue international growth through the increased use of immunized international alliances, an examination of their domestic circumstances is necessary to show that international expansion is no longer just economically sound, as has always been true, but has become an economic necessity. The network airlines financial posture in the domestic market has been seriously weakened, and cannot be stabilized by domestic operations alone due to the carriers inability to attract the premium revenues needed to compensate for their high cost structure. The key to stopping the continued decline in domestic capacity (and reestablishing profitability) for these carriers is developing the revenue flow support that results from continued international expansion. Industry pundits often link the network airlines poor financial results that have persisted throughout this decade to the crisis of the moment, and then maintain that when the crisis du jour is over, financial recovery will follow. The airline industry has indeed been confronted with a well known litany of external challenges that have affected profitability and led to inconsistent financial results. But the network airlines financial plight is more fundamental than the steady stream of foreign policy crises, fuel price shocks and recession. The network airlines have fared considerably better internationally than domestically despite the fact that both arenas have been affected by most of the same exogenous factors. Domestic losses have continued in the face of large scale capacity reductions designed to stem those losses, yet international profits have recovered and steadily increased to record levels during a period of sustained strong growth. While network airlines were suffering domestically, the low cost airline sector profitably expanded throughout this decade. Clearly, something other than the crisis of the day is producing the very different domestic results experienced by the two groups of airlines. Figure 1.1 shows network airline operating profits or losses for the past 23 years broken-down by domestic and international operations. This is illustrated on a 12 months ended basis for each calendar quarter to provide a more complete picture of trends. Operating profits, rather than net profits, are used in order to eliminate most out-of-period and major special items that serve to obscure the airlines true ongoing performance. 2 2 In a few instances, some components of reported expenses have been adjusted to remove unusual or out of period expenses in an effort to normalize comparative operating expenses. Gerchick-Murphy Associates LLC

88 14 Network Airlines Domestic Problems Continue $10,000 $7,500 Operating Profit or Loss (in $millions) $5,000 $2,500 $0 -$2,500 -$5,000 -$7,500 -$10,000 Domestic International -$12,500 -$15,000 MJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJS Months Ended Figure 1.1 Operating Profit or Loss (in $ millions), Large Network Airlines, Domestic and International Operations, 12 Months Ended Data at Each Quarter Given the extended period of massive losses in the domestic market, it is remarkable that the large network airline sector remains essentially intact. Although the focus in Figure 1.1 is drawn to the enormity of the change between the eight-year period through calendar 2000 and the following six-year period, this graph illustrates other important developments. First, the abrupt change in the financial picture began before September 11. On a year-over-year basis, the change between June 2000 and June 2001 was the largest decline in profitability ever experienced by the industry. Also, international operations returned to profitability at the beginning of 2004, more than two years earlier than domestic operations. Indeed, even as international profits continued, domestic profitability remained elusive, and again declined through early 2005, before beginning to recover and eventually returning to profits for a brief period starting in late Once both domestic and international operations returned to profitability, international profits were far larger, a very different result than during the 1990s. Domestic profits were only one-third those experienced during the 1990s, while international profits doubled. Once international operations turned profitable, they reached record levels, and then remained positive despite the fuel crisis, at least through the year ended third quarter of Figure 1.2 shows that the network airline domestic financial struggles are the result of both the imbalance in their own revenues and costs, as well as their very Gerchick-Murphy Associates LLC

89 15 Network Airlines Domestic Problems Continue different cost structure compared with that of the low cost airline sector. This graph compares RASM (operating revenue per available seat mile) and CASM (operating expenses per available seat mile) which are commonly used industry benchmarks for assessing changes in unit revenue and unit costs. 3 These comparisons are limited to the domestic entity. $0.140 $0.135 $0.130 $0.125 $0.120 Network CASM Unit Rates (in$) $0.115 $0.110 $0.105 $0.100 $0.095 $0.090 $0.085 $0.080 $0.075 Network RASM Low Cost RASM $0.070 $0.065 $0.060 Lo w Co st CASM MJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJS Months Ended Figure 1.2 RASM and CASM (excluding transport related revenues and expenses) (in $), Domestic Operations, Large Network and Low Cost Airlines, 12 Months Ended Data at Each Quarter Figure 1.2 shows that the network airlines massive losses beginning in early 2001 were compounded by the fact that they were driven by having both unit revenues and unit costs trending sharply in the wrong direction. It shows, for example, that just before the collapse of profits there was a two-year period of rapid cost escalation that was accompanied by similar increases in unit revenues that lasted for only one year. Therefore, while the upward trend in unit costs continued, the trend in unit revenues first flattened out for the year-ended first quarter of 2001, then suddenly collapsed, and continued to trend downward for more than a year. Revenue only began to recover significantly three years later. Equally (or perhaps more) important, this graph shows the remarkable differences between the network and low cost airline sectors. The large network airline profitability problem at the beginning of the decade was not limited to their own imbalance of revenues and expenses, but was compounded by the presence of a lower cost industry component that had continued to be profitable despite experiencing a sharp decline in RASM. This competitive presence not 3 The calculations exclude transport related revenues and expenses in an effort to concentrate on mainline unit rates. Gerchick-Murphy Associates LLC

90 16 Network Airlines Domestic Problems Continue only reinforced the need of the large network airlines to quickly and significantly realign the balance between their revenues and expenses, but also weighed on their alternatives for doing so. For example, when the large network airlines reduced capacity, both by handing off service to regional affiliates and by cutting service in many markets, their calculus had to consider the longer-term consequences of reducing capacity in the face of a growing competitor. Just as the presence of the low cost airline sector undoubtedly discouraged network airlines from even steeper capacity cuts while they strove to regain a better balance in costs and revenues, the continuing growth of the low cost sector made the capacity constraining strategy employed by the network airlines very risky. Perhaps the most important takeaway from this graph is the fact that the network sector has sustained massive losses despite having unit revenues that far exceed those of the low cost sector, yet the low cost airlines have remained profitable. Clearly the low cost airlines lower costs explain this result, and the cost gap between them and the network airlines is once again increasing. This and other trends that will later be examined provide an explanation of why the network airlines have continued to reduce domestic capacity throughout this decade, while low cost airline capacity has more than doubled. Although this macro examination of trends in unit cost and revenue presents a sharply negative picture for the large network airlines in the domestic market, it actually understates the problem they are confronted with. The improvement in RASM, for example, is not explained by a return of high-end demand that the network model has relied upon, but is explained by other factors that have now been largely exhausted. RASM is influenced by both changes in price and by changes in load factor. Increased load factors from late 2002 to early 2008 account for 60 percent of the increase in RASM that occurred during that period. The other component of RASM, or yield (average fare per mile), is not directly affected by load factor, but reducing capacity will tend to result in higher average yields because fewer seats are sold at the lowest fares. Yet, even with these capacity reductions, average yields regained just over half of the decline experienced from the high point in The network airlines also appear to have exhausted another tool they have used to try to stabilize domestic financial results. They have greatly expanded the use of their regional partners, who now serve a large part of their domestic networks. Although regional ASM capacity grew from early 2001 until 2008, in 2008 regional capacity began to decline steeply for the first time since Therefore, both mainline and regional capacity are now being reduced, and at a rapid rate. Gerchick-Murphy Associates LLC

91 17 Network Airlines Domestic Problems Continue Network airline RASM has also benefited from increased revenues flowing through their domestic networks by international bound passengers whose travel also includes a domestic segment, or so called DPIJ (Domestic Portion of Passengers traveling on International Journeys). Absent DPIJ revenues, the network airlines would not have achieved profitable results in 2006 or Indeed, the network airlines increase in DPIJ revenue has accounted for the majority of domestic profits. While the network airlines ability to further improve profitability through increased load factors and the greater use of regional affiliates is limited at best, increased international flows continue to be an option for a new source of revenue to support domestic operations. The network airline domestic capacity reduction strategy has limits. For these airlines, network strength comes from increasing the depth and breadth of their networks, not shrinking capacity. It now seems they cannot rely upon domestic expansion to rebuild their network systems. This explains the ongoing international expansion that has become the means for economic survival of the network airline sector. Rebalancing the relative sizes of domestic and international operations has been necessary, and given the remarkable capacity cuts in the domestic entity subsequent to calendar year 2000, international growth, rather than continued reductions in domestic services, must become the focus of that effort. The network airlines inability to compete profitably in the domestic market stems from three factors: the strong and growing presence of the low cost airlines, the loss of high-end demand that the network airline business model has traditionally relied upon, and the continuing significant gap in the cost structures of the network airlines compared with the low cost airline sector. The data do not suggest that any of these circumstances will become more favorable for the network airlines. Revenue Effects of Low Cost Airline Expansion: Although the impact of low cost airline entry on incumbent network airlines in individual markets has been widely acknowledged and understood for many years, the cumulative effect that low cost competition has had on network airlines is not well understood. When considered in the context of explosive growth in low cost service during this decade, together with the continuing cost gap between the two airline sectors and the collapse in high end demand, both of which are examined later, the effect on the network airlines has been fundamental. The reason for this is best demonstrated by illustrating the very different demand segments these airlines target, at least when low cost airlines are not present. The data used are from the Origin and Destination Survey Data reported by the airlines to the Department of Transportation. This survey is based on a ten percent sample of all passengers carried by the reporting airlines. The Gerchick-Murphy Associates LLC

92 18 Network Airlines Domestic Problems Continue information for each passenger includes the airports of origin and destination, connecting points where applicable, the operating and marketing airlines for each segment of travel, and the amount of the fare paid. This information typically is used to examine price and traffic changes in city-pair or airport-pair markets and differences in price between markets. In these types of examinations, the analysis is typically based on data that has been significantly aggregated. The graphs that follow here and in a later section of the report use the most detailed information available in that Survey, or record level market data for travel between cities. 4 Multiple passengers are included in the same record only if every detail of their information is identical. Thus, by using record level detail, information for every single passenger that is reported in the Survey is included. The fares charged for each passenger in the Survey are then compared to an often used industry benchmark, the Standard Industry Fare Level (SIFL), and the number of passengers and related revenues are aggregated based upon the relationship of the individual fares paid with the calculated SIFL in 10 percent blocks. 5 This provides the clearest picture of the distribution of met demand in the airline industry. Figure 1.3 shows the very different nature of met demand separately for the large network airlines in markets without competition from low cost airlines and for the low cost airlines for the second quarter of Quarterly data are used because during a period of rapid change some of the change would be averaged out using annual data. Second quarter data are used because that captures a period before the severe financial deterioration of the network airlines began. Figure 1.3 illustrates the revenue distributions based on the relationship of the fares passengers paid relative to the SIFL. 4 This analysis is limited to passengers traveling between cities within the 48-contiguous states. 5 The SIFL is a mileage based formula that reflects the regulated full coach fares that were in effect as of July 1, The development of the SIFL was required by the Airline Deregulation Act of 1978 for the purpose of evaluating post deregulation changes in airline pricing. It has been updated since to reflect changes in airline operating costs and has been used as a benchmark for examining price trends by DOT, the Transportation Research Board, and by highly regarded academics who have authored numerous airline industry studies. The tendencies revealed by the SIFL related analyses presented in this study are also supported by more traditional city-pair based data analysis. Gerchick-Murphy Associates LLC

93 19 Network Airlines Domestic Problems Continue $0.6 $0.5 Revenues (in $ billions) $0.4 $0.3 $0.2 Low Cost Network $0.1 $ Relationship to SIFL (%) Figure 1.3, Large Network Airlines (with no low cost competition) and Low Cost Airlines, Revenues (in $billions) Distributed Based on Percentage Relationship to SIFL, Second Quarter 2000 Because this type of presentation appears twice in this study, some clarifying discussion is provided here to be certain the reader understands the presentation. As shown, the X axis is labeled Relationship to SIFL. Each increment indicated represents the upper end of the SIFL increment. Increments preceded with a minus sign are populated by data for passengers who paid less than the SIFL. Toward the left of the axis, the SIFL increment of -80 indicates that the passenger (or revenue ) plots above that increment are for passengers who paid fares that were 70 to 80 percent below the SIFL. Similarly, toward the center of the graph, the SIFL increment of 200 indicates that passenger (or revenue) plots above that increment are for passengers who paid 190 to 200 percent more than the SIFL, or up to three times the SIFL. This graph provides a clear picture of network high-fare revenue that had not been affected by low cost airline competition during the second quarter of This graph shows that for the low cost airlines, 80 percent of revenue is accounted for by passengers who paid fares that were less than the SIFL, and only five percent of revenue was accounted for by passengers who paid fares that were 30 percent above the SIFL. In contrast, for the large network airlines only 35 percent of the revenue was accounted for by passengers who paid less than the SIFL and 57 percent of revenue, or $4.9 billion (nearly $20 billion annually), was accounted for by Gerchick-Murphy Associates LLC

94 20 Network Airlines Domestic Problems Continue passengers that paid fares that were 30 percent or more above the SIFL. Much of this revenue later proved to be vulnerable to low cost airline expansion. Clearly the large network airlines model relies heavily upon market price segmentation, and the dilemma that is created for network airlines when low cost airlines enter a market is obvious. Figure 1.4 shows trends in total passengers and revenues in markets that have a low cost airline competitor. 90% 80% 70% 60% Passengers Revenue Percentage Share 50% 40% 30% 20% 10% 0% D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S Months Ended Figure 1.4, Percentage Share of Total Passengers and Revenue in Low Cost Markets, 12 Months Ended Data at Each Quarter Over this more than two decade period, low cost airlines have steadily gained market-share in city-pairs they serve, and expanded into more city-pairs, bringing price competition to increasing numbers of passengers. Figure 1.4 shows that the increasing trend accelerated after In 2000, which ties back to the revenue distribution shown in Figure 1.3, just over 40 percent of network airline revenues were in markets with a low cost competitor. By the year ended September 2008, following years of strong low cost airline growth, over 70 percent of network airline revenues (and almost 80 percent of passengers) were in markets with a low cost airline competitor. The graph in Figure 1.5 compares the distributions of network airline revenues in markets that do not have a low cost competitor for the second quarters of 2000 and Gerchick-Murphy Associates LLC

95 21 Network Airlines Domestic Problems Continue Figure 1-5, Network Airline Revenue (in $billions) Distributed Based on Relationship of Fares Paid to SIFL, With Low Cost Competition, Second Quarters of 2000 and 2008 The drop in revenues in markets without low cost competition is severe, and across the board. But a large proportion of the gap is for markets where passengers paid the high-end fares in 2000, and totals $11.3 billion on an annual basis. Moreover, the 2008 distribution shows that the high-end bubble that began at about 40 percent above SIFL in 2000 has disappeared even in markets that do not have low cost airline competition. This high-fare revenue decline, whether due to low cost airline expansion or other factors, directly challenges the focus of the network airline business model of attracting higher fare paying passengers to cover their higher costs of operation. Although the number of network airline passengers carried in markets with low cost airline competition has increased by about 50 percent since 2000, and revenues have greatly increased as well, virtually none of that revenue growth was from passengers who pay high fares relative to the SIFL. To fully appreciate this loss of high-fare demand by the network airlines the continuing and significant capacity reductions for the seven year period since calendar year 2000 shown in Figure 1.6 have to be considered. All else equal, one would expect that capacity reductions would lead to higher average fares. If nothing else, large scale capacity reductions would enable spilling of more price sensitive passengers, and this alone would push up average fares. But if the network airlines had meaningful pricing power, such capacity reductions would also enable them to increase prices for some other passengers. But this has not Gerchick-Murphy Associates LLC

96 22 Network Airlines Domestic Problems Continue happened. It is clear that the network airlines now have very limited pricing power in the domestic market. Figure 1.6 shows ASM trends following calendar year 2000 for the network airlines and also for network airlines and their regional affiliates combined Network Mainline Network Mainline and Regional Affiliates M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D Figure 1.6, ASM (in billions) Growth Trends, Large Network Airlines Mainline Operations and Large Network Airlines Mainline Operations Plus Regional Airline Affiliates Operations, 12 Months Ended Data at Each Quarter While network airline ASM capacity steadily increased until calendar year 2000 (not shown), Figure 1.6 shows that it has steadily declined through the year ended first quarter 2008, other than a small increase during Moreover, spurred in part by the fuel crisis, network airline capacity cutbacks were significantly accelerated during the fourth quarter of Network airlines domestic mainline capacity has declined by 28 percent since 2000, and 18 percent if the capacity of their regional partners is included. Part of the decline in mainline capacity is related to transferring a large amount of service to regional affiliates, and while capacity for affiliated regional partners has grown throughout much of this period, that trend changed in early Also, as noted, the capacity cuts in network airline mainline operations and regional operations show a marked increase during the fourth quarter of Compared with a year earlier, for the fourth quarter of 2008 mainline ASM capacity was down by 12 percent and regional capacity was down by 10 percent. These capacity cuts have continued into With these recent capacity cuts, on an annual basis, network mainline capacity is now down by more than one third Gerchick-Murphy Associates LLC

97 23 Network Airlines Domestic Problems Continue since calendar During this same period, low cost airline ASMs have slightly more than doubled. The Cost Gap With Low Cost Airlines: Despite successful efforts by the large network airlines to achieve huge savings in operating costs earlier in this decade through bankruptcies and other dramatic steps, a large cost gap continues to exist and is now steadily and significantly increasing. This fact, standing alone, does not establish that they cannot profitably compete with the low cost sector, because higher costs are only an issue to the extent that higher revenues do not compensate. But as just demonstrated, the network airlines ability to compensate for higher costs has substantially changed, and the continuing unsatisfactory financial performance in the domestic market shows that the imbalance in their revenues and expenses remains problematic. Figure 1.7 shows CASM for the network and low cost airlines, adjusted for distance, and shown separately with and without fuel costs. $0.14 $0.13 $0.12 $0.11 Network CASM Unit Cost per ASM ($) $0.10 $0.09 $0.08 Low Cost CASM $0.07 Network CASM (excluding fuel) $0.06 $0.05 Low Cost CASM (excluding fuel) $0.04 MJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJ SDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJS Months Ended Figure 1.7 Network and Low Cost Airlines, Total Operating Cost per ASM (CASM) ($) (excluding transport related expenses and adjusted for distance), With and Without Fuel Expense, Domestic Operations, 12 Months Ended Data at Each Quarter While the network airlines efforts to reduce operating costs initially resulted in a cost gap that was lower than existed before the cost run-up that began in 1999, the cost gap has since increased by 32 percent. Of course fuel price increases Gerchick-Murphy Associates LLC

98 24 Network Airlines Domestic Problems Continue account for a major part of the increase, but the gap in non-fuel costs has also increased by 22 percent. As of September 1999, or just before the spike in network airline costs, their CASM was 3.73 cents higher than that of the low cost airlines. In March 2002, when at its peak, the cost gap was 5.02 cents. The cost gap then steadily declined until it bottomed out in March 2007, when the actual difference was lower than in 1999, or 3.42 cents. Since then, network airline costs have increased by nearly one third to 4.5 cents. When considered in the context of the ongoing poor domestic financial performance and the revenue-side problems that continue for the network airlines, the return of an increasing cost gap with the low cost airline sector is particularly problematic. The other important comparison that stands out from this graph is the more recent trends in non-fuel CASM. The downward trend in network airline costs flattened out at the end of 2005, and then began to steadily increase in early In comparison, low cost airline non-fuel CASM has remained remarkably stable for the past several years. These data clearly do not offer any suggestion of convergence between the costs of network and low cost airlines during the foreseeable future. If, as seems likely, a significant narrowing between the network and low cost airlines CASM does not occur, what does this say about the competitive balance between the network and low cost airlines? An assessment of all the trends now in play and the marketplace circumstances that are driving those trends indicate that network airlines are well positioned to remain strong global competitors, but that they will not regain the competitive position in the domestic market that existed during the decade of the 1990s. The network airlines very different prospects globally and domestically are reflected in a combination of their financial results and growth trends. As demonstrated earlier, in the domestic market, despite massive capacity cutbacks that have limited capacity growth, it took nearly six years after the 2001 financial collapse to edge into the black, and that did not last. In contrast, the network airlines have continued to profitably expand international operations since the early 1990 s, consistently reporting profits except during a three-year period following the 2001 terrorist attacks, when operations were also affected by SARS and other terrorist concerns at the beginning of the Iraq War. There is no evidence to suggest that these broad indicators are misleading. For these reasons, the network airlines have had to look outside the domestic market for ways to reestablish network strength that will enable them to compete at home and system wide. Their international expansion strategy has been driven by economic and competitive realities at home. And the strategy of making better use of their own primary competitive advantage, their global networks, also appears to be the most economically sound response to the circumstances they face. Gerchick-Murphy Associates LLC

99 25 Section II Alliance Integration Overview The Role of Networks in International Service Antitrust Immunity Gerchick-Murphy Associates LLC

100 26 Alliance Integration ALLIANCE INTEGRATION The key for airline growth and consumer benefits Overview: The information presented in Section I shows that the network airlines have lost their once industry-leading position in the domestic market, and that the low cost airlines, not the network airlines, now drive change in both capacity and price. It also explains why this transition of competitive influence will not be significantly reversed in the foreseeable future, if at all. This section will provide information that will show why the key to rehabilitating the stability of the network airlines overall systems, including their domestic networks, is continued international expansion. However, the prospect for continued successful international growth is highly dependent upon the continuing ability to integrate multinational alliances which, in turn, is highly dependent upon having antitrust immunity (ATI). As with virtually any other business, an airline s ability to successfully serve and compete in a market is directly related to its ability to improve price and service to consumers. Scheduled airline service is a highly complex undertaking that requires careful integration of detailed information and ideas from a broad and diverse set of disciplines in order to optimize decision making, and therefore the appeal of the product to customers. In international markets, this undertaking is complicated by the fact that the vast majority of city-pair markets can only be effectively served by connecting service, the type of service that is most conveniently provided by multinational alliances. At best, alliances are imperfect business model substitutes for cross-border mergers, a type of transaction that is prohibited by foreign ownership laws. Alliances have difficulty in reaching decisions on most activities that affect consumer price and service because all decisions must be joint; no one is fully in charge. But decision making is even more severely hampered by the fact that, absent antitrust immunity, alliance partners cannot jointly develop the core components of airline service scheduling, pricing, and capacity, plus a litany of other matters that affect their ability to operate, market, and sell a better global product without risk of antitrust liability. Government policies that prohibit cross-border mergers and also limit alliance integration could be said to doubly handicap cooperation and efficiency. While there is nothing that can be done to remove the need for joint decisionmaking in multinational alliances, antitrust immunity at least enables the alliance partners to use all the tools and information that an individual airline can use to Gerchick-Murphy Associates LLC

101 27 Alliance Integration maximize its ability to better serve customers and compete with other market participants. Without question, the key to optimizing the effectiveness of joint networks is integration. Although some measure of cooperation can be achieved in the absence of immunity from the antitrust laws, that cannot match the degree of integration made feasible with ATI. In the absence of antitrust immunity, too many decisions that contribute to the effectiveness of the joint service have to be made independently by the alliance partners, and such decisions are likely to be focused on self interest rather than the consumer benefits that could be better met by involving the other alliance partner. DOT has recently focused more of its attention on the benefits of network integration. It recently conditioned the grant of ATI for two alliances with multiple large airline partners on their participation in revenue-sharing joint venture agreements. This shifts the focus of the alliance carriers from self interest to promoting alliance development, and will result in a more integrated network with consumer benefits. Unfortunately, airline alliances are sometimes treated as something other than what they are -- a business model that is designed to fill a need created by government interference in the market. Tens of thousands of city-pair markets and millions of passengers in the transatlantic market alone can be effectively served only by alliances. Being able to use all of the tools required for providing effective service and pricing options, through network integration, are no less important for alliances (and consumers) than for individual airlines. When the competitive implications of linking networks are positive, it is not productive to prevent the alliance partners from jointly integrating in order to enhance the alliance s ability to offer improved service and price. Put another way, when merger analysis concludes that combining multinational networks would not have an anti-competitive result, then failing to facilitate network integration and collaboration through the grant of antitrust immunity would harm both consumers and airlines alike. To the extent multinational networks would not be able to integrate effectively in the absence of immunity; the consumer would be the first to suffer from the resulting service and price limitations. And where such service and price limitations affect a sufficiently large number of passengers, overall demand is diminished. Less consumer demand means less airline service, fewer airline jobs, fewer travel-related jobs, and, ultimately, less economic growth in the communities that bear the brunt of less service. These effects would not be limited to depressed growth in international markets, but would also spill over into other areas that are served by the respective domestic networks of the alliance partners. For example, for the U.S. network airlines, 55 percent of their international passenger revenue is generated from behind gateway cities, and such revenue flows through and supports service in the carriers domestic networks. While purely domestic network airline revenue Gerchick-Murphy Associates LLC

102 28 Alliance Integration has decreased by approximately 10 percent (not adjusted for cost inflation) since 2000, international revenue from behind gateway cities has increased by 50 percent during that same period and now provides revenue support for domestic services of $2.5 billion annually. Expanded international operations, therefore, provide substantial and much needed support for the network airlines otherwise unsatisfactory domestic performance. The significance of the international growth strategy that the network airlines have now pursued for much of this decade is fundamental. As with network industries generally, the large network airlines have natural economic incentives to expand. The more destinations served by an airline network, the greater its value to customers. A key reason for this is that the primary advantage of a network, indeed its most basic and perhaps only sustainable competitive advantage relative to other commercial airline business models, stems from the fact that every discrete part of the network supports and, in turn, is supported by revenue flows generated by the rest of the network. International expansion, therefore, has not just resulted in improving system financial results for network airlines, but also in increasing flow revenues that support the rest of their system of networks, particularly the carriers domestic networks. As detailed in this study, the natural incentive for the large network airline industry to expand internationally has been sharply elevated in recent years by their weakened competitive position in the domestic market. But alliance integration is a costly, time-consuming process that takes years to implement. The commitment of alliance partners to the success of such an effort is therefore affected by the perceived stability of their alliance. Although there are no guarantees that alliances will continue, any governmental actions that suggests a reversal of long-standing U.S. government policy that recognizes the value of providing protection from the antitrust laws for pro-competitive alliances could certainly chill alliance integration. This would, in turn, diminish the benefits immunized alliances provide to consumers, and much needed networkenhancing growth for the network airlines. Stability is particularly important for alliances since they require more time to mature than mergers because the partners remain fully independent companies that are merely cooperating. Whether or not the large network airlines are able to continue to expand their international operations will have far reaching consequences. According to the Air Transport Association of America, its member airlines estimate that the loss of antitrust immunity for their international alliances would result in service cutbacks and the loss of 15,000 jobs. Further, given their domestic competitive position, limiting their ability to grow internationally could well lead to much more substantial job loss and also the reduction of service to hundreds of smaller cities that receive all (or virtually all) of their service from network airlines and their regional affiliates. Destabilization of DOT s immunity policy would certainly increase the pressure on U.S. network airlines to merge in an effort to strengthen their networks. Gerchick-Murphy Associates LLC

103 29 Alliance Integration The Role of Networks in International Service: Immediately after the formation of large domestic network systems of service following U.S. deregulation, domestic networks began to affect international aviation. Figure 2.1 shows domestic and transatlantic RPM growth from calendar 1978 through calendar 2000, indexed to Atlantic RPMs Scheduled RPMS Indexed (12/31/1978 = 100) Domestic RPMs DMJ SDMJ SDMJ SDMJ SDMJ SDMJ SDMJ SDMJ SDMJ SDMJ SDMJ SDMJ SDMJ SDMJ SDMJ SDMJ SDMJ SDMJ SDMJ SDMJ SDMJ SDMJ SD Months Ended Figure 2.1, Domestic and Atlantic RPM Growth (in billions), Indexed to 1978, Twelve Months Ended Data at Each Quarter. Both the transatlantic and domestic markets experienced strong traffic growth with the formation of domestic networks during the mid-to-late 1980s. Directly linking U.S. network airlines domestic networks to large European cities resulted in superior service compared to that previously provided, and stimulated demand. Transatlantic growth remained strong before beginning to taper off in 1989, and then, even after the demand-reducing problems in the early 1990s had been resolved, remained relatively flat until 1996, when immunized alliance activity greatly accelerated. The U.S. network airlines not only experienced strong growth, but also gained market share. The shift in competitive balance as the U.S. network airlines became larger providers of U.S. international service prompted a response by European airlines to develop their own domestic hub networks to feed their own international operations. The early stages of network development, therefore, led Gerchick-Murphy Associates LLC

104 30 Alliance Integration to the beginning of a fundamental structural change in the nature of international operations on both sides of the Atlantic. Network systems of service are necessary for the vast majority of longer-distance international markets. Using the transatlantic as an example, about 450 cities in the U.S. and an equivalent number in Europe receive scheduled passenger service, and more than 40,000 of the potential 100,000 city-pair markets show up in the Origin and Destination Survey reported by U.S. airlines. Foreign airlines do not report information that allow us to know how many additional city pairs they serve that are not reflected in the U.S. carrier reports, but the number is in the thousands. But the reality is that a very small number of either cities or city pairs between the U.S and Europe have sufficient passenger volumes to sustain nonstop scheduled service, and even fewer markets can sustain competitive nonstop service. In the summer of 2008, only 30 U.S. cities received nonstop service to Europe, and only 45 European cities received nonstop service to the U.S. 6 Most of these cities were served as spokes for the larger U.S. or European network hubs. Only 220 city-pairs received nonstop service across the Atlantic. All of the cities that received service were either network hubs or were linked to network hubs. Therefore, absent the presence of networks, only a fraction of even the very small numbers of cities, city-pairs, and hubs that received service would have been able to sustain any nonstop, scheduled service. Even with the extensive support of network flows, most city-pairs that received nonstop service were served by a single airline, and the vast majority of those were served with a single round trip a day. Only about 65 nonstop segments received nonstop service by more than one airline, and only 15 were served by three or more airlines. The domestic networks of the U.S. and E.U. network airlines enable convenient connecting services from behind their respective domestic gateways across the Atlantic. U.S. airlines independently provide connecting service to 12,000 of the more than 40,000 city-pairs included in the DOT traffic reports. We estimate that, combined, less than one half of the transatlantic city pairs can be served by single U.S. or foreign airlines and that the remaining city pairs depend upon joint services of two or more international airlines. Although independent airline networks effectively serve large numbers of markets and passengers, these data show the impact of aero-political and economic limitations. Single carrier systems cannot independently serve tens of thousands of international city pairs that account for millions of passengers. As a result, various forms of multinational alliances have led the way in extending improved service and price to transatlantic air travelers, particularly for customers in smaller connecting markets. However, passengers traveling in virtually all long 6 Service of at least five frequencies per week. Gerchick-Murphy Associates LLC

105 31 Alliance Integration distance international markets, including even the largest with existing nonstop service, benefit from having increased connecting options. Multinational alliances range from relatively modest marketing agreements, to various types of code-share alliances that provide more meaningful service enhancements, and ultimately to immunized strategic alliances that combine the full scope of large domestic networks on different continents in a manner that comes much closer to a true online service. At the beginning of 1996, there was an array of alliances in place across the Atlantic, only one of which had been granted antitrust authority: Northwest -- KLM. In mid-1996, two more alliances that had been limited to code-share relationships received antitrust authority from DOT: United Lufthansa, and Delta Austrian/Swissair/Sabena. This is the most obvious change in transatlantic circumstances that would have prompted the strong growth trend shown in Figure 2.1 that lasted until the terrorist attacks of DOT has issued two reports that examine the effect of immunized alliances in the transatlantic market. 7 DOT concludes that the surge in traffic that began in 1996 was indeed the result of immunized alliance activity; that immunized alliances had stimulated demand and fueled an enormous increase in connecting traffic (both in markets that had historically suffered from poor quality interline service and by providing service alternatives in markets that already had the benefits of seamless online service by individual airlines); and that the very strong growth resulting from immunized alliance development provided overwhelming evidence that they had greatly expanded consumer welfare. The first multinational alliance that received antitrust immunity from the Department, Northwest and KLM, provided strong evidence of the appeal of such alliances to consumers. Each of these airlines was a mid-sized network airline, but their alliance became a model for alliance development that led to enormous expansion. Before receiving immunity, Northwest did not provide daily service from any of its domestic network hubs to KLM s Amsterdam hub. Within a very short period of time, however, Northwest provided multiple frequencies from both of its primary hubs, Minneapolis/St. Paul and Detroit, and also added service from other U.S. cities, such as its relatively small hub at Memphis and other smaller cities such as Portland, Oregon. Northwest now provides 15 daily nonstops between eight U.S. cities and Amsterdam. This greatly expanded service is clearly attributable to network traffic flows that also support each airline s domestic network. These increased network flows lead not only to increased transatlantic travel, but strengthen the reach of the network hubs. KLM, in particular, greatly facilitated its ability to expand significantly the catchment area of its Amsterdam hub, and this allowed KLM not only to serve 7 International Aviation Developments: Global Deregulation Takes Off, December 1999, and International Aviation Developments, Transatlantic Deregulation, The Alliance Network Effects, October Gerchick-Murphy Associates LLC

106 32 Alliance Integration more transatlantic markets, but enabled Northwest to serve Amsterdam from several non-hub U.S. cities. It is now clear that international alliance development was still in an early stage at the time the DOT reports were prepared in 1999 and Subsequently, Delta terminated its immunized alliance with Austrian, Swissair and Sabina and entered into a much more far-reaching alliance with Air France. New airlines received code-share authority, and new partners were added to existing immunized alliances. The result of these and other changes is shown in Figure 2.2, which provides an updated comparison of domestic and transatlantic growth from 1994 through 2008, indexed to Scheduled RPMs Indexed (3/31/1994 = 100) Network Airlines' Atlantic RPMs Network Airlines' Domestic RPMs M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D Months Ended Figure 2.2, Network Airlines Domestic and Atlantic RPMs, Indexed to 1994, Twelve Month Ended Data at Each Quarter Although domestic RPM growth slightly exceeded Atlantic growth for the first four years of this period, the faster rate of transatlantic growth that began in late 1996 eventually surpassed domestic growth in mid-1998 and then quickly and steadily pulled away until mid The gap between domestic and transatlantic growth declined somewhat until the end of 2003, and then steadily expanded at a very rapid rate through calendar year 2008, as transatlantic RPMs increased by 54 percent while domestic RPMs, after increasing slightly until late 2005, declined to their lowest level since Continental Airlines, which has not been an immunized participant in a transatlantic alliance, also greatly expanded across the Atlantic during this period. Ninety percent of its growth was from its Newark hub, Gerchick-Murphy Associates LLC

107 33 Alliance Integration Transatlantic growth trends are strong not just in comparison with domestic trends, but also when compared with other international operations. Figure 2.3 compared RPM growth for all three international reporting entities since Scheduled RPMs (billions) M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D M J S D Months Ended Network Atlantic RPMs Network Latin American RPMs Network Pacific RPMs Figure 2.3, Network Airline RPMs (in billions) by International Entity, Twelve Months Ended Data at Each Quarter Although traffic has grown in all three entities, the Atlantic entity, which has experienced the most immunized alliance development, shows by far the strongest growth, particularly during the last five years when immunized alliance activity has significantly intensified. Antitrust Immunity: Gaining U.S. government approval of a request for immunity from the antitrust laws is by any measure an extremely intense and vigorous exercise. 9 In making which has an enormous local traffic base and is perfectly positioned to flow passengers from large destinations throughout the U.S. to Europe. No other airline could duplicate this kind of unilateral growth. While its unique Newark hub provided Continental an opportunity to expand in the absence of an immunized alliance, its growth was in part the result of its code share relationship in the SkyTeam alliance. Continued strong growth will depend upon participating in transnational alliance development. 9 The Secretary of Transportation may exempt an airline from the antitrust laws after reaching the decision that it is required by the public interest, 49 U.S.C (b). In practice, the DOT uses a two step process. It first considers approval of the alliance agreements under 49 U.S.C by conducting a competitive analysis to determine whether the alliance agreements would substantially reduce or eliminate Gerchick-Murphy Associates LLC

108 34 Alliance Integration its determination to grant or deny immunity, DOT conducts a thorough, formal, public examination. The information relied upon in the competitive analysis includes the extensive detailed data routinely filed by the airlines with DOT, plus large quantities of information (much of it highly business sensitive) requested of the joint applicants by DOT. Any party that wants to either support or oppose a request for immunity can formally participate in the DOT public proceeding. Applications for antitrust immunity are also reviewed by the Department of Justice, which occasionally also participates as a party in the formal review process. Immunity requests are also subject to independent review by foreign antitrust agencies and are routinely commented upon by elected officials. This process is thus informed both by the type of confidential examination that the Department of Justice conducts for domestic airline mergers and the formal record developed by DOT s own extensive information requests, and occasionally by other parties participation in the public proceeding. Unlike the Hart-Scott-Radino process used by the Department of Justice in evaluating mergers, DOT s alliance review process is more transparent and open to participation by all interested parties. Although the Department has not conducted a formal study of the effects of alliances since its 1999 and 2000 studies, it continuously monitors their development using extensive data required to be filed by airlines that receive immunity. The Department also, in effect, examines alliance benefits each time an application comes before it. With over a decade of in-depth alliance development analysis, the Department has gained an understanding of how alliances work in practice and what commercial activities actually require immunity. Based on that knowledge and experience, the Department holds new applicants to a stringent standard, requiring them to demonstrate the activities they believe require antitrust immunity, and to document the public benefits that would only directly result from such activities. As noted, DOT has also increased its focus on the benefits of revenue-sharing joint venture agreements for larger alliance partners in order to facilitate the implementation of the network integration needed to produce consumer benefits. This last step helps ensure that antitrust immunity will result in important consumer benefits from greater coordination of schedules, pricing, and capacity management. This level of cooperation impacts other areas of decision-making that produce service benefits, efficiencies and cost savings that can be passed on to consumers in the form of lower prices. competition. If DOT approves the alliance agreements, it will then grant immunity from the antitrust laws under 41308(b) if it is persuaded that such immunity would facilitate the implementation of the alliance agreements and the realization of public benefits. DOT can condition approval upon enactment of remedies, and it always requires foreign alliance partners to provide the type of Origin and Destination Survey data routinely filed by U.S. airlines to allow for review of alliance results. Gerchick-Murphy Associates LLC

109 35 Alliance Integration Absent antitrust immunity approval by DOT, the alternative option for alliance partners is to gain joint venture approval in a Business Review Letter signed by the Department of Justice. However, the airlines themselves believe this alternative approach, while helpful, would have the effect of restraining activities necessary to optimize alliance integration. An important reason for this is the threat of private law suits. Private suits would not have to prevail in court to succeed in chilling legitimate pro-consumer alliance activities. The extent to which a court would be influenced by the existence of a Business Review Letter could also be affected by the fact that alliance development is a very dynamic process that could open the door to challenges. This fact could also cause alliance partners to be less aggressive in their efforts to grow their alliances. There are sound policy reasons to support the use of antitrust immunity in the case of multinational airline alliances. The use of antitrust immunity is important because it is the only vehicle available to airlines to effectively provide integrated joint services in international markets that cannot be served by single independent airlines without running a very real risk of being subject to private antitrust litigation. Despite the economic benefits that could result from an endto-end merger between a U.S. airline and a foreign airline, mergers between U.S. and foreign airlines are prohibited under existing U.S. ownership laws (and most foreign laws as well). Receipt of immunity from the antitrust laws allows alliance partners to collaborate in efforts to integrate, and to the extent possible act as one in order to expand output and improve service to consumers. There is no evidence that immunity has led to alliance carriers engaging in anticompetitive behavior against other airlines. Transborder mergers are not the only form of international network expansion limited by government legal barriers. U.S. network airlines cannot set up subsidiaries in the EU to operate a system of feeder flights to support intercontinental US-EU services, and the same is true for EU carriers in the US. Neither US nor EU carriers can invest in existing foreign carriers and gain control of them because of foreign ownership limitations. US and EU carriers cannot partner with financial institutions to set up new carriers outside their home jurisdictions unless they are prepared to cede to the financial partners control of the venture. Any of those structures would be much less risky from an antitrust point of view than a contractually-based JV with a horizontal competitor, where the degree of network integration is by definition going to be less. Legal restrictions on U.S. airline ownership constitute an intractable barrier to international expansion. They prevent the deployment of economic tools that are commonly used to develop domestic networks, thereby forcing carriers to resort to less efficient means to serve passengers in international markets. These constraints harm consumers, airlines, and their employees. To the extent that airlines can develop JVs in international markets, those restricted opportunities are further constrained by the antitrust risks involved. In this context, DOT s antitrust authority provides a means whereby airlines can overcome those limitations in order to implement JVs that demonstrably benefit consumers and Gerchick-Murphy Associates LLC

110 36 Alliance Integration enhance efficiency. In the absence of DOT s antitrust authority, the airline industry could well decide to take somewhat greater antitrust risks, but it is unlikely that they would implement the broad-based alliances that have proved to be so valuable. The development of a successful plan to operate a broad-scale aviation network is a complicated challenge for even a single airline. The starting point for a successful airline business plan is targeting demand and developing a means of profitably serving and competing for that demand. But in addition to developing a sound business concept, a seemingly endless array of problematic choices have to be successfully dealt with in order to put the plan in motion, and these choices must be monitored and adjusted over time as necessary to meet changes in demand, costs, market opportunities, and competition. For all large network airlines these doing-business choices include capacity, scheduling, pricing and yield management strategies, along with the IT platforms, information systems, personnel, and processes that are necessary to monitor and adjust these critical business components. Airlines operating network systems must also have strategies, practices, and processes for marketing, sales, distribution, customer service, reservations, seat inventory, frequent flyer programs, and a long list of other activities that compose the full range of activities that combine to complete the business model. Even though many of these activities are exceedingly difficult to plan, manage, and implement, they must be fully coordinated to meet the overarching service and pricing goals of the airline. The complexities of successfully meeting all of these needs increases at a geometric rate as the geographic size of an airline increases, particularly for airlines that use network systems of service. A large hub network, for example, can offer connecting service in thousands of individual city-pairs, but cannot provide service to all potential markets or provide optimum service, either in terms of minimizing connecting delays or offering preferred times of service, to all city-pairs it serves. Since an airline cannot simultaneously optimize these choices for all city-pairs it serves, it must make choices in deciding how to structure the services it provides to maximize the overall revenue potential of the network by better meeting the needs of the greatest numbers of consumers. This is further complicated by the fact that every large U.S domestic network airline operates multiple hub-and-spoke networks in the domestic market that can serve the same city-pairs, including foreign destinations. For a network airline to maximize revenues and its competitive position vis-a-vis other market participants, it cannot optimize service alternatives for connecting consumers on a hub-by-hub basis, but must integrate the potential synergies of routing passengers over multiple hubs. Gerchick-Murphy Associates LLC

111 37 Alliance Integration For example, American Airlines operates hubs at Chicago and Dallas, and each of those hubs provides connecting service alternatives to literally thousands of the same city-pairs. From either a customer or airline point of view, it would not make any sense for American to decline to consider, and indeed coordinate, the services it provides over both of those hub networks when selling service in any of the city-pairs each hub serves. The customer might well have better price and service options by traveling over one hub in one direction and over the other hub in the other. Moreover, because of greater availability in capacity at one of the hubs at any particular time, greater pricing alternatives might well be available to the customer if both hubs are a travel option. The appeal from American s point of view is that by considering all service alternatives and thereby offering more service options to consumers, it increases its competitive position with other airlines that may serve the same city-pairs. In airline parlance, the greater the number of paths that are available and effectively coordinated, the greater the odds are of satisfying a customer s price and service needs. Obviously, both the consumer and the airline benefit from the fact that the latter is able to develop the full potential of its service and pricing models through careful integration of all aspects of that service. In other words, by integrating the various hub-and-spoke components of their broader systems all independent airlines are able to provide better service and also provide service to more communities and passengers. All network airlines have developed elaborate technology platforms and information systems that are designed to accomplish this. These systems, and the data inputs and decision points that trigger adjustments in these systems, are constantly reviewed throughout the organization. The key to making all this work effectively and efficiently is the constant communication of information and ideas, and the opportunity to work for consensus where differing views emerge. The need to accomplish the same level of network integration is no less important for the joint networks that are needed to provide quality service for the great preponderance of international city pairs. Indeed, if anything, it is more important because a far greater proportion of total traffic in intercontinental markets is connecting traffic that relies on careful scheduling, pricing, revenue management, and product integration. Because these are core areas upon which carriers compete, antitrust immunity is required to facilitate this cooperation; without it the consumer benefits from the integration would not be possible. For immunized alliance partners, in fact, local passengers account for relatively small proportions of total traffic, and are often outnumbered by the so called bridge traffic -- passengers traveling in the thousands of smaller city pairs that neither alliance partner can serve independently. However, the network integration needed to better serve these and other connecting passengers is far more complicated and difficult for alliance partners to accomplish than for an individual airline for several reasons. In particular, alliances must integrate larger networks that serve different cultures, operate with different corporate cultures, Gerchick-Murphy Associates LLC

112 38 Alliance Integration and have different service and pricing strategies, as well as support systems and processes. United and Lufthansa, for example, have eleven hub-to-hub transatlantic routes that create multiple connecting options for passengers, especially those that can only be served by joint services. Both the need for integration and the problems of accomplishing that are clear. But it is also clear that large numbers of passengers have been attracted to the integrated services they jointly provide. They have increased frequencies on the two hub-to-hub routes they operated at the time they were granted immunity, and they have added service on nine additional hub-to-hub routes, several of them involving Lufthansa s smaller hubs of Munich and Düsseldorf that rely heavily upon connecting traffic flows to support the service. This is also clear evidence that network integration is a dynamic process that is sparked by the receipt of antitrust immunity, which allows such integration to proceed without risk of private antitrust litigation. As the immunized alliance succeeds in jointly creating the necessary policies and procedures for successful integration of existing services, the groundwork is developed for new services that likely would not otherwise have been viable. When these new services are integrated with the broader network, additional passengers are attracted to the alliance, which leads to further capacity increases not only on pipeline routes, but also on additional intercontinental routes from the alliance partners respective hubs. In addition, due to the effective joint service that evolves, the carriers also benefit from the halo effect resulting from each alliance partner s experience in their respective home markets. The increased revenue flows across the Atlantic also contribute to the alliance partner s respective homeland networks. The continued strong growth in the transatlantic market where alliance development has been the most robust supports the concept that alliance development is both a dynamic and long-term process. The successful alliance development has not only benefited consumers and airline profitability, but has resulted in increased jobs for the airlines and travel related industries. This type of successful integration simply cannot result from unilateral adjustments made by the respective alliance partners; it requires extensive collaboration of ideas and sharing of business sensitive (competitively sensitive) information that would not occur in the absence of antitrust immunity because of the risk of private antitrust litigation. Just as schedule integration made possible by antitrust immunity triggers significant growth, capacity management decisions necessary for accommodating that growth are also greatly enhanced by immunity. The right capacity on pipeline routes is the key to achieving the maximum potential connecting benefits of the alliance, as well as scheduling to enhance service for local passengers traveling on the pipeline routes. Only with immunity can the alliance partners fully collaborate on developing an understanding of how improved schedule coordination and pricing will likely affect increased Gerchick-Murphy Associates LLC

113 39 Alliance Integration connectivity and, thus, the capacity needed to accommodate increased demand over the network linking pipeline routes and elsewhere in the broader network system. Conversely, the lack of immunity would not just limit the ability of the alliance partners to offer improved schedule coordination and price; it would also increase the likelihood that the critical pipeline capacity would not be set at the optimum level. The benefits to consumers of an immunized alliance s capacity decisions are not limited to scheduling improvements. The heightened ability to collaboratively adjust alliance capacity is also advantageous to customers and airlines when seasonal changes are needed and when significant disruptions occur that require rapid network adjustments. An example of the former is shifting capacity in order to adjust to differing seasonality changes throughout the broader network. Examples of the latter include spikes in fuel prices, economic downturns, acts of terrorism, and SARS-type issues. All of these circumstances are best addressed by jointly-developed action plans designed to accommodate them. For example, when severe cost or demand fluctuations such as fuel-cost spikes or economic downturns necessitate capacity reductions, collaborative decisions by alliance partners are likely to leave more capacity in a market and better distributed capacity than would result if each partner acted independently to bolster profitability. In addition, immunized alliances have the incentive and authority to make pipeline scheduling decisions that are totally focused on consumer needs, rather than the parochial interests of each alliance partner. For example, instead of each partner scheduling flights at prime times, they would be motivated to better distribute flights to offer alternative timing that will give some consumers more convenient schedule options. A better understanding of how much traffic each partner forecasts over its system helps both partners estimate their capacity needs for the pipeline hub. Developing effective pricing policies and strategies is another core function that is enhanced by antitrust immunity. Until the development of alliances, large numbers of long-distance international markets were poorly served and suffered from high prices. This combination of charging more for less is not an intuitive result, but there are sound economic reasons for this, and also for the fact that the reverse is true at the other end of the price/quality-of-service spectrum. On the one hand, when selling a joint service that offers little or no service enhancements to stimulate demand, each airline tends to mark up price to provide a profit for its portion of the joint service. Conversely, successful implementation of service enhancements made possible by antitrust immunity shifts the focus from pricing to protect profits for small volumes of passengers to increasing profits by growing demand with lower, more competitive prices. The network airlines efforts to enhance their competitive positions through international market expansion have a long history of marked change. Before Gerchick-Murphy Associates LLC

114 40 Alliance Integration broad-based alliances came of age, airlines jointly served markets that could not be served by independent airlines primarily through poorly coordinated interline service that often had little or no enhancements that benefited consumers. Interline services had very limited competitive significance. Such service was difficult to market, and involved little, if any, lasting commitment by the partners. Carriers were reluctant to invest the time and money required to enhance such service. Although some interline service was enhanced with other forms of cooperative arrangements, such as joint marketing agreements, and that likely produced somewhat better results, the more significant change in selling joint services came in the form of code-share alliances. Code-share service typically has a higher value than pure interline service, particularly when the code-share partners successfully integrate their services so that they have some of the same advantages of true on-line service, such as better service coordination, single check-in, and coordinated baggage handling. This limited integration significantly changes the competitive posture relative to pure interline cooperation and leads to more competitive pricing. However, codeshare partners continue to have incentives to focus more on their individual interests than on growing the alliance s network of service. This tendency is best overcome where the partners enter into a metal-neutral agreement that shifts profit focus so that alliance growth is the path to greater profits for all participants. The price effects of differing levels of integration are substantial. Between Washington and Istanbul, for example, recently the lowest prices for joint service that included United ranged from $2,107 for interline service jointly with British Airways connecting over London, to $947 for service jointly with its immunized Star partner, Lufthansa. In between, the price for code-share service jointly with a non-immunized Star partner, Turkish Airlines, was $1,299. The higher fare with British Airways reflects the fact that United does not have an alliance relationship with British Airways, and the quoted fare was simply the sum of their respective local fares. Consumers have little incentive to use that joint service unless other alternatives are not available. Compared with its joint price with Lufthansa, the higher price with its non-immunized code-share partner, Turkish Airlines, reflects the fact that, in the absence of immunity, both carriers have the incentive to charge a mark-up on their respective segment s of the itinerary. This double marginalization is wide spread for joint fares that do not involve immunized partners. As the airline model for serving long-distance international markets has evolved from one virtually devoid of effective joint collaboration to one with full collaboration, it is clear that the degree of alliance integration significantly effects price in a manner that is helpful to consumers. It is also important to understand that, with immunity, in making joint pricing decisions, alliance partners have not just the authority, but also the incentive to share business-confidential information, experience, and expertise to fully collaborate in developing specific prices that are designed to maximize consumer appeal and therefore increase Gerchick-Murphy Associates LLC

115 41 Alliance Integration competitive pressures on others in markets they can only serve jointly. But specifically because of the potential competitive effects on others, engaging in these types of activities without antitrust immunity, even where the joint venture has been approved by the Department of Justice, would be risky. The ability of multinational alliances to produce price benefits has been documented in airline studies that link the conceptual theory with an analysis of detailed price data reported by the U.S. airlines. Compared with interline pricing that would otherwise occur, Jan Brueckner showed a price reduction of 17 to 30 percent for the combination of code-sharing and antitrust immunity. 10 Tom Whalen showed that, Immunized alliance fares are 19% lower than non-alliance fares and statistically no different from fares for single carrier service in most specifications. Code sharing alliances, by comparison, are associated with only 11% lower average fares. 11 Immunity also creates incentives for airlines to engage in other customer-friendly actions, some of which could occur without immunity, but nevertheless are much more likely in an immunized relationship given the traffic and revenue growth that results and the increased ability of the alliance partners to collaboratively examine the revenue impact. An example of this is co-location of alliance partners within airport terminals. With or without immunity, co-locating in terminals provides easier connections for consumers by reducing the time required to make connections between flights. But co-location is a timeconsuming and costly process that is much more likely to be financially viable for immunized alliances. Finally, revenue-management systems, which are essentially methodologies or processes that are designed to identify and match demand to an airline s available capacity, are the vehicle for maximizing value both for travelers and airlines. They enable an airline to better serve customers by identifying schedule paths and fares that meet their service and pricing needs, and in doing so attract revenue to the airline. Revenue-management systems, therefore, affect and are affected by other core components of an airline and must continuously interact with those areas in a variety of ways to be effective. Here again, although this is a complex matter even for a single, independent airline, it is much more difficult for alliance partners. To be fully effective, they cannot simply apply their respective systems to the broader network, even 10 Brueckner, Jan K., International Airfares in the Age of Alliances: The effects of Codesharing and Antitrust Immunity, Department of Economics and Institute of Government and Public Affairs, University of Illinois at Urbana-Champaign 11 Whalen, W. Tom, Constrained Contracting and Quasi-Mergers: Price Effects of Code Sharing and Antitrust Immunity in International Airline Alliances (May 15, 2003). EAG Discussion Paper No Available at SSRN: or DOI: ssrn Gerchick-Murphy Associates LLC

116 42 Alliance Integration assuming successful integration of service and pricing policies, but must integrate their revenue-management systems as well, and it is highly unlikely that their respective revenue-management systems are compatible. In addition, they must have the same interaction with other core components of the alliance model as required by an independent airline, and these systems are key to maximizing the ability of the alliance to better serve customers. Absent immunity from the antitrust laws, alliances simply cannot fully coordinate these and many other types of critically important activities that independent airlines routinely perform in their efforts to increase the appeal of their product to consumers. Since multinational alliances are the only effective way to serve tens of thousands of city-pairs and millions of passengers, it does not make either any economic or public policy sense to deny them the ability to do so as effectively and efficiently as possible. For example, a revenue-sharing environment may give a carrier the necessary incentive to invest in adjustments to its facilities to accommodate a joint-venture partner if it can benefit from the revenue created by this effort. The implementation of Atlantic Plus gave United the necessary incentive to make the investments necessary to accommodate Lufthansa at Chicago O Hare s Terminal 1. The transatlantic traffic trends shown in Figures 2.2 and 2.3 are clearly consistent with a competitive market. Figures 2.4 and 2.5 provide additional evidence of a competitive market. Figure 2.4 compares non-fuel CASM for the network airlines in the Domestic and Atlantic entities Domestic CASM (non fuel) Atlantic CASM (non fuel) M JS D MJ S DM J SD M JS D MJ S DM J SD M JS D MJ S DM J SD M JS D MJ S DM J SD M JS D MJ S DM J SD M JS D MJ S DM J SD M JS Figure 2-4, Network Airline Non-fuel CASM in the Domestic and Atlantic Entities, Twelve Months Ended Data at Each Quarter Gerchick-Murphy Associates LLC

117 43 Alliance Integration Although these CASM calculations are based on nominal expenses, the Atlantic non-fuel CASM has nevertheless trended down since 1996, when United and Delta became the second and third network airlines to receive antitrust immunity. Note also the differences since 2006; non-fuel CASM is up in both entities, but in the domestic entity the CASM increase is twice that in the Atlantic entity. Figure 2.5 shows network airlines RASM and CASM in the Atlantic entity Unit Rates (in $) MJSDMJSDMJSDMJ SDMJSDMJSDMJSDMJSDMJSDMJSDMJ SDMJSDMJSDMJSDMJSDMJSDMJSDMJSDMJ S Months Ended Network RASM (less trans rel rev) Network CASM (less trans rel exp) (current) Figure 2-5, Network Airline Atlantic Entity RASM and CASM, Twelve Months Ended Data at Each Quarter RASM has exceed CASM since early 2004, but the difference reflects operating margins of only about 5 percent (except for the period in late 2006 and 2007 when fuel prices briefly declined). Moreover, subsequent to 2004, the increases in passenger revenues reflected in the RASM calculations do not compensate for the increases in fuel prices. The relatively steep increases in RASM that have occurred during the past five years do not reflect pricing issues, but are in line with increases in costs. One other factor to consider is that U.S. E.U. open skies did not become effective until early in This certainly increases the prospects for new competition to discipline price, at least once the economy recovers. The solution to any pricing issues that might evolve should not be governmental actions that would jeopardize the pro-consumer effects of alliances, but rather to encourage and rely upon other sources of increased competition. Gerchick-Murphy Associates LLC

118 44 Alliance Integration The continuing balance between cost and revenue trends in the transatlantic market is partly due to the inability of network airlines to increase price even on the nonstop overlap routes (pipeline routes) that connect their respective networks. These routes are typically the focus of antitrust concerns and may tie back to concerns about the very substantial fare premiums for time-sensitive passengers that led to the domestic high fares at hubs phenomenon during the 1990s. Certainly the marketplace and competition are typically considered to be the controlling factors influencing price in any city-pair market. But in the absence of competition, an important tactic used to raise price is to limit capacity. In a citypair where there is both an absence of competition and enough passengers who are willing to pay very high prices, airlines could limit capacity to maximize their ability to extract rents from such passengers. But if too many seats are made available, it is difficult to adequately segment the market. While reducing capacity in some domestic markets did allow price increases until the emergence of the now-dominant low-cost airline sector, there are five fundamental reasons why this strategy does not make economic sense on hubto-hub pipeline routes, which attract most of the antitrust scrutiny. First, a much smaller proportion of transatlantic traffic is local traffic versus connecting or flow traffic. As a result, the proportion of time-sensitive local traffic to total traffic is substantially lower in transatlantic markets. It is therefore most unlikely that the profit maximizing strategy for an alliance carrier would be to limit capacity for the purpose of increasing local prices to time-sensitive passengers, rather than increasing capacity in order to increase connecting revenues. On the very few pipeline routes that have large amounts of local traffic, the existence of other competition also prevents unilateral reductions in capacity by the alliance partners. Second, reducing pipeline capacity would be contrary to the network carriers ongoing efforts to restructure the balance of their domestic and international networks by moving capacity from the domestic market to where it can be more profitably operated. Increased international operations also increase global traffic flows over the domestic system in an effort to stabilize their system of networks. Third, as immunized alliances continue to integrate, they will improve the effectiveness of connecting services, both in terms of convenience and price. This will increase the appeal of connecting services to time-sensitive passengers, particularly in markets with minimal frequency, but in other markets as well. Improved connecting services will provide incentives for alliances to use price to attract time-sensitive passengers from nonstop providers, particularly if the nonstop carriers attempt to increase price for such passengers. Gerchick-Murphy Associates LLC

119 45 Alliance Integration Fourth, the connecting market sector that can only be serviced with joint service remains a prime growth area that is not likely to be sacrificed in order to focus on small numbers of local time-sensitive passengers. Despite the alliance growth that has occurred in recent years, international connecting markets are still the most underserved sector. With recent transatlantic deregulation, new capacity by existing airlines, new airlines, and new business models will eventually materialize, particularly in larger markets. Therefore, small and faster growing connecting markets will become even more important for network airlines because those are the markets that are less likely to attract new competition. Finally, the robust traffic and capacity growth illustrated for the transatlantic market are clear evidence of the choice the network airlines have made. In an effort to stabilize their financial viability they have unanimously chosen growth rather than higher prices for limited numbers of passengers. Although continued international expansion clearly is important for the network airlines, there are no guarantees that this will occur even if they retain immunity for their international alliances. The deregulation of air service between the U.S and the E.U. will almost certainly change the competitive dynamic in the U.S. network airlines largest and fastest growing international market. They are already confronted with stronger, faster growing, and better financed foreign competitors than in the past. These competitors are now far more entrepreneurial and market oriented. Furthermore, new business models are evolving overseas and the varieties of innovation and competition that deregulation has bred domestically are likely to come into play in international markets. Historical perspective strongly suggests that the resulting change in the competitive landscape is likely to be extensive and in forms not anticipated. Multinational alliance development has created an alternative to consolidation for airlines that is likely to become more important as international deregulation continues to unfold. The ultimate effect this has on domestic consolidation on different continents around the globe will, to a great extent, depend upon how effectively multinational alliances are able to tap into the massive connecting market segments that can only be served by a well-integrated network. Gerchick-Murphy Associates LLC

120 EXHIBIT 4

121 Analysis of Europe-US connecting markets RBB Economics, 03 September 2009 RBB Economics Page 1

122 1. Introduction This note presents the results of our analysis of transatlantic connecting markets. The reason for analysing passenger traffic on connecting markets is to assess the extent to which being part of an integrated alliance confers a competitive advantage. In principle, members of an integrated alliance will not only coordinate their fares, leading to the elimination of the doublemarginalization, but they will also jointly optimise their scheduling and develop seamless connections. All else being equal, integrated alliance members will offer a more attractive product to passengers on these connecting markets than a non-integrated alliance, such as oneworld, or non-allied carriers. First, we analyse traffic data on connecting markets between 30 European countries and the US. Second, we also estimate the size of transfer traffic at key European hub airports. The analysis is based on MIDT data provided by American Airlines (AA). In summary the results show: Transatlantic connecting markets transport about 11.4 million bookings (YE March 2009), or 43% of all passenger traffic between Europe and the US. This high share indicates that connecting flights play an important role in long-haul passenger air traffic between Europe and the US; STAR and SkyTeam each have a much larger share of overall bookings on connecting markets between the 30 European countries and the US compared to oneworld. We estimate the SkyTeam and STAR shares to be 42% and 38% respectively. In contrast oneworld s share is 16%; For this analysis Continental was treated as a member of SkyTeam (which it was throughout the period covered by the data). However, Continental will soon be joining STAR. If instead we assume that Continental is part of STAR, that alliance s pro forma share of all bookings increases to 46%, whilst SkyTeam s falls to 34% (although conceivably their shares will exceed those levels as they reorganise their schedules and maximise connection opportunities). Whether CO belongs to STAR or SkyTeam does not improve oneworld s pro forma share, which is still well below that of the two other alliances; A more granular analysis of the booking data on connecting markets between the US and Spain (Iberia s home market ) on the one hand and the US and the UK (BA s home market ) on the other hand shows that oneworld has a smaller share of bookings than SkyTeam in each case (33% vs. 41% and 24% vs. 46% respectively); STAR and SkyTeam are the leading alliances on many connecting markets between Europe and the US. For example, our analysis shows that SkyTeam is the leading RBB Economics Page 2

123 provider in more than 3,400 connecting markets that transport in total more than 1.7 million passengers a year; We estimate that post-alliance, the elimination of the double-marginalisation should enable oneworld to contest the leading position of STAR and SkyTeam in between 150 and more than 600 connecting markets between the European countries and the US; oneworld has a relatively weak market position in thousands of transatlantic connecting markets (based on a share of bookings less than 15%). In many of these markets oneworld already interlines, which means that following the implementation of the JBA its service offer on these markets will be more competitive for the benefit of millions of transatlantic passengers; The STAR and SkyTeam hub airports in Europe, namely FRA (Frankfurt), AMS (Schiphol) and CDG (Paris Charles De Gaulle), together accommodate far more transfer passengers on the selected Europe-US connecting markets involved in this analysis than LHR (London Heathrow). They record 3.4 million transfer passengers versus 1.1 million at LHR; Analysing traffic on these connecting markets shows that FRA and AMS see individually more transfer passengers from their national airline alliances than LHR; Unlike FRA, AMS and CDG (where transfer passengers are essentially transported across the Atlantic by one single alliance - either STAR or SkyTeam), at LHR a significant portion of transfer passengers (14%) relies on STAR members to fly across the Atlantic. We find that oneworld is in a much weaker position on transatlantic connecting markets relative to the two other international alliances. SkyTeam and STAR on the other hand appear far more successful in attracting passenger flow across the Atlantic. The proposed JBA will likely turn oneworld into a more effective competitor, by removing double marginalisation and raising the quality of service offered by its members. Lower prices and better service will bring benefits to millions of connecting passengers. RBB Economics Page 3

124 2. Analysis of Europe-US connecting markets The dataset is based on MIDT data covering the period April 2008 to March 2009 and include all bookings between the US and 30 European countries. The European countries correspond to the European Economic Area (EEA) countries and Switzerland. 1 The data also includes information on airport sequence, operating and marketing carrier for each leg Connecting markets Connecting markets are defined as city-pair markets without a nonstop service. That is, passengers flying on any city-pair markets that are included in this analysis will have to make at least one stop to reach their destination. For example, the city-pair Newcastle Chicago is a connecting market. This is because there is no direct flight on this market. Below we provide some examples of the various possibilities of reaching Chicago from Newcastle. 1 The EEA countries are Austria, Belgium, Bulgaria, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Norway, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden and the United Kingdom. RBB Economics Page 4

125 Origin NCL City pair: Newcastle (NCL) Chicago (ORD) No Direct Flights Destination ORD One connection in Amsterdam (AMS), two segments NCL AMS ORD Two connections in Paris (CDG) and New York (JFK), three segments NCL CDG JFK ORD These examples show: There are no direct flights between Newcastle and Chicago. The only way to fly from Newcastle to Chicago is via connecting airports; To fly from Newcastle to Chicago it is possible to connect in Amsterdam. This is a twosegment flight, the first segment from Newcastle to Amsterdam, and the second segment from Amsterdam to Chicago; To fly from Newcastle to Chicago it is also possible to connect in Paris and in New York. This is a three segment flight. The dataset used for this report contains all city-pairs where directional travel involves four segments or less. 2 The reason our analysis focuses on passenger traffic of connecting markets is that any integrated airline alliance member should have a clear competitive advantage relative to other carriers. This is because the price and non-price aspects of the offer will be close to that of an on-line journey rather than an inter-line journey. This implies that all else equal, members of SkyTeam and STAR would in principle provide air transport service for a lower fare and a better quality of service on these connecting markets. 2 We excluded O&Ds where the connecting airport was in the Middle East, Russia or Africa. These markets involve a very small number of bookings. RBB Economics Page 5

126 2.2. Overview of Connecting markets Figure 1 below presents the number of city-pair markets in each of the 15 largest European countries for connecting routes to the US. A country-pair is just a simple grouping of city-pair markets between the US and one European country. For each country-pair we present the number of single-connect markets, i.e. there is at least one airline/alliance that offers a service with only one stop, and the number of double-connect markets, where passengers have to connect twice or more. Overall, the dataset contains about 36,500 connecting markets between the 30 European countries and the US, and more than 27,000 of these are double-connect markets. Figure 1: Number of connecting city-pair markets by country-pair 4,500 Number of connecting city pair markets by country 4,000 3,500 3,000 2,500 2,000 1,500 1, Number of single connect markets Number of double connect markets Note: Country-pairs are ranked by total flight bookings on connecting markets. Only the 15 largest country-pairs are shown. Figure 1 shows: The country-pair Germany US has the largest number of connecting markets. For this country-pair the dataset records 4,135 connecting markets, of which 1,357 are singleconnect, i.e. there is at least one airline/alliance that offers a service with only one stop, and 2,778 markets where passengers have to connect twice or more; RBB Economics Page 6

127 Similarly for the other country pairs, the number of double-connect markets is larger than the number of single-connect markets. Not surprisingly it is possible to reach more destinations with a two-stop journey compared to a one stop. For each country-pair Table 1 below presents the number of bookings for all connecting markets and that for thin connecting markets. We define a thin market as a city-pair within which fewer than 5,000 passengers a year travel. Table 1: Number of bookings for connecting markets Country pairs Total bookings Total bookings for thin markets (bookings<=5,000) Germany US 2,058,217 1,013,288 United Kingdom US 1,466, ,637 Italy US 1,454, ,271 Spain US 925, ,598 France US 923, ,282 Switzerland US 488, ,485 Sweden US 428, ,387 Denmark US 394, ,315 Greece US 365, ,427 Belgium US 346, ,555 Ireland US 324, ,278 Norway US 295, ,780 Netherlands US 287, ,829 Poland US 241, ,705 Austria US 234, ,082 All Europe US 11,423,128 5,521,503 Note: Country-pairs are ranked by total flight bookings on connecting markets. Only the 15 largest countrypairs are shown. The main results of this analysis are: In total there are 11.4 million passengers travelling on connecting markets between the 30 European countries and the US, and the thin markets have seen close to half of the total number of passengers; The five EU countries with the largest domestic population account for nearly 7 million passengers a year on these connecting markets. 3 To highlight the significance of passenger traffic on connecting market, our data analysis shows that 15 million passengers travel on direct markets between Europe and the US. This means that bookings on connecting markets represent 43% of overall transatlantic traffic. Table 2 below shows the number of direct markets between each European country and the US. As expected the number of these markets is much lower than the number of connecting 3 These countries are Germany, the UK, France, Italy and Spain. RBB Economics Page 7

128 markets. The analysis shows that 8 European countries 4 do not have a direct service to the US, and do not appear in the table below. In addition, for a number of other European countries the number of markets served by a direct service to the US is smaller than 5, and countries like Norway, Hungary, Latvia and Romania have just 1 direct service to the US. This means that many transatlantic passengers need to fly indirect. Table 2: Number of direct markets and total bookings for direct markets Country pairs Number of direct markets Total bookings Germany US 44 2,369,872 UK US 40 5,250,242 Norway US 1 37,833 Spain US ,575 Sweden US 4 148,854 Denmark US 5 176,608 Belgium US 5 308,752 Italy US 22 1,257,967 France US 19 1,746,374 Netherlands US 19 1,028,376 Austria US 4 126,432 Czech US 2 59,885 Finland US 2 60,312 Greece US 3 173,926 Hungary US 1 62,582 Iceland US 5 29,238 Ireland US ,107 Latvia US 1 10,278 Poland US 5 303,917 Portugal US 7 167,006 Romania US 1 32,731 Switzerland US 9 494,161 All Europe US ,070,028 Note: Only the 22 European countries with direct connections to the US are shown. 4 The eight European countries with no direct connection to the US are: Bulgaria, Cyprus, Estonia, Lithuania, Luxembourg, Malta, Slovakia, and Slovenia. RBB Economics Page 8

129 2.3. Competitive analysis of transatlantic connecting markets In this section we look at the relative strength of each airline alliance on connecting markets between Europe and the US Analysis of overall traffic Figure 2 below presents the share of passenger traffic on connecting markets between the 30 European countries and the US. Figure 2: Share of bookings for Europe to US connecting markets Non Allied, 4% oneworld, 16% SkyTeam, 42% Star Alliance, 38% This analysis reveals that: SkyTeam and STAR are the two leading alliances with shares of 42% and 38% respectively; oneworld has only 16% of these bookings, which is due in part to the fact that it a less integrated alliance than STAR and SkyTeam. RBB Economics Page 9