33; ‘.
.4‘,

'.§;,,

 

.III.

n
,. px #1:...3}.

 

THE“

2 004
5;: .3

This is to certify that the
dissertation entitled

A Merger By Any Other Name? Empirical Evidence of the
Anticompetitive Effects of Domestic Airline Alliances

presented by

Aisha Rafiqui-Masroor

has been accepted towards fulfillment
of the requirements for the

PhD. degree in Economics

 

 

/W 5-5.,“

Major Professor’s Signature

”Avg/1Q w;

 

Date

MSU is an Affirmative Action/Equal Opportunity Institution

 

 

 

 

LIBRARY
Mlchigan State
University

 

 

 

PLACE IN RETURN Box to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/ClRC/Date0ue.p65-p. 15

 

A MERGER BY ANY OTHER NAME? EMPIRICAL EVIDENCE
OF THE ANT I-COMPETITIV E EFFECTS OF DOMESTIC AIRLINE ALLIANCES

By

Aisha Rafiqui-Masroor

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Department of Economics

2003

 

 

A MI
ANT

In
that were
airline in
regulator
Carriers i
second b
far being

POWEI' a1

ABSTRACT
A MERGER BY ANY OTHER NAME? EMPIRICAL EVIDENCE OF THE
ANTICOMPETITIVE EFFECTS OF DOMESTIC AIRLINE ALLIANCES
By

Aisha Rafiqui-Masroor

In this paper we empirically investigate the behavior of the three airline alliances
that were announced in early 1998 among the six most dominant carriers in the domestic
airline industry. These alliances were announced at a time when antitrust law and
regulatory concern made mergers between large and financially successful domestic
carriers impossible. A natural question was therefore whether these alliances were a
second best strategy: had they allowed these carriers to realize cost synergies that were so
far being forfeited by expected antitrust intervention, or had they been a source of market
power and served as a loophole in current antitrust law?

Starting with the premise that the motivation for the formation of these alliances
would be reflected in their long-term fare behavior, strong evidence of the exercise of
market power was found in all three cases. Though the timing of the fare responses
highlighted some differences between the individual alliances, the uniformity of results
supports the generalization of arguments against each of these arrangements.

This paper also presents strong evidence that airport dominance (a familiar source

of market power in the airline industry) was greatly enhanced due to the formation of

 

these
that t]
was c
Airwa

more i

these alliances. Finally, division of the full sample into important sub-samples revealed
that the increased market power of the Delta-United and Northwest-Continental alliances
was derived from a reduction in the travel options available to travelers. For the US
Airways-American alliance, demand complementarities or multi-market contact were

more important for the realization of increased market power.

 

Idlfl
Tmmuxt
Dr. Borer
lhmk>t
innnense

AIM
DLBCy
continue
Dnhhfl
ifinnfil
DLINC
DI-D:

early i

Dr. M

Your

the ‘

ACKNOWLEDGEMENTS

I’d like to thank the staff at the Department of Justice and the Department of
Transportation who led me to the data when I first started work on this dissertation.
Dr. Borenstein shared a data set that was used in a preliminary version of this paper.
Thank you. Kathleen Citro at the Embry Riddle Aeronautical University was also
immensely helpful in this regard.

At Michigan State, I would like to thank:

Dr. Boyer, thank you for being upbeat and encouraging. Your support helped me to
continue on a difficult path.

Dr. Mullin, thanks for fulfilling your commitment to this dissertation and for keeping
it in mind.

Dr. Wooldridge, thanks for teaching micro-econometrics and for your input.

Dr. Davidson, your no-nonsense teaching style converted me to microeconomics
early in the program, thank you.

Dr. Matusz, thanks for being a ‘risk-taker’.

At home, I’d like to thank my husband Saqib. This dissertation is a reflection of
your commitment. Every time I asked, “stop?” you said, “go”. You saw the light at
the end of the tunnel through incredibly hard years. Thank you. This is for you.

Faraz and Kiran: Thanks for waiting.

iv

TABLE OF CONTENTS

List of Tables ..................................................................................... vii
Chapter 1.
The Evolution of the Airline Industry: From Mergers to Alliances .................. l
1.1 The Post Deregulation Years ........................................................................................ 1
1.2 The Introduction of Frequent Flyer Programs .............................................................. 3
1.3 The Need for International Airline Alliances ............................................................... 5
1.4 The Formation of Domestic Airline Alliances .............................................................. 7
1.5 The Question of Interest and the Need for this Research ........................................... 12
1.6 Layout of Paper ........................................................................................................... 14
Chapter 2.
The Structural Model and Empirical Review ....................................................... 16
2.1 The Structural Model of Equilibrium ......................................................................... 16
2.2 Efficiency versus Market Power- The Primary Hypothesis ....................................... 20
2.3 Previous Empirical Evidence ...................................................................................... 22
2.3.1 Airline Mergers and their Impact on Fares ....................................................... 22
2.3.2 The Fare Impact of Airport and/or Route Dominance ...................................... 24
2.3.3 Frequent F lyer Programs ................................................................................... 25
2.3.4 International Airline Alliances .......................................................................... 27
2.3.5 Domestic Airline Alliances ............................................................................... 29
2.4 Synopsis and Conclusion ............................................................................................ 30
Chapter 3.
Alliance Formation: Efficiency versus Market Power ....................................... 32
3.1 The Price Equation ...................................................................................................... 32
3.2 Explanatory Variables and Expected Signs ................................................................ 34
3.3 Data: Source and Description ..................................................................................... 37
3.4 Market Definition ........................................................................................................ 38
3.5 Treatment and Control Firms ...................................................................................... 39
3.6 Airports Considered .................................................................................................... 40
3.7 Time Period Selection ................................................................................................. 41
3.7.1 The Timing of Fare Changes ............................................................................. 43
3.8 Evidence: Pooled OLS Estimates of the Impact of Alliance Formation on Fares ...... 44
3.9 Mitigating a Potential Econometric Problem ................................................ 48
3.10 Evidence: First Difference Estimates of the Impact of Alliance Formation on
Fares ............................................................................................ 52
3.11 Hub-Specific Evidence: First Difference Estimates of the Impact of Alliance
Formation on Fares ........................................................................... 55
3.12 Synopsis and Conclusion .......................................................................................... 63

 

Chapte
Alliant
4.1 The
4.2 £pr
4.3 Met]
4.4 Evid
Dorr

4.5 Som
4.5.2
4.5.2
4.5.3

4.6 Preli;
Forrr.

4.7 The l
4.8 Evidc
Airpc
4.9 Syno“

Chapter
Sub-Sar
5.1 The P
5.2 H)I)Ol

5.2.1 }

5.2.2 1
5.3 Methc
5.4 Sub-S

I:Ol'ma
5.5 SFHOp

Chapter 1

Apmndu
Appendi:
Appendi;

Chapter 4.

Alliance Formation and Airport Dominance ......................................................... 65
4.1 The Fare Impact of Airport Dominance ...................................................................... 65
4.2 Explanatory Variables and Expected Signs ................................................................ 67
4.3 Method of Inquiry ....................................................................................................... 69
4.4 Evidence: Pooled OLS Estimates of the Impact of Alliance Formation on Airport
Dominance .................................................................................................................. 72
4.5 Some Econometric Issues ........................................................................................... 74
4.5.1 Collinearity Between Explanatory Variables .................................................... 74
4.5.2 The Omitted Variables Problem ........................................................................ 75
4.5.3 Endogeneity of an Explanatory Variable .......................................................... 76
4.6 Preliminary Evidence: First Difference Estimates of the Impact of Alliance
Formation on Airport Dominance ............................................................................... 77
4.7 The Persistent Endogeneity of the Explanatory Variable ........................................... 80
4.8 Evidence: First Difference-2SLS Estimates of the Impact of Alliance Formation on
Airport Dominance ..................................................................................................... 81
4.9 Synopsis and Conclusion ............................................................................................ 90
Chapter 5.
Sub-Sample Analysis: Disentangling the Market Power ................................... 91
5.1 The Price Equation ...................................................................................................... 91
5.2 Hypotheses Tests on Route Sub-Samples ................................................................... 92
5.2.1 Efficiency versus Market Power ........................................................................ 92
5.2.2 Size versus Concentration .................................................................................. 94
5.3 Method of Estimation ................................................................................................. 94
5.4 Sub—Sample Evidence: First Difference Estimates of the Impact of Alliance
Formation on Fares ..................................................................................................... 95
5.5 Synopsis and Conclusion .......................................................................................... 101
Chapter 6.
Summary and Conclusions ...................................................................................... 103
Appendices .................................................................................................................. 108
Appendix A: Data Base Construction. ........................................................................ 109
Appendix B: The 45 Busiest Airports .......................................................................... 110
Appendix C: Gate Controlled and Slot Constrained Airports. .................................... 113
Appendix D: Cities Served by More Than One Airport .............................................. 114
Appendix E: Hub Specific Pooled OLS Estimates of log (fare in) ............................. 115
Appendix F: Pooled OLS Estimates of log (AM in), Route Market Share Excluded...121
Appendix G: Variable Descriptive Statistics by Sub-Sample .............................. 122
Appendix H: Pooled OLS Estimates of log (fare in), for Route Sub-Sample ............ 125

Appendix I: First Difference-ZSLS Estimates of log (AMm) for Route Sub- Samplele8

Bibliography ................................................................................................................ 13 1

vi

 

Table
Table
Table
Table .
Table 1
Table 4
Table 4
Table 4
Table 4

Table 4.
Table 4.

Table 5.

Append

Append.

Appendi
Appendi

API-‘end i

Table 3.1:
Table 3.2:
Table 3.3:
Table 3.4:
Table 3.5:
Table 4.1:
Table 4.2:
Table 4.3:
Table 4.4:

Table 4.5:

Table 4.6:

Table 5.1:

Appendix E:

Appendix F:

Appendix G:

Appendix H:

Appendix I:

LIST OF TABLES
The Sample of Airline Alliances ................................................ 43
Variable Definitions ............................................................... 45
Pooled OLS Estimates of log (farem) ........................................... 46
First Difference Estimates of log (farein) ....................................... 53
Hub Specific First Difference Estimates of log (farein) ...................... 56
Variable Definitions ............................................................... 70
Variable Descriptive Statistics ................................................... 71
Pooled OLS Estimates of log (AMm) .......................................... 73
First Difference Estimates of log (AMm) ...................................... 79
First Difference-ZSLS Estimates of log (AMin),
lag (MSm) as Instrument ......................................................... 82
Hub Specific First Difference-ZSLS Estimates of
log (AMin), with lag (MSm) as Instrument .................................... 84
First Difference Estimates of log (farem), for Route
Sub-Samples ....................................................................... 96
Hub Specific Pooled OLS Estimates of log (fare in) ........................ 115
Pooled OLS Estimates of log (AM in), Route Market Share
Excluded ........................................................................... 121
Variable Descriptive Statistics by Sub-Sample .............................. 122
Pooled OLS Estimates of log (fare in), for Route Sub-Sample ............ 125

First Difl‘erence-ZSLS Estimates of log (AMm) for Route Sub-Sample3128

vii

 

Chap

airline
the cor
discuss

carriers

Chapter 1. The Evolution of the Airline Industry: From Mergers to Alliances

This chapter briefly traces out the events taking place in the post deregulation
airline industry that played major roles in the state of competition there. Beginning with
the competitive impact of deregulation of this industry in 1978, this chapter ends with a
discussion of the industry's current characteristic of increasing cooperation between

carriers.

[.1 The Post Deregulation Years

The Airline Deregulation Act of 1978 deregulated the US. airline industry
through the adoption of a gradual system of deregulation. Part of this process was the
governance of the industry by the Department of Transportation (DOT) fi'om 1985 to
1988. During this period, the DoT assuming contestability,‘ permitted many airline
mergers.2

A merger wave in this industry thus took place in the mid 1980's as larger carriers
began to acquire smaller ones, especially those that could provide substantial feed traffic
to their designated hubs.3 This enabled the larger carriers to extend their network and

scale of operations. Many airlines abandoned smaller and more competitive markets in

favor of consolidating their operations into regional hubs and city-pair markets where

 

' Contestability of airline markets has since been rejected. See for instance Hurdle, et. al (1989), Hurdle
(1989), Levine (1987), Call and Keeler (1985) and Graham, Kaplan and Sibley (1983).

2 Most mergers took place in 1985-86. In 1986 alone, 14 airline mergers were approved by the Department
of Transportation. Some of these mergers were justified under the failing firm clause of the Merger
Guidelines. Clougherty (2002) explains the favorable antitrust review received by some cross-national
domestic airline mergers with that regulators may have considered their international competitive effects.

they could function as oligopolies, or even monopolies. This is not to claim that this was
their prime motive: cost efficiencies possible through economies of density in a hub
based network arrangement have been documented.4

Deregulation not only gave fieedom of entry, but also of exit: along with an
increase in acquisitions through mergers in the period immediately after deregulation,
there were a large number of bankruptcies as well.5 This widespread and rapid entry and
abandonment following deregulation was not checked by the Department of
Transportation and it helped carriers realize dominance at their designated hubs, resulting
in the creation of fortress hubs. There is evidence that during this period of increasing
consolidation of the industry, gate constraints and slot controls,6 already genuine physical
constraints at some of the most important airports in the country, began to be used as
tools by the ‘dominant carriers’ to compete with rival carriers.7 Dominant carriers were
able to provide better flight frequency and/or flight convenience and inhibit other firms
from obtaining landing slots, thereby impeding their entry and expansion in major

markets.8

 

3 Large hub airports me those (as defined in the us. Code) with at least 1% of total annual passenger
enplanements. These hubs are not necessarily the same as the hubs designated by carriers. According to
the code, thirty one airports qualify as hubs, while twenty one of these are airline-designated hubs.
‘ See for instance, Brueckner, Dyer and Spiller (1992) and Caves, Christensen and Tretheway ( 1984).
5 The main reason for firm collapse would be its failure to adopt appropriate yield management techniques
that allowed it to collect fares specific to the elasticity of the consumer. That is, uniform pricing practices
prevented many carriers from achieving minimum load factors.

A gate is a physical asset at an airport. A slot refers to the right to land or takeoff from an airport at a
certain time of the day.
7 A carrier is ‘dominant’ if its market share at an airport is greater titan 50%. See US. General Accounting
Office, GAO/01-518T.
' See us. General Accounting Office, GAO/01-518T.

1.2 The Introduction of Frequent F lyer Programs

The introduction of loyalty inducing programs is an important aspect of the post
deregulation evolution of the airline industry. The Frequent F lyer Program was
introduced by American Airlines in May 1981.9 Under the AAa‘vantage program,
members accumulated mileage and could redeem credit either though free/discounted
travel or service upgrades. A week later, United Airlines responded by announcing its
own Frequent F lyer Program. Within the next six months, six other carriers had joined in
with similar programs of their own and competition had begun (especially among the
major cm'riers) to offer the most lucrative Frequent Flyer Program.

What followed ranged from credits for free travel to promotional tie-ins with car
rental agencies, hotels, cruise liners and credit cards. The Frequent Flyer Program
therefore became an important tool for an airline to compete with rival carriers and to
secure the demand of its most lucrative travelers.

The 'Principal-Agent Problem' is the basis for the main criticisms of these
programs. That is, these programs can induce choice distortions by the traveler if the
person accumulating and receiving benefits due to firm choice (the agent) is not the
entity paying the airfare (the principal). 1° The outcome is that the employee makes
decisions to maximizes his/her own travel benefits rather than the profits of the firm
paying the air fare.

In the airline industry, the Principal-Agent problem surfaces due to a number of
reasons. First, the complexity of the award structure makes it difficult for the principal

to categorically ascertain whether the agent has indeed made an inefficient carrier choice

 

9 Aviation Week & Space Technology, "American Establishes Travel Bonus Program" Page 41, May 18
1981.

 

 

or to re:
would I!

5

their trai
Frequent

a compa
carriers L'
smaller r.
Frequent
Created a:
and also 5
Ar
between 1
means (h;
only allov
also Provi
1h.-
FreCluent 1:
potency of
their (”Wei

andFOXn!

\

it See Wall St.

FQr l
m

9&102 ore on

or to reclaim these awards from the agent. Of course if this were possible, the program
would lose its effect on carrier choice.ll

Second, the non-linearity of the award system encourages members to concentrate
their travel with a single firm, even when this choice is sub—optimal. Third, a carrier's
Frequent Flyer Programs is not only able to discourage firm entry (unless the entrant has
a comparable route structure), it can also discourage a fare challenge from other rival
carriers that already serve in many of the same markets. Fare competition initiated by a
smaller rival carrier may not serve its purpose if, in order to avoid loss of their awards,
Frequent Flyer Program members do not make a firm-choice switch. Switching costs are
created as benefits are awarded only once a certain mileage threshold has been reached
and also since after award redemption, the member is awarded some initial bonus miles.‘2

And finally, these programs can also encourage the agent to take indirect routings
between trip origin and destination, as credit is given for each 'segment' traveled.l3 This
means that the size of a carrier's network is of great importance: a wider network not
only allows Frequent Flyer Program members more opportunities to accumulate credit, it
also provides them with more destinations at which to redeem travel awards.

There is evidence that the above mentioned distortions were induced by the
Frequent Flyer Program. For instance, a survey by Toh and Yu (1988) confirmed the
potency of these programs: they found that program members believed in concentrating
their travel with one program to maximize their travel reward accumulation. Stephenson

and F ox (1987) found that travel managers were concerned not only with the inefficient

 

'° See Wall Street Journal, "Greed Gets Most Mileage Out of Airline Credits" October 10, I985.

" See Levine (1987).

'2 For more on the features of Frequent Flyer Programs, see US. General Accounting Office, GAO IRCED—
90402.

carrier choices made by their traveling employees, but that these programs encouraged
travelers to adopt longer itineraries to accumulate mileage, increasing their travel time
and company expenses. The single most important management concern though was the

effect of these programs on airfares.

I. 3 The Need for International Airline Alliances

The first international airline alliance was formed in 1986 between Air Florida
and British Island. And then in 1987, British Airways and United Airlines proposed a
code-sharing alliance that was later exempted fi'om Department of Transportation
scrutiny. This set the stage for changes in bilateral agreements between countries and
also paved the way for antitrust immunity for such agreements.

The 1990's are characterized as a decade of increasing cooperation among
carriers, on both the international and domestic fi'ont. While cooperation had been a
feature of the industry since its inception, most was technical in nature, that is, in the
exchange, leasing and pooling of aircraft and aircraft parts, and in the maintenance of
aircraft and engines. Interlining agreements had also developed due to the tight
regulatory framework that confined carriers to operate within the boundaries of their
state.

The 1990's were also a decade of rapid growth in demand for air travel between
the United States and the rest of the world, particularly Europe, creating the need for
carriers to find ways improve their global network strengths and also to remain

competitive at home. However, restrictive bilateral air service agreements continued to

 

‘3 A 'segment' is defined as travel under a single flight number, so both direct and non-stop flights are

exist, creating immense legal, political and institutional constraints on mergers of carriers
of different countries. Route authority restrictions prevented a single carrier to serve
more than a handful of major international destinations. The response from major
domestic carriers was the formation of strategic alliances with international partners that
would allow them to expand in these markets.M

These major international alliances enjoyed substantial legal exemptions. For
instance, some of the big international alliances enjoyed the freedom to set fares, to
coordinate schedules, to cooperate on revenue pooling and on marketing. Effectively,
these arrangements allowed two carriers to function as a single entity while enjoying low
commitment burdens. An alliance between United Airlines and Air Canada and one
between American Airlines and Canadian Air, enjoyed such antitrust immunity.

Once a carrier formed an alliance with its international counterpart, the firms
aimed to project themselves as a single entity.15 The strategy was to provide travelers
with the sense and experience of 'seamless' service, whereby the passenger feels no
difference in firm identity and service between trip origin and destination.

It would be in this decade that the Frequent Flyer Program gained even more

importance as a device that would allow major carriers to stay ahead of their competitors.

 

single segment flights while connecting flights provide the opportunity to accumulate more credit.

” These international alliances enjoyed antitrust immunity. For instance, the alliance between Northwest
Airlines and KLM and between United Airlines and Lufthansa enjoyed antitrust immunity.

'5 Advertising and program promotion was done extensively with an alliance logo instead of individual
ones. For instance, KLM and Northwest developed the [QM-Northwest World Wide Reliability logo,
which incorporated the logos of both partners.

1. 4 The Formation of Domestic Airline Alliances

The successes realized upon formation of international alliances were followed by
domestic carriers pursuing similar arrangements at home. In early 1998, management at
Northwest Airlines and Continental Airlines announced that a link-up was being planned:
an extensive and integrated alliance, with the member firms code-sharing, merging their
Frequent F lyer Programs, joining use of their lounge facilities and swapping equity.

Thus in late 1998, Northwest Airlines purchased a majority stake in Continental
Airlines, despite a Department of Justice lawsuit challenging the acquisition."5 This
alliance was to involve Northwest purchasing 51% of the voting stock of Continental
Airlines. But due to objections by the Department of Justice, the plan was scaled back to
a 46% stock acquisition. However, members of both Northwest and Continental's
Frequent Flyer Programs could claim award flights on the other carrier's system, both
domestic and international. Code-sharing, (that is the practice of using an airline’s two-
letter code on another airline’s flight) was permitted on limited (non-hub) routes.l7

A broad marketing alliance was soon announced by American Airlines and US.
Airways and this was to include a joining of their Frequent F lyer Programs and code-
sharing between their regional partners. Frequent F lyer Program members could
combine their (domestic) travel miles with those earned on their partner's flights, though
mileage earned on the partner's flights would not qualify transfer. This alliance was
announced in April 1998. By September of the same year, Delta Airlines and United

Airlines had embarked on similar plans of their own, allowing their Frequent F lyer

 

'6 The Department of Justice did not seek a temporary injunction against the transfer of voting control to
Northwest Airlines. The lawsuit was dismissed in January 2001 when Northwest agreed to divest all by
7% of its voting interest in Continental Airlines.

'7 This alliance was examined as a hill merger and the 5% market share increase provision in the Merger
Guidelines was used to limit the routes on which the two carriers could code-share.

Program members to claim award flights for domestic travel on the other partner's
system. Both alliances would grant joint access to airport lounges. ‘8

The institutional setting of domestic alliances was to be significantly different
fiom that of international alliances. For one, antitrust immunity was not expected for
domestic partnerships, though initially the level of integration planned had been
significant. While the main motivation for international alliance formation had been to
gain entry and to bypass ownership barriers, such constraints did not exist in domestic
markets.

A diverse area of business was planned for domestic airline alliances. For
instance, it was planned that it may include joint sales and marketing, joint purchasing
and insurance, joint passenger and cargo flights, code-sharing,19 block-spacing,20 links
between Frequent Flier Programs, management contracts, and joint ventures in catering,
ground handling and aircraft maintenance.21

While an inherent feature of domestic alliances was the low commitment pressure
on the member firms, these carriers placed emphasis on their longevity. Anecdotal
evidence suggests that partner firms recognized alliances as being an important part of
firm dynamics and an effective means of gaining a competitive edge over rivals. Thus,

while the initial emphasis of contemporary alliances had been on marketing, eventually

they seemed increasingly strategic in nature.

 

" Both us Airways-American and Delta-United dropped their plans to code-share after the attention
received by the Northwest-Continental code-sharing agreement.

'9 Code-sharing is the practice of using of an airline’s two-letter code on another airline’s flight. It allows
a carrier to expand its network without substantial costs.

2° Block-spacing is the purchase and marketing of a certain number of seats on another airline’s flight.

2' A joint venture is a separate and independent organization set up by the partners to carry out specific
tasks.

r‘

While the discussion above points to the difference in both motivation and scope
between domestic and international alliances, the various motivations for alliance
formation between domestic carriers, were only seemingly diverse.22 In each, the basic
elements of the desire for firms to achieve either improved efficiency or greater market
power, the two competing results possible due to alliance formation, can be identified.

First, given that the current regulatory climate toward domestic airline mergers is
far different from what it was immediately following deregulation, it may have served as
an important impetus to alliance formation between domestic carriers.23 That is, while a
merger between large and financially successful domestic carriers would not be feasible
under current antitrust laws, domestic alliance formation among the country's largest
carriers may have been a second best strategy for carriers who seek either cost synergies
or market dominance. If either of these effects is realized, the carrier is able to achieve
significant advantages over its rival carriers.24 Rival firms, especially those that do not
serve a comparable number of important markets, will be unable to compete or at the
very least, will be placed at a significant disadvantage.25 Also, alliance features such as

code-sharing and mergers of Frequent F lyer Programs can play an important role in

 

22 For instance, management at Delta Airlines and Continental Airlines have expressed that these alliances
were formed in response to a need to increase shareholder value and to increase convenience for the
consumers through the offer of a wider network, greater frequency etc. See The Avmark Aviation
Economist, "Virtual Mergers Regulatory Headaches", May 1998.

23 While domestic carriers are actively restrained fi'om an outright merger with another domestic carrier,
the United States Congress has authorized the Department of Transportation to impose a 30 day waiting
period (extendable to 150 days for code-sharing alliances) before certain joint venture agreements,
(including Frequent Flyer Program links) are finalized. The Department of Transportation also "has the
authority to prohibit airline practices as unfair methods of competition if they violate antitrust principles,
even if the practices do not constitute monopolization and attempted monopolization under the Sherman
Act.” For more on the designation of authority between the Department of Justice and the Department of
Transportation and other aspects of federal oversight on the airline industry, see US. General Accounting
Office, GAO/Ol-518T.

2‘ See Boresntein (1989). Again, this advantage could be increased market power or improved efficiency.
2’ See for instance, US. General Accounting Office, GAO/RCED-90-102.

contributing toward passengers disregarding the distinct identity of the two partner
firms.26

Second, major domestic carriers may have pursued alliance formation since the
costs associated with an outright merger can easily eliminate any value that may have
resulted from the agreement. For instance, increasing costs that partner carriers could
face include customer service disruptions, costs of repainting the fleet, the costs of
severance packages as the new firm tries to eliminate redundancies, wage increases due
to any labor contract negotiations, etc. And this list of costs associated with an outright

7 Therefore a low commitment

merger is rift with unusually high complexities.2
agreement among parties that is afforded through an alliance may be more desirable.28

The 'strategic paralysis' that can be expected to take place after a merger is one
such complexity that merging firms need to contend with and that can have an impact on
firm profitability due to higher costs. It refers to a time period during which the two
merging firms are unable to stay on top of day to day strategic business decisions due to
the distraction of the merger. In an industry where such strategic moves are an important
part of their interaction with rival firms and where such a hiatus could easily be observed,
this set back can have consequences that the new merged firm may find hard to recover
fiom.

Third, and as noted earlier, competition between carrier's marketing programs

(such as their Frequent F lyer Programs) had been growing soon after their introduction in

 

2‘ Other marketing practices also contribute to providing the traveler with the sense of 'seamless' travel and
they include block spacing, franchising, schedule coordination, proximate placement of gates between
connecting flights, etc.

27 See Airline Business, McKinsey & Company, "Making Mergers Work.” Page 110-114, June 2000.

2’ Rhoades and Lush (1997) present the general conditions for stability and duration of alliances. They find
that less complex arrangements between carriers contribute favorably toward their stability and
maintenance.

10

1981 . We consider this as an important motivation for their eventual mergers.29 In fact,
soon after the introduction of the first Frequent F lyer Program in 1981, other airlines
operating jet propulsion aircraft began to offer similar programs, with a trend toward each
being competitively obligated to match the promotional offers of its major rivals.30

Fourth, there was a growing trend toward greater cost containment by the time the
three alliances were announced in early 1998. This had translated into the major carriers
having less control over their business travel segment, thus far a comer stone of airline
industry revenues. In fact, results of a survey by Bender and Stephenson (1998) of both
corporate travel managers and business travelers, support this. They found that cost
cutting on the corporate travel side was being achieved by ensuring that employee travel
was both necessary and economical and through the growing popularity of video
conferencing.

Fifth, the absence of sufficient economies of scale in the airline industry has also
been established in previous empirical literature, while evidence of the presence of
economies of scope has been presented.31 This is a second important cost based
motivation for alliance formation on the domestic fi'ont. These scope economies are
related to the size and structure of the partner carrier's networks. Also, Frequent F lyer
Programs (and therefore allied Frequent F lyer Programs) are more effective marketing
strategies when the partner carriers offer a large network over which points can be

collected and redeemed. Also, with respect to the realization of economies of density, as

 

29 For instance, in November 1987, Delta Airlines introduced a mileage plan under which 1988 mileage
would triple automatically. Most major carriers responded within weeks with similar incentives of their
own. In January 1990, a Frequent Flyer award war took place between United and American Airlines when
travel awards were to be distributed after only a few trips. Northwest, Continental, Delta, TWA followed
with their own generous reward programs. See Wall Street Journal, "War to Win Frequent F lyer Sizzles"
3t Asra Q. Nomani, January 26‘“, 1990.

See Hu, Toh and Strand (1988).

11

traffic volume increases on individual routes, higher load factors lead to lower costs.
This increased traffic density could allow the use of a larger aircraft that could be
operated at lower unit costs.

Thus, there is a diverse list of possible motivations that can explain and justify the
formation of alliances among major domestic carriers. However, the ex-ante
identification of the exact motivation for domestic alliance formation is not necessary
since in each, the basic elements of the desire for firms to achieve either improved

efficiency or greater market power, remain valid.

1. 5 The Question of Interest and the Need for this Research

Alliances between the six most dominant carriers in the domestic airline industry
were announced at a time when both academic and regulatory concern regarding the state
of competition in this industry, including the use of anticompetitive means by major
carriers to secure demand at major airports, was high.32 Since an outright merger
between large and financially stable carriers was impossible, a natural question was
whether these arrangements were a second best strategy to an outright merger. Had they
enabled these carriers to realize some cost synergies or was increased market power the

dominant effect?

 

3 ' See Caves, Christensen and Tretheway (1984).

32 While the Department of Justice did not file its complaint against the Northwest-Continental alliance
until October 1998, regulatory concerns regarding these alliances had been presented by the General
Accormting Office in written testimony before the US. Senate in June 1998. See US. General Accounting
Office, GAO/'1” -RCED-98-215.

12

Improved efficiency could have been realized perhaps through economies of

33 These carriers could have realized cost savings due

density and/or economies of scope.
to the former if the agreement allowed them better market access and/or traffic feed, and
due to the latter if the alliance partners re-configured their networks after alliance
formation.34

On the other hand, increased market power could also have been realized. For
each domestic alliance, the two marketing teams and their formerly individual marketing
programs, had been major rivals. Now the former may have found a platform to make
mutually beneficial strategic decisions and the latter may now be virtually
indistinguishable to an important and lucrative demand segment. The promise of pooled
program mileage may be sufficient to allow these firms increased control over this
demand segment and the 'elbow-room' to raise fares. 35 In such, these alliances may have
been a second best strategy to an outright merger between the carriers: they are
permissible arrangements that allowed member firms greater market power, while
maintaining the freedom that comes with a low commitment arrangement.

While these marketing alliances among the major domestic carriers received

considerable attention in the press and in regulatory circles, so far there has been no in-

depth empirical study of their competitive effects. Thus it remains to be seen whether

 

33 See Caves, Christensen and Tretheway (1984) on the importance of economies of density in the airline
industry.

3’ It was the extensiveness of the agreement between Northwest and Continental Airlines that was at the
heart of the belief in regulatory circles that network synergies may allow the realimtion of welfare
improvement.

3’ The three major domestic alliances had announced their plan to code-share, a plan later dropped by US
Airways-American as well as by Delta-United. Northwest and Continental Airlines had also responded to
Department of Justice concerns by dropping code-sharing on seven hub to hub markets. Plans to code-
share on domestic routes had also created problems for these carriers among their labor unions. For
instance, Delta and United Airlines decided to forego code-sharing when Delta's pilot unions refused to
consider it. Pilots at American Airlines had responded similarly. American Eagle and US Airways
Express also had labor discontent.

13

.Y ~'.-. "

each of these alliances allowed its member carriers the realimtion of forfeited cost
synergies or of greater market power.

The answer to this question hold relevance for the business traveler (a cornerstone
demand segment of airline industry revenues and profits), rival carriers (that may either
benefit from the fare 'umbrella' provided to them, or be placed at an unfair advantage for
instance, by being denied market entry), ticket purchasing firms (the Agent) as well as
'watchdog' agencies like the US. General Accounting Office and the Department of

Justice}6

1.6 Layout of Paper

This chapter provided a narrative history of the evolution of the domestic airline
industry, fiom the mega-mergers of the mid 1980's to the increasing cooperation between
large carriers in the 1990's, including the eventual combination of previously distinct
Frequent Flyer Programs. In the next chapter, the structural model is presented and
empirical literature on both airline mergers and (international and domestic) airline
alliances is reviewed. Chapter 3 begins the analytical query in order to determine the
answer to the hypothesis of increased efficiency versus greater market power due to

domestic alliance formation. In Chapter 4, the relation between each alliance event and

 

’6 Currently, both the Department of Transportation and the Department of Justice are responsible for
ensuring the working of competition in the airline industry, with the latter responsible for taking actions
against mergers that may stifle competition. Thus the Department of Justice has the authority to initiate
any proceedings against an acquisition proposal if it violates the Clayton Act, which applies to any merger
or acquisition that may substantially lessen competition in a relevant market, or tend to create a monopoly.
The Department of Justice also has the authority to enforce the Sherman Act, which prohibits unreasonable
restraints of trade and attempts to establish and maintain monopoly power. See Clougherty (2002) who
notes that since over 90% of US airline revenue is domestic, the international competitive effects of
domestic airline mergers should have less political-economic weight in this country than in other nations.
Therefore, antitrust authorities will be primarily concerned with the domestic competitive effects of
domestic airline mergers.

14

the changes in airport dominance that it affords, is examined. Chapter 5 focuses on some

important route sub-samples and Chapter 6 bears the concluding remarks.

15

Chapter 2. The Structural Model and Empirical Review

The two main institutional arrangements that have had a profound effect on the
airline industry in recent history have been the mergers that dominated the structural
landscape of this industry immediately after deregulation and the more recent trend of
alliance formation. While the motivations and institutional impact of these and other
main events in the post-deregulation airline industry were documented in the previous
chapter, here we will discuss the documented empirical impact of these important events.
We begin though with a structural model of firm behavior that captures and defines our

main hypothesis.

2.1 The Structural Model of Equilibrium

Our structural model of profit maximization captures both firm specific responses
as well as inter-firm rivalry. Consider an airport-pair market in which there are two firm
'types', that is, alliance and rival firms, both producing a homogenous service.” The ifl‘

firm's route (r) and time (t) specific output is qt“. Total market output is therefore
2
Qt = Z Qin-
i=1
The market inverse demand function is given by:
PIT = D(Qfl9 xdrls Add, 8 dfl)

(2.1)

 

37 This is an arguably inappropriate assumption for the airline industry. However, it is one that has been
often adopted in its previous empirical literature.

16

where Xdin are the variables effecting demand, Adm represents the vector of variables that
affect a firm's (route and time specific) demand due to alliance participation or non-
participation, as the case may be. 8 “in is the error term.
The firm level total cost function is dichotomized into:
Cirt = C(qcirta X cirt, Bern, 8 cin)
(2.2)
Where X ‘2“ are the factors that affect firm costs, B°m is a vector of variables that directly
affect firm costs due to alliance participation or non-participation. 8 °m is the error term
to capture the cost impact of time varying unobservable factors on firm costs. Note that
de and X cm may also capture the impact of route specific measures that are time and
firm invariant, for instance airport specific infrastructure constraints.
Given these demand and cost specifications, firm i's route and time specific profit
frmction is:
“in = D(Qrt9 Xdirtr Adm, 8 dirt)<1irt - C in
(2.3)
With Q“ as the choice variable, the Cournot-Nash equilibrium is represented by

the following first order condition that yields the market price:
Pitt = MCirt (Q cht, X cht, Bcilta 8 din) - (6 P/ 5Q) clirt 9m

(2.4)
The introduction of a route subscript in equation 2.4 draws especial attention to
the fact that fare behavior in the airline industry is expected to not only depend on firm

identity and on the time period selected, but also on route specific characteristics. In this

17

setting, equilibrium is achieved when each firm optimizes its output, given the output of
the other firm.38

The standard Cournot model predicts that if the alliance firm restricts its output
after the event, its rival firms will respond with an increase in their own output.
However, rival firm response may be restricted due to their non-participation in the
alliance. That is, for firms remaining outside the agreement, demand and/or costs may
change (the former captured by Adm and embedded in equation 2.4, the latter by 8%) in
the post-alliance period due to their non-participation. We therefore want to allow for the
possibility of alliance and rival firm fare differentials, despite their producing a
homogenous service.

In equation 2.4, 9m = (dqj / am), where (j at: i), and is the firm, route and time
specific Conjectural Variations parameter, or according to Bresnahan (1989), is the index
of "the competitiveness of oligopoly conduct. " The Cournot solution corresponds to zero
conjectural variation and if the firm behaves more collusively than Cournot, 9m >0.
Conversely, 0m <0 depicts a more competitive state than captured by the Cournot setting.
In an oligopolistic setting therefore, film conduct can be inferred from its fare behavior.
If price, rather than quantity is the choice variable, then Bertrand solution yields marginal
cost pricing, that is, the same as under perfect competition. At the other extreme, (tacit)
collusion between two (identical cost) firms with an aim to maximize their joint profits

implies that 0111:].

In the Cournot model, the market price moves toward the competitive price if the

number of firms is large enough. Then each firm realizes (approximately) zero profits and

 

3’ An alternative fiamework is when each firm views the output of the other firm as a function of its own

18

acts as a price taker. However, when firm products are differentiated, the firm no longer
faces infinite demand elasticity at equal prices. This means that positive price cost
margins are possible due to product differentiation.

Equation 2.4 can be re-written as:

[Pu - Met. (0.... x... B“... s “all IPn = [(1+ 9m) Msa 1 n
(2.5)

Where MS“, is firm i's (route and time specific) market share and r] = -(6 Qn/ 6
P“) (Pn/ Q“) is the (positive) price elasticity of market demand. Equation 2.5 shows that
price is a (non-linear) fimction of marginal costs and a 'mark up' that is positively
correlated with its market share, and that also depends upon the conjectural variations
parameter. That is, an increase in either market share or the conjectmal variations
parameter will increase price.

Keeping with the accepted protocol of new empirical industrial organization, we
proceed with that firm level marginal cost is not directly observable. Rather, the price-
cost margins will be inferred fi'om experiments of firm behavior.39

Writing out the model in terms of the Lerner Index (equation 2.5) shows that the
spread between a firm's prices and its marginal costs (or the 'price distortion') will rise if
its prices rise and its consumers adjust their demand downward only slightly. Thus low
demand elasticity induces strong price distortions. That is, even when the firm increases

its unit price, the corresponding demand response from its consumers is low: market

 

output decision.
39 For more on the current framework of empirical oligopoly, see Bresnahan (1989).

19

share adjusts downward, but does not go to zero and there is a large monetary transfer
from consumers to the firm.
We can also write firm is perceived elasticity of demand as:
mu =n/ MSm (1+ Him)
(2.6)
Equations 2.5 and 2.6 show that each firm's price-cost margin ratio is determined
by the reciprocal of its perceived elasticity of demand. In the monopoly state (or with
cartel formation in an oligopoly setting) this implies that the firm's own perceived price
elasticity of demand is equal to the market elasticity 1], otherwise m > 11. In an oligopoly

for instance, in may or may not equal 1].

2. 2 Efliciency versus Market Power- The Primary Hwothesis
Equation 2.4 captures the primary hypothesis of this paper: did the alliance film's
fare response indicate that it had realized an improvement in efficiency due to alliance

formation, or had increased market power dominated?

Factors that directly influence the alliance firm's demand after alliance formation
as well as those that can affect its costs, are embedded in this equation through Adm and
8%, respectively. Specifically, if the alliance firm realizes improved efficiency/cost
synergies through alliance participation, lower marginal costs will lead to lower fares, at
least beyond the short term and the price gap between alliance and rival firms will
decrease in the period corresponding to alliance formation.

There are two potential sources of efficiency gains especially worth mentioning

within the fiamework of domestic alliances: economies of density and economies of

20

scope.40 Substantial cost savings due to the former can be realized if increased market
access and/or traffic feed can be achieved, which in turn depends on the degree of
network integration between the partner carriers. And the latter, while relevant in the
airline industry, will not be realized unless partner carriers re-configure their networks

after alliance formation.41

However, if increased market power is the dominant outcome from alliance
formation, alliance firm fares will be rising (through higher Adm), ceteris paribus, and the

price gap between alliance and rival firms will increase.42

To the extent that rival firms are able to take advantage of the umbrella of alliance
firm fare increases, it will indicate a collusive environment (that is, 0 in > O in equation
2.4) and will translate into a post alliance price gap similar to the one before alliance

formation.43

In summary, it is the direction of growth of alliance firm fares that will signal the
dominance of either improved efficiency or of increased market power.44 Even if some

of both these effects are realized, the direction of change in their fares will indicate the

 

4° See Caves, Christensen and Tretheway (1984) on the importance of economies of density (rather than
economies of scale) in the airline industry.

4' We assume that new entry does not take place in response to alliance fare decreases.

‘2 This outcome can result either from the larger size of the 'new' firm, or through reduction in competition
due to the elimination of a competitor. See for instance Borenstein (1990) who finds evidence of increased
market power from the Northwest-Republic merger at its Minneapolis hub.

’3 We note that concerns regarding collusion the airline industry have been often raised and are augmented
not only due to the dominance of key airports by a few carriers, but also due to the often documented fare
implications of multi-market contact in the airline industry. Secondly, changes in fares and quantities are
easily observable in this industry. Therefore alliance formation can potentially not only provide formerly
competing carriers with a forum for a collusive stance, but a host of other factors pre-exist that create
strong incentives for firms to abide by the terms of some (tacit) agreement. See Alam, Ross and Sickles
(2001) on how a stable price relationship between firms can signal successful dynamic oligopolistic
interactions.

21

dominant effect.45 This then defines the test of the primary hypothesis of efficiency
improvements versus increased market power from alliance formation and this exercise
will be detailed and undertaken in the next chapter. First though, the more recent
empirical literature on the main events that have shaped the post-deregulation airline

industry is reviewed.

2. 3 Previous Empirical Evidence

2. 3. 1 Airline Mergers and their Impact on Fares
The effects of the post -deregulation trend of growth by merger in the airline
industry have been documented in a number of research papers. For instance, Borenstein

(1990) aimed to uncover the effect of the 1986 merger of Northwest and Republic Airlines

and TWA's acquisition of Ozark Airlines. This paper uses the same basic principles of our

model of equation 2.4 and of our hypothesis of Section 2.2 where we presented that firm
fare changes indicate the dominance of either improved efficiency (through lower B°m in
our equation 2.4) or increased market power (through a higher Adm). In Borenstein
(1990), a relative fares ’6 difference in dtfierences technique detects an empirical link
between the event and increased market power in the Northwest-Republic merger case.

And at the hub airports of TWA-Ozark, the paper shows that there was no systematic

 

“ Evidence of increased market power was found by Borenstein (1990) for the Northwest-Republic merger
of 1986 at its Minneapolis hub, and by Kim and Singal (1993) who examined fourteen airline mergers that
took place during the airline merger wave in the mid 1980's.

‘5 This reasoning assumes that alliance participation the only difference between the firms. Recall that to
the extent that such participation creates sufficient product differentiation, fares may not have to be
downward responsive to realized cost synergies.

‘6 Relative fare is defined as the ratio of the fares of the merging carrier on its major hub airports, to
average industry fares for routes of the same distance.

22

difference between the fares of this merged firm and that of other carriers operating in the
same markets, at least in markets where the merged carrier faced competition.

The basic principles of our behavioral model of equation 2.4 and the dlflerence in
dijferences approach is also the method of choice in Kim and Singal (1993) who examine
14 airline mergers that took place between 1985 and 1988. First, relative fare changes
are calculated between sample and control routes and between the two (before and after)
time periods in order to determine whether the dominant effect was improved efficiency
or increased market power. Results show that these mergers were associated with a 10%
increase in airfares and that rival firms had followed with fare increases of their own,
showing the dominance of increased market power fi'om the mergers and the presence of
the umbrella efibct, respectively. Next, econometric estimation revealed a significant
positive correlation between fare changes and changes in concentration for both merging
and rival firms.

Farrell and Shapiro (1990) analyze horizontal mergers in a Cournot oligopoly and
find that for a merger to lower prices, considerable economies of scale or learning need to
be realized. McAfee, Simons and Williams (1992) show that when firms engage in
spatial price discrimination, their equilibrium post-merger prices increase.

Boyer (1992) has shown that a merger results in a decrease in output in those
markets where competition is eliminated and the resultant decrease in marginal cost of
the merging firm causes harm to non-participating firms. This harm to rival firms is
presented as a sufficient index of harm to social welfare. This model is one of
oligopolistic interaction in which firms function in overlapping but related markets and
where the good is homogenous but price charged depends on market conditions. The

possibility of separate markets and of non-uniform pricing creates conditions for the

23

absence of an umbrella efi’ect for rival firms: there is an increase in competition even
outside the core markets dominated by the merging firms where rival firms are placed at
a price disadvantage, though not because of their inefficiency. The harm to consumers
and to rival firms is sufficient to characterize the merger as harmful. The two important
assumptions to reach this result are that firms in the industry have strengths in different
markets (that is, they operate in different locations), and that they face capacity limits and

inter-market connections.

2. 3.2 The Fare Impact of A irport and/or Route Dominance

Borenstein (1989) has estimated the impact of route and airport dominance in the
degree of market power exercised by the nine largest domestic carriers in the third quarter
of 1987. After controlling for some important measures of quality and cost, results show
that both route and airport market shares are relevant in determining the degree of market
power afforded to a carrier: a 1% increase in the carrier's route market share is estimated
to increase its fares by between 0.03% to 0.22%. The fare effect of airport dominance
was also found to be strong, especially for high-end fares. Results show that a dominant
carrier47 charges 6% higher median and high-end fares and that smaller carriers were
unable to benefit from the 'umbrella' created by the dominant carrier, a discrepancy
explainable by marketing devices (such as Frequent Flyer Programs) that favor the
dominant firm.

Evans and Kessides (1993) also sought to answer the same question: that of the
impact of airport and route dominance on the ability of a dominant carrier to raise fares.

Examining 1988 data for 22 carriers for the top 1000 most heavily traveled routes and

24

 

using fixed-effects estimation, they found that while airport dominance does create
substantial benefits for the carrier in allowing it to raise it fares, a result also noted by

Borenstein (1989), no pricing power was found to derivable from route level dominance.

2. 3.3 Frequent F lyer Programs

As noted earlier in Chapter 1, the 1980's trend in the airline industry of growth by
merger was followed by the introduction of programs designed to allow carriers to
achieve growth by capturing consumer loyalty. These programs allow the carrier to earn
a disproportionate share of its net revenue from a specific passenger type. Specifically, it
is the low fare/ high time valuation traveler that is their most important source of revenue.
Given the recent interest by businesses to curtail their travel costs, benefits provided
under a carrier's Frequent F lyer Programs became a vital means of achieving revenue
growth for carriers operating in a tight economic environment.48 They may even have

been a way for carriers to realize increased load factors."9

In any case, the programs
served an important source of revenue growth.

Banerjee and Summers (1987) modeled Frequent Flyer Programs as collusion
facilitating devices that allow firms to split the market and then charge higher fares fi'om
the price inelastic and time sensitive class of consumers. The creation of an artificial

switching cost allows positive economic profits, as consumers may redeem their benefits

only upon remaining loyal to the firm.

 

’7 Airport dominance refers to an airport market share of at least 50%.

4’ See for instance, Stephenson and Fox (1987).

‘9 Load factor is defined as the ratio of the number of seats filled to the total number of seats on the aircraft.
Higher load factors can lower fares through a lower X2, in Equation 2.4, or even raise them through higher
X in.

25

Through a two-firm (A and B), two-period (Fl,2) model they show that these
programs can increase the payoffs to all participants in the game. The two firms produce
a homogenous good at zero marginal costs and no fixed costs, compete at the price level
(P‘.) so that o s P’. _<_ 1 (where firm i is the price leader in the tth period) and in the non-
transferable coupons offered (CA2 and C32). The coupons can be used for discounted
travel in the second period. Prices are set sequentially in each period (PAr, P31, PAZ, RE).
The price leader in the first period is picked randomly and in period 2, the firm with the
largest market share leads fare setting. Consumers have a reservation demand of 1 unit
and a homogenous reservation price of 1. Each consumer makes a firm choice decision
upon observation of both fares and coupons and maximizes his or her expected utility
over the two periods.

First consider that the second period game is such that an equal proportion of
coupons from each film have been selected in the first period. Thus, in the second period
the two firms evenly split the market. Their market shares are therefore, 11A] = 1/2 and
“a! = 1/2. If firm A is the price leader and it seeks to cooperate by not undercutting, then
it will set PA; such that PA; 5 20‘; + C82. IfA sets PA; = 20‘; + C32 (where 20‘; + C32
S 1), then it seeks to enforce cooperation. While setting PA; > 1 will lead to zero profits
for A (as B will undercut), setting PA; = 1 dominates setting P“; < 1. This means that
firm profits will be an increasing function of coupon size. F irms benefit from an increase
in their rival's coupons as this deters undercutting. Thus when coupon size is sufficiently
large, the firms split the market and the joint monopoly outcome results.

Second, consider the scenario that firm A had the larger market share in period 2
and is therefore the price leader. Once firm A's second period price is set (PAz), carrier B

can set its fares such that each firm retains its first period consumer base. This implies a

26

lower limit on firm B's second period fares (PA; -CA <PBz <PA2 +CB) if it desires greater
profits. That is, while firm B could set a low price and capture the entire market, it does
better to set a higher fare if the value of the coupons is high. Similarly, it is in the interest
of firm A to set its second period fares such that firm B sets a moderate fare, allowing
both to retain their first period customer base. Thus second period profits are an
increasing function of the value of the coupons set in the first period. They show that an
equilibrium is CA=1= CB,PA1=1= P3,, P"2 = 2 = P52.

This model shows that Frequent F lyer Programs allow firms to abandon
incentives to undercut and start a price war. As long as each firm has the potential for
significant positive profits in the second period, it is in its interest to share the market.

Cairns and Galbraith (1990) have shown that an incumbent firm offering a
Frequent F lyer Program is able to create a barrier to entry despite zero cost advantages.
It does so through the creation of an "artificial compatibility" among its own services,
that is, by creating an incentive for the consumer to buy from a single firm. Offering a
more attractive rebate program allows the incumbent to maintain positive profits as well
as to deter entry. Even if the rival firm offers its own rebate program, the scope of the

incumbent's network makes the valuation of its rebate program higher.

2. 3. 4 International Airline Alliances
Recent empirical work on international airline alliances has consistently noted
their welfare improving effects. Brueckner's (2001) theoretical model shows that in

interline markets, code-sharing will improve welfare through cost efficiencies achieved

27

through cooperative pricing between the two partner carriers.50 This benefit in interline
city-pair markets will be greater than the increased market power in inter-hub markets,
where competition will be reduced.“

If international alliance partners coordinate their operations to feed passenger
traffic to each other's hub airports, it allows both partner firms to effectively enter new
city-pair markets without any investment in new resources. Brueckner, Dyer and Spiller
(1992) show that increased traffic density translates into economies of density.

Using a model similar to our equation 2.4, where the dominance of efficiency
versus that of increased market power has divergent effect on fares, Park and Zhang
(2000) present an empirical analysis of international airline alliances. They find that
these partnerships generated additional passenger demand and lower equilibrium level
fares. These welfare improvements were led by the cost reductions achieved after a
reduction in the redundancies among partner firms, who integrated activities and linked
networks. Overall, code-sharing among the partners allowed the realization of
efficiencies and improved convenience for interlining passengers.

Similarly, Brueckner (2003) has examined the change in the interline fares paid
by international passengers due to the effects of code-sharing and antitrust immunity.
The results show that code-sharing between the carriers had the effect of reducing
international interline fares 8°/o-17%, and that antitrust immunity reduced fares by 13% to

21%.

 

5° Code-sharing is the practice of using of an airline’s two-letter code on another airline’s flight It allows a
carrier to expand its network without substantial costs and is a common feature of international airline
alliances.

5' That is, markets between the hub cities of the partners.

28

2. 3. 5 Domestic Airline Alliances

The formation of alliances in 1998 involving the six largest domestic carriers in
the United States (which accounted for about 70% of the domestic airline traffic then)
received attention in the press and raised concerns in regulatory circles of their potential
anticompetitive effects. However, we are not aware of any empirical study that has
examined the actual competitive impact from these three partnerships: the most recent
academic and regulatory inquiry on the airline industry focus almost exclusively on
international airline alliances. There were two exceptions.

First, in early 1999 the US. General Accounting Office presented an examination
of the likely effects of the alliances formed between the big six domestic carriers, with a
focus on the anti-competitive effects likely if all three were to also implement code
sharing.52 Overall, they found that with such a scope, all three alliances could have
positive effects for consumers in terms of additional flight frequencies and points served
but would likely harm new entrants and non-alliance carriers. It was recognized that if
competition were reduced between the carriers once they entered the agreement, it would
create competitive harm for consumers. The three alliances were not expected to
stimulate demand for air travel, but to allow demand shifts away fi'om non-alliance
carriers.

The tabular analysis of this study was based on pre-alliance (1997) data and is
therefore predictive. An accurate identification of the effects of these alliances should be
based on the structural changes brought about by the event, using both pre and post
alliance data. The conclusions of this study therefore remain to be empirically confirmed

or refuted.

29

Second, Bamberger, Carlton and Neumann (2001) examined the code-sharing
alliances between Northwest and Alaska Airlines (which began in 1995) and between
Continental and America West Airlines (which began in 1994). This paper is part of their
testimony in favor of these two alliances.53

Since code-sharing by the carriers in question took place on only a handful of
routes, fare (and traffic) changes were examined on routes on which code-sharing had
taken place (the treatment group) versus those on which it had not (the control group).
Specifically, the treatment group of routes included those city pairs on which a code-
sharing agreement converted a potential interline flight into a code-shared flight. Using
the first difference method of estimation, they found that the growth in the price gap on
sample versus control routes was negative.54

Thus while their method is compatible with our model of equation 2.4, where
improved efficiency and increased market power have opposing effects on fares, alliance

firm fare behavior is not examined and the question of the realization of efficiency versus

market power at the firm level, remains unanswered.

2. 4 synopsis and Conclusion

The hypothesis of the attainment of efficiency improvements following domestic
alliance formation versus the attainment and exercise of market power on important
domestic routes was explicitly presented in this chapter through a price equation. This

model has been the basis of empirical work in the airline industry for both merger and

 

52 US. General Accounting Office, RCED-99-37.
’3 Testimony was presented by Carlton and Lexecon Inc. and was sponsored by Northwest and Continental
Airlines.

30

—-'-

 

alliance analysis and shows how the two competing effects possible from mergers and
alliances, are nested in profit maximization theory. This model then forms the basis of an

examination of alliance fare behavior that will be undertaken in the next chapter.

 

“For more on panel data methods, see Wooldridge (2002).

31

Chapter 3. Alliance Formation: Efficiency versus Market Power

In the previous chapter it was discussed that within the general fiamework of
Bresnahan (1989) improvements in efficiency versus the exercise of market power have
divergent effects on firm fares. Thus the primary hypothesis is whether alliance fare
behavior indicates that they realized higher efficiency/cost synergies after alliance
formation or did greater market power dominate? Note that the analysis of the
competitive impact of these alliances assumes that the two partner carriers behave as if

they had merged.55

3.1 The Price Equation

A price equation that represents a reduced form from the previously specified
structural model of Chapter 2 (equation 2.4) is specified and it is assumed that the
equilibrium price is affected by both endogenous and exogenous variables. In equation
2.4, it was noted that the price charged by a carrier on a route depends on factors that
influence demand and supply. The assumption of unchanged supply factors means that
the price equation can be thought of as a reduced form specification, with demand
characteristics included as explanatory variables. Thus the results of this chapter, given
the previously specified structural model in Chapter 2, will be seen as tests of the primary

hypothesis.

 

’5 An approach that follows the literature on airline alliances. See for instance Brueckner (2003) and
Bamberger, Carlton and Neumann (2001).

32

A log transformation of equation 2.4, where price is a non-linear function of the
marginal cost, plus a markup term (which in turn is shown to depend on market share and

the conj ectural variations parameter) provides the linear approximation for estimation:

Ln Pitt = Ln MCirt (Qirt, X in, Berri, C in) ' Ln I 1 + (1/ 11) 9m MSirt l
(3.1)
Equilibrium fares are estimated after replacing the last term of equation 3.1 by a
linear function of other factors affecting fares in the airline industry, shown by the vector

Xi“. Therefore equation 3.1 is rewritten as:
Ln Pitt = Ln MCIfl (Q1119 X i119 Bcifla e II!) + X111

(3.2)
Again, the accepted protocol in new empirical industrial organization is that firm

level marginal cost is not directly observable. The basic model is:

In Pirt=Xirt 3+ Uirt

(3.3)
where Xm is a vector of regressors, including a constant. Equation 3.3 is the price
equation that will be estimated on the balanced panel and where i = 1... n (subscript for
firm), r = 1,...n (subscript for route) and t = 1,2,3,4 (subscript for quarter). The
dependent variable is (the log of) the one way average fare of carrier i on route r in
quarter t.56 It is therefore half the average coach class fare paid for a round trip during a

particular quarter, outbound from the base airport.

 

’6 For the alliance, this is the average fare of the two alliance members for each identical itinerary.

33

3. 2 Explanatory Variables and Expected Signs

To test the primary hypothesis of the dominance of efficiency versus market
power from alliance participation, the empirical analysis relates the log of average fares
(lfarem) to alliance formation and to other variables that form the vector of structural

variables (Xrn in equation 3.2). Thus the price equation contains a constant plus:

1. An Alliance Participation dummy variable (ALLYtn) that indicates whether the
firm is a member of the alliance in question. The sign and magnitude of this explanatory
variable is the main focus of the empirical analysis of this paper as it will provide the
answer to the primary hypothesis. Specifically, a negative coefficient on this dummy
variable (ALLYm < 0) means that alliance formation was associated with lower average
fares, or that the event generated efficiency gains. Conversely, a positive dummy
variable (ALLan > 0) is evidence of the realization of increased market power from
alliance participation.

It is defined as one when the firm is a partner of the domestic alliance in question,

and zero otherwise. (Expected Sign: Indeterminate).

2. A Cost Component. The structural model of equation 2.4 shows that the
equilibrium price also depends on a cost component. In air travel, generally the most
important cost component of a trip is its distance, that is the non-stop mileage (in statute
miles/ 5,280 feet) between the origin and destination airports, and is taken in logs
(IDIST,). There is also a direct relation between the distance between two airports and
the lack of good substitute modes of transportation available to the consumer, and

therefore demand for air travel. That is, substitute modes of transportation become less

34

attractive for longer distance travel and so firms may realize greater fare leverage on
longer distance routes. However, costs should not increase with distance in a linear

fashion.57 (Expected Sign: Positive. Expected Elasticity: Less than one).

3. Time Dummy Variables. T-l time dummy variables capture the effects of
macroeconomic factors on the dependent variable. Given that air travel is highly pro-
cyclical, the inclusion of time intercepts permits control of the seasonal influences on
fares. Their exclusion would force one to assume that the change in the dependent
variable (lfarein) is due to alliance participation (captured by ALLYin) when in fact it may

have actually been due to external, economy-wide effects.

4. A measure of competition. The count of the number of rival firms operating on a
given route 'r' at time 't' will proxy the degree of competition there (Nfirmsin). (Expected

Sign: Negative).

5. A dummy variable to identify routes on which at least one endpoint is gate
constrained or slot controlled (Gate/Slob). It takes a value of 1 on airports that are gate
constrained or slot controlled, and 0 otherwise.58

Slot controlled airports are those where the Federal Aviation Administration has,
since 1969, placed limits on take off and landings in order to minimize flight delays.
While these slots were initially allocated in 1985, the Department of Transportation had

allowed carriers to buy them from and sell them to other carriers. However, the DoT had

 

’7 Shorter distance markets generally tend to have a higher per mile fare than longer distance ones.

35

maintained the ownership of 5% of the slots at these four airports, and later distributed
them through lottery to those carriers that had few or no slots. Many of these were
carriers that subsequently went out of business resulting in the increase in slot ownership
by a few firms.59

Similarly, gate constrained airports are those where gate facilities are limited
through long term and exclusive-use leases. Again, these leases tend to be owned by
larger carriers.

In terms of the impact of these time invariant constraints on the dependent
variable, note that when the ownership of these factors is controlled by one or a few
firms, market demand will be more easily concentrated with a few firms, an affect
captured through a higher X"r in equation 2.4.60 These firms may refiise to sell or lease
these factors to carriers wanting to expand at these airports, imposing higher costs on
these rivals, an effect captured through a higher X“r in equation 2.4. For firms owning the
majority of these scarce resources at an airport, positive fare changes represent the
extraction of market power and for those who do not, the scarcity rents due to non-

ownership of sufficient airport gates and slots. Both these conditions create an

expectation for a positive parameter estimate for this variable. (Expected Sign: Positive).

6. A dummy variable to indicate if at least one endpoint airport is geographically

'isolated'. This is the second route specific and time invariant structural factor that is

 

’8 Scarcity rents can be expected to accrue to firms that own these factors. Evans and Kessides (1993) find
airport capacity constraints augment a carrier's market power. See Appendix C for a list of the gate and slot
constrained airports within our route sample

59 See US. General Accounting Office, GAO/RCED-97-4 on slot ownership at dominated airports.

6° See Morrison and Whinston (2000), who find that much of the fare difference between hub and non-hub
airports can be accounted for by the fact that low cost carriers avoid congested airports.

36

important to consider in its effect on our dependent variable. It takes on a value of l for
routes on which at least one endpoint airport is 'isolated' and is 0 otherwise.

At secondary airports, costs tend to be low since they have lower Passenger
Facility Charges (PFCs). Here, delay costs should also be lower due to lack of
congestion problems.

This variable proxies for the rents accruing to carriers serving such 'isolated'
airports, an effect captured by a higher X"r in terms of the structural model and equation
2.4. It also proxies for the constraints faced by firms in expanding and competing with
incumbent firms in such a metropolitan area. The costs of these firms will be higher due
to this constraint, captured by a higher X“r in equation 2.4. The expected sign on the
ARPT, coefficient is therefore positive.

However, there is a second effect to consider with respect to this explanatory
variable. Recall from Chapter 2 that it is expected that the domestic alliances in question
are aimed at low elasticity consumers with high time/low price valuation. If this demand
segment has stronger airport location preferences, then the absence of a competing airport

may prevent the carriers to segment demand by elasticity. This then forms the basis of

61

the expected sign of the ARPT, dummy variable to be negative. (Expected Sign:

Indeterminate).

3.3 Data: Source and Description

The Department of Transportation collects data from all large certified air carriers

conducting scheduled passenger service. For each quarter, the raw data-base (the Ticket

 

6' See Tirole (1994) on second-degree and third-degree price discrimination.

37

Dollar Value Origin and Destination, or 'O&D' data bank) contains domestic economy-
class airfares and number of passengers on a route, identifies the carriers, the point of
origin, intermediate airports, point of destination, distance and fare class. Since July
1987, the 0&D has been based on a stratified, scientific sample of at least 1% of tickets
in major domestic markets and 10% of tickets in all other domestic city-pair markets.

An important feature of the data is that post 1998, carriers were subject to new
requirements of reporting both the operating and marketing carrier codes on each coupon
of each record in their survey filings (as opposed to just the marketing carrier, which they
had reported until the previous year). This feature allows the elimination of noise in the
data due to any code-sharing. That is, credit is given to the carrier on which the
passenger actually traveled instead of the carrier that marketed the flight"2 Further
details regarding data base construction form Appendix A.

Data is aggregated at the firm level within each route. Fare data represents the
average coach class fare paid each way by (local and connecting) passengers for a round
trip during a particular quarter, outbound from the base airport, such that at least one

member of the alliance operated there.

3.4 Market Definition

The most relevant definition of the market for the purposes of this paper is the
origin to destination air transportation market. Note first that this definition of a market
is distinct from that of a route: a market represents the actual trip origin and destination,

while a route represents the actual path flown by the passenger between his/her trip origin

 

‘2 Pre-l998 data used in this paper has been adjusted to reflect this change in reporting practice.

38

and destination. Since passengers flying in the same market can fly a variety of different
routes, travel from point A to point B is considered to be a different market than travel
from point B to point A.

Second, a market is defined as travel between a unique airport pair, rather than a
city pair. That is, travel from multiple airports in the same city is not aggregated.

Reference will however be made to markets and routes interchangeably.

3. 5 Treatment and Control Firms

The domestic operations of the 10 largest carriers in the domestic airline industry
are considered. Just before the alliances were annormced in 1998, these firms served over
87% of the domestic passenger traffic flying with major, low cost or regional carriers.
For every alliance analyzed (the treatment group), the control includes the other eight
largest carriers, that is, exclusive of only the two participating in the alliance. The big six
carriers that formed the three major marketing alliances in 1998 were Delta-United,
Northwest-Continental and US Airways-American. Non-participant rival carriers that we
include in the control group are Alaska Air, America West, TWA and Southwest
Airlines.

Other than the firms participating in the domestic alliances in question, the
inclusion of only their four largest rival carriers was in part due to the understanding that
these large and dominant firms represent a distinct group worthy of separation from their
smaller rivals and from regional carriers. Evidence supporting this belief is common in
the literature on the airline industry, for instance, Borenstein (1991) has shown that the

advantages enjoyed by the largest firms in the industry flow not only due to cost and

39

quality differences, but also their distinctly successful marketing strategies and reputation
advantages.

This choice was further supported by the fact that the market shares of these four
major rival carriers are already so low in many of the busiest markets that we are
interested in, that they serve only as sporadic competitors to the big six carriers forming

domestic alliances.

3. 6 Airports Considered

The pool of airports considered is also a focused one. Specifically, alliance and
rival firm operations on routes between each alliance hub airport and the busiest 45
airports of the country are considered.63 These busiest airports of the country are ranked
according to passenger enplanements in the 1999 ACI Monthly Traffic Statistics and are
listed in Appendix B.

The airport selection criteria reflects the belief that these busy airports represent a
distinct market.64 For one, the competitive forces at play at these airports can be distinct
from those at other smaller airports that a major carrier may service. Second, it is
expected that the passenger traffic at these airports makes them the most relevant, not

only for the benefits/burden from the realization of cost synergies/market power that may

 

‘3 The Federal Aviation Administration (FAA) defines a large hub as a geographical area in which its
airports account for at least 1% of the total annual enplaned traffic. A canier’s network hub is the
designated 'central' airport from where passengers are redistributed to their intended destinations.

6" An argument supported in airline industry literature. For instance, see Borenstein (1988 and 1989) and

Berry(l990).

40

result from an alliance agreement, but also with respect to the type of travelers most
likely to be affected by the formation of domestic airline alliances.65

The sample of airports considered in this paper contains all those vital gateway
cities on which carrier operations are constrained due to gate constraints or slot controls.
Scarcity rents at these airports can be expected to affect fares there, especially for smaller
carriers that are unable to expand in these markets,66 as opposed to dominant carriers who
tend to own their majority.67 Airports with gate constraints or slot controls are listed in

Appendix C and Appendix D lists the airports that have a second airport within the same

metropolitan area.

3. 7 Time Period Selection

In order to obtain a satisfactory answer to our primary hypothesis, an extended
time period is selected. This is especially important in industries in which consumer
loyalty plays an important role.68

Three specific periods for each alliance were earmarked: the quarter before the
alliance was announced in the press, the quarter after alliance agreement but before its
consummation, and the quarter nearly a year after the alliance had been operating.

The Northwest-Continental alliance was announced in the press on January 27th

1998 and started operations in December 1998. Therefore, the second quarter of 1998

 

‘5 This subset of markets is also of particular recent interest to agencies like the General Accounting Office.
See for instance, US. General Accounting Office, GAO/RCED 90-102, 93-171, 97-4 and US. General
Accormting Office, GAO/01 -5 18T.

66 See Dresner, Windle and Yao (2002) who find that these constraints have a significant impact on carrier
yields.

‘7 See US. General Accounting Office GAorr-RCED-98-l 12.

68 Focus on an extended time period allows a distinction between medium and long term goals of an
alliance. Specifically, while the medium term goal of an alliance may be an increase in market share, its

41

was the period after alliance agreement, but before its consummation. The third quarter
of 1999 was selected to capture firm behavior (nearly) a year after the alliance was
operating.

The selection of quarters for the other two alliances required more judgement
since both the Delta-United and the US Airways-American Airlines alliances were
announced and consummated in two consecutive quarters. Specifically, the US Airways-
American alliance was announced in the press early on April 5th 1998 and started
operating in the middle of the third quarter of 1998, specifically in August 1998.
Therefore the previous (that is, the second) quarter was selected as one that was 'after
alliance agreement but before its consummation'. Similarly, the Delta-United alliance
was announced on June 30th 1998 and began operation at the end of September 1998.
Therefore, the third quarter of 1998 was selected as one that was 'after agreement but
before consummation'. And for both, the third quarter of 1999 was selected to capture
firm behavior a year after alliance operations. Table 3.1 shows these time periods for

each alliance.

 

long term goal can be focused on improving the bottom line, be it through reduced costs or through
increased revenues.

42

Table 3.1: The Sample of Airline Alliances

 

 

Enplaned Pagers a Time Period b
Before After One Year
Alliance: Airline 1 Airline 2 Negotiations Agreement After Alliance

 

DL-UA 27,549,985 21,941,274 Q2, 1998 Q3, 1998 Q3, 1999

NW-CO 13,679,584 560,776 Q4, 1997 Q2, 1998 Q3, 1999
US-AA 1,533,858 20,909,503 Q1, 1998 Q2, 1998 g3, 1999
' Passenger data is for scheduled and unscheduled passengers, before or during preliminary alliance
negotiations.
b Time period is shown as Quarter,Year. See Chapter 3, Section 3.7 for time period selection criteria.

 

3. 7.1 The Timing of Fare Changes

Given that three quarters (t) are selected for each alliance, note that at t = 1 none
of the firms participated in these alliances. By t = 2, the two firms had already
announced the agreement and rival firms were excluded. Thus the announcement eflect
is definedasthe fare change betweent= 1 andt=2. The change infares betweent=2
and t = 3 is defined as the alliance completion effect: at t = 2, the two firms had not yet
begun participation in the program, while at t = 3, the program was operating for about a
year. The fill eflect considers fare changes between all three quarters that is, from before
alliance announcement to a year into its full-fledged operations.69

In terms of our primary hypothesis of Section 2.2, consider that when two former

competitors join operations or decision making at any level, potential efficiency

improvements cannot be realized until the alliance is actually formed and operating.70

 

‘9 This demarcation of time periods will be adopted throughout this paper and is similar to that in Kim and
Singal (1993).

7° There is a potential for improvements in efficiency after alliance formation, for instance by integrating
redundancies betweeen the two firms and through economies of density. For instance, cost synergies can
be realized due to better load factors, better coordination of ground crew, of flight arrival and departures
and of gates and slots, through economies of scope and/or density, etc. See Bradley, Desai and Han (1983)
who discuss synergistic gains and firm acquisition.

43

This points to the importance of the completion effect and the filll eflect for identification
of greater cost synergies fi-orn alliance formation.71

However, the exercise of market power by a firm does not have to wait until the
alliance is actually formed: fare changes can take place even on alliance announcement if
the two carriers seek such gains. Consider the scenario in which two (formerly
competing) management teams now have the opportunity to discuss and agree on
mutually beneficial strategies. The importance of the announcement eflect, the
completion effect and the fill] eflect is therefore established for identification of increased

market power fiom alliance formation.

3. 8 Evidence: Pooled OLS Estimates of the Impact of Alliance Formation on Fares

Independent and explanatory variable definitions are shown in Table 3.2. Table
3.3 shows pooled OLS estimation results with heteroscedasticity adjusted standard errors
and White t-statistics.

Results in Table 3.3 show that the distance (lDIST,) coefficient is unsurprisingly
positive: longer trips have higher fares.72 Since this and the dependent variable are in
logs, it indicates that, for instance in terms of the full effect for the US Airways-
American alliance network, that a one percent increase in distance raises average fares by
about 25%. It is Similarly positive and significant on other alliance networks and for all

three time periods considered.73

 

7' See Kim and Singal (1993) for a similar analysis of airline mergers.

72 Distance is measured in statute miles/ 5,280 feet.

73 The coefficient on Distance remained positive and significant in both linear and logarithmic forms. The
latter is reported.

44

Table 3.2: Variable Definitions ’

Variable
lafarern

ALLY in

IDIST ,

Gate/ Slot ,

ARPTr

Nfirrns in

D2/D3

Definition
Average one-way coach class fare for each round trip on carrier 'i',
route 'r' and at time 't', taken in logs.

Alliance participation dummy variable indicating whether the firm 'i'
on route 'r' was a member of a major domestic alliance at time ‘t'.

The one-way non-stop and straight line distance (DIST,) between
endpoint airports of route 'r', taken in logs.

Dummy variable that indicates whether at least one endpoint of a route 'r'
was gate controlled or slot constrained.b

Dummy variable that indicates whether at least one endpoint airport of a

route 'r' does not have second airport in the same metropolitan area.c

The count of the number of rival firms of carrier 'i' operating on a route 'r'
at time 't'. It is between 0 and 8 for alliance carriers and between 1 and
8 for the other major carriers.

Time dummy variables for the second/third quarter.

(a) See Appendix A for data description and data screening.
(b) See Appendix C for list of airports within the sample that are gate controlled or slot constrained.
(c) see Appendix D for list of airports within the sample that are not geographically 'isolated'.

The parameter estimates for the gate constraint/slot scarcity dummy variable (Gate/Slog)

indicates that these time invariant constraints had a quantitatively important and

statistically significant effect on fares of firms operating on all three alliance networks.

45

.82-. Ems-oi bozo-8.2 ._o>o_ .8. as .8 as; as a aaoaeme 888838 N... as ....
.mowSEo one he meant 05 we cameraman 8 com Em :ouoom .m .8325 new 8:38:63 038:8, bounce—axe .58 NM eBa- oom 3v
.Eotu 23:58 8.32 gone-33.8.58: 55» A5

 

 

 
 

888.8 58.8 @888 888.8 888.8 888.8
88 8828 888 888 .888 .8888 .88- - .828 5.8:
888.8 888.8 88.8 8888 88.8 888.8
88 8888 88...- 3888- .388 .888 .8888- - .8888 8-32
88.8 8888 8:88 888 58.8 888.8
88.: 888 888 .8888- .888 .888 L888- - .888 5525
Ecum— 53295.5
$888 58.8 888.8 818.8 888.8 888.8
88 8828 88.... 8888 .888 .888 - .888- .828 firms
A888 888 288.8 888.8 828.8 888.8
88 $88 888 888 .888 .888 - 8888- 288- 8-32
8888 A888 588 $88 88.8
88.: 8:8 888.8 .888- .888 - - 888- .888 .585
“gamma uflOEOQfl-Sflfl4
888.8 88.8 288.8 888.8 828.8 8888 $8.88
88.2 828 8888 888 .888 .888 .828- 8:8- .888 5.8:
88.8 888 A888 8288 58.8 88.8
8.8.8 828 888 - .8888 .888 .888- 888 .888 8.82
888.8 988.8 888.8 888.8 88.8 8:88 8:88
83: 8:8 888 .288- .838 .8888 .888- 888 .828 ass-no scum =5.
38382
2 am 8588 .85. .8885 .88: 8 8

 

 

46

Parameter estimates of Gate/Slot, were quantitatively similar in all three time
periods. Its full effect parameter estimate ranged fiom 6% on the Northwest-Continental
alliance network, to 10% on the Delta-United alliance network. With the average one
way coach fare of $212.56 on the former network and $200.94 on the latter, the fare
premium due to the scarcity of gates or slots was $13 and $20, respectively. These
findings are roughly in conformity to previous research: Borenstein (1989) reported
finding a range of premiums 1% to 7% at such airports in 1987 and Abramowitz and
Brown (1993) found that carriers operating on slot controlled airports were able to extract
a 4% fare premium in 1988.74

The statistically significant negative coefficient of ARPT on the Delta-United
alliance network shows that this constraint served to depress fares there. Again, both its
sign and magnitude were similar in all three time periods considered. Specifically, in the
full effect period, fares were 7% lower on the Delta-United network when at least one
endpoint airport was 'isolated'. Recall final section 3.2 that this finding is interpretable as
that the absence of a competing airport in the same metropolitan area had served to
prevent demand segmentation by elasticity, thereby keeping fares lower on such routes
within the Delta-United network. On the networks of the other two alliances, the
statistically insignificant coefficient on ARPTr shows that this constraint had no effect on
fares there.

Parameter estimates of Nfirmsin remained consistently statistically insignificant,

indicating that this proxy for the competitive state of a route had no impact on our

 

7’ Given that at least one member of the three major domestic alliances in question dominates ownership of
gates at some of the country's busiest airports, this point estimate reflects for these carriers, the market
power (as opposed to scarcity rent) associated with such ownership and control. See US. General
Accounting Office GAO/1' -RCED-98-l 12, for the percentage of slots that were owned by major carrier
groups between 1986 and 1996.

47

dependent variable.75 It was therefore excluded as an explanatory variable from this and
subsequent regressions. Results also remained qualitatively unaltered due to its
exclusion.

The parameter estimate of principal interest is the Alliance Participation dummy
variable (ALLY-n). Results in the first column of Table 3.3 show that it was positive and
strong on all three networks. This result was evident even in the period corresponding to
alliance announcement. For the full effect (defined earlier in Section 3.7.1), Table 3.3
shows that alliance participation was associated with fare increases ranging fiom 3% for
the Northwest-Continental alliance to 19% for the US Airways-American alliance. These
findings clearly show that that domestic alliance formation, resulting in pairings among

the six most dominant carriers in the country, has lead to the realization of market power.

3. 9 Mitigating a Potential Econometric Problem

The pooled OLS regressions in Table 3.3 may suffer from an econometric
problem that will make previously discussed results unreliable. Specifically, pooled OLS
estimates may be inconsistent due to the omitted variables problem since it ignores the
impact of time constant unobserved effects on the dependent variable. For instance, there
may be market specific and/or firm specific characteristics that affect fares (the
dependent variable) and that have thus far been relegated to the idiosyncratic error term

(Urn in equation 3.3).

 

7’ Recall that the sample of firms consists of the top ten major domestic carriers, two of which are linked
through an alliance at any given time. The sample of routes consists of those between each alliance hub
airport and the busiest 45 airports of the country, such that at least one alliance member firm operated there.
Therefore Nfirms is bormded between zero and eight for alliance carriers and between one and eight for
their rivals.

48

In term of the price equation, the basic model of unobserved heterogeneity in

which such unobserved effects are explicitly included is:
In Put =- Xirt I3 + Uirt+ Cir

(3.4)
where the dependent variable is the (log) average fare of firm 'i' on route 'r' at time 't'. X1,-
is the vector of regressors, including a constant. Here, Cr, is the unobserved
heterogeneity and its inclusion allows its correlation with alliance participation. This is
especially important in the context of the alliance event since alliance participation was
not randomly assigned, rather, member firms 'self-selected' into an alliance. At the firm
level, it captures innate features such as managerial quality and the route level, for
instance the 'mix' of passengers between leisure and business types. These features are
unobserved characteristics that can be viewed as (roughly) constant over the time period
of interest.

Within the methods available under the class of models of unobserved
heterogeneity, the method of choice is the first difference transformation (FD). This is
the panel data equivalent of the difirerence in diflErences approach and the general
intuition behind it is to examine the impact of some 'treatrnent' on a firm and to compare
its performance to a group of firms on which the treatment was not applied (the control).

The FD transformation lags the elements of the dependent and independent

variables and subtracts them. If two quarters are considered, that is t = 2, then

AIan=92+AXmB+AUirt

(3.5)

49

where A In Pm = 1n Pm - In P“, (.1 and 02 is the second period intercept.

After the FD transformation, the first time period considered for each cross
section is lost since there is no first difference for these observations. Similarly, any time
invariant element of the X,“ vector, as well as time invariant (unobservable) elements of a
firm's managerial quality (captured by Cr, in equation 3.4), will drop out due to the
application of the FD transformation. The adoption of the FD method thus also allows an
abstraction fi-om the influence of hub specific effects on the dependent variable.“5 To the
extent that the prime interest is in time varying explanatory variables, this feature does
not represent a significant limitation.

After the FD transformation, the parameter estimate of the time dummy variable
measures the growth in average fares for firms that are in the control group and over the
period, due to aggregate factors in the economy, ceteris paribus. If B; is the coefficient
on the Alliance Participation dummy variable (ALLY-,1) at t = 2, then [32 + 0 2 measures
the growth in average fares for the treatment firms during this period.

The parameter estimate of the Alliance Participation dummy variable (ALLY-n)
shows the difference in the growth of average fares between the treatment and control
firms (or the growth of the price gap), ceteris paribus. Thus again, the sign of the
coefficient on the Alliance Participation dummy variable (ALLY-n) provides the answer
to the primary hypothesis of improved efficiency versus increased market power resulting

fi'om the formation of a major domestic alliance. If greater efficiency/cost synergies are

 

7° The latter refers to the fare premiums carriers may charge on travel to and from its hub airports, relative
to those on the rest of its route system. Borenstein (1989) documents such a hub premium at dominated
airports. See also Lee and Prado (2003) who examine the effect of the mix of pmsengers by fare class on
reported hub premiums. They find that while a hub premium does exist, much of it is explainable by
passenger mix. Specifically, they find that failure to control for passenger type, inflates the average hub
premium by 11.9% to 20.8% for restricted coach passengers. They also find that die hub airports with the

50

 

fix

CE

l0

Cor
Passer

realized fiom alliance participation, then lower costs (Win in equation 2.4) should act to
lower the growth in the price gap between alliance and rival firms (ALLY in < 0). On the
other hand, if the market power effect dominates then the growth in the price gap
between alliance and rival firms will be positive (ALLY in > 0).

Thus while this method is comparable to that of Bamberger, Carlton and
Neumann (2001) reviewed earlier in Chapter 2 (Section 2.3.5) in that improved efficiency
and increased market power have opposing effects on fares, there are important
differences in the approach. First, the nature of the alliances they examine is different in
that code-sharing alliances were not system-wide arrangements. Thus their sample and
control could be selected on the basis of specific routes on which the alliance did and did
not operate, rather than at the firm level.

Second, they considered a large pool of firms, consisting not only of major
carriers but also of the smaller commuter carriers operating on the sample and control
routes of interest. 77 In fact, not only is information from smaller (and arguably distinctly
different) firms pooled with that of the larger carriers, a very large route network of these
carriers was considered. And third, they examine the impact of these code-sharing
alliances from before the alliances began to a year after they had been operating, while in

Section 3.7, analysis over a longer time period is defended.

 

largestpremiumsaresmallercitiesthatservethinnerroutesandthatuseaircrafithataremoreeostlyto
grerate. Also see Liu (2003).

Commuter carriers are defined as those that operate predominantly propeller-driven aircraft in scheduled
passenger service and in predominantly short haul service.

51

3.10 Evidence: First Diflerence Estimates of the Impact of A Iliance Formation on Fares

Table 3.4 reports the FD estimates of equation 3.4, after time invariant
explanatory variables were dropped.78 Parameter estimates in this table reveal that our
results are quite robust to the method of estimation. For all three alliances, the parameter
estimates of the explanatory variable of prime interest (that is, ALLYm) changed slightly
in magnitude between pooled OLS and FD, showing that controlling for unobserved
route and firm effects was somewhat important. However, it was again positive and
strong in magnitude. While standard errors were higher after first differencing, it
maintained statistically significance at the 1% level.

Comparing the parameter estimates of the Alliance Participation dummy variable
(ALLYm) from pooled OLS estimation (Table 3.3) with those from first differencing
(Table 3.4), for instance for the full effect for the US Airways-American alliance, shows
that the fare impact (lfarem) from participation in this alliance varied from 19% to 22%.
For Delta-United, the association between the dependent variable and the Alliance
Participation variable was lower after first differencing, varying from 13% from pooled
OLS estimation to 11% after first differencing. However, the strength of the evidence
remained largely unaltered: the consistently positive and statistically significant
parameter estimate of the Alliance Participation dummy (ALLYm) provides strong
evidence that alliance participation is responsible for generating fare premiums for the

participant carriers.79

 

7' Results are reported with heteroscedasticity adjusted standard errors and White t-statistics.

79 To check the sensitivity of our results to our method of selection of the control firms we also estimated
the impact of alliance participation, taking the members of the other two major domestic alliances jointly #
warranted at the time period under consideration. Results showed robustness to this alternative
construction and only one set of results is reported here: with the other four alliance carriers taken as
separate firms at each time period.

52

.882 885.838 8258a: 83.... .8. Ba an .8. 28 a “80883 83838 H... 85 a...

 

 

.8885 88.8 .3 wEEu 05 .«o noun—v.83 a .88 En 8283 .m 8830 cam 8:03.83 038:8, bean—835 28 8358.8 .5.“ N6 038,—. com SV
.55 2.88% .382 868880502 83 E

 

 

88 888 .8888. 2:888 - .8888 $-83
888.8 58.8 288.8
88 8.888 :88 888 - .288 8-32
58.8 888.8 888.8
88.: 8888 888 3888. - .888 «5-3
tuba :etoifieU
828.8 888.8 888.8
88 E88 888 - 28388 .888 33:
888.8 888.8 888.8
88 :888 8888 - 58.88- .288 8-32
588 8:88 8888
88.: :88 888 - .288 .838 £825
Ehnm 23500525:
82888 $88 @888 888.8
88.8 888 8888- .888 .8888 .888 58:
838.8 58.8 888.8 58.8
88.8 2888 8.88- 8888- :88- .888 8-32
2888 E88 58.8 888.8
83: 8888 8888 :288 .8888 .858 5-89 .85. :8..—
”xuozoz
z m .528 8 S s 5,2 8... 2,8 8. a§5> 88:88

 

[

 

 

53

While the Northwest-Continental alliance was the first to be publicly announced,
the US Airways-American alliance was the first to begin operations. The first column of
Table 3.4 shows that for each time period examined, the growth in the price gap between
the US Airways-American alliance and its major rivals was the greatest among the
alliances examined. For this, as for the other two major domestic alliances, results
indicate the absence of an 'umbrella effect' to benefit rival carriers.80

Table 3.4 also provides evidence that the realization of market power took place
even before these alliances were up and running: in the alliance announcement period, the
Alliance Participation dummy variable was positive and statistically significant for all
three major domestic alliances. From the first column of Table 3.4 it is seen that
immediately after alliance announcement, (average) fare growth of the Delta-United
alliance was 15% more than that of its largest rival carriers serving the same markets.“
This timing of growth in the price gap shows that increased market power was realizable
even at announcement of alliance participation.82 This could have been made possible
either since alliance announcement allowed the conversion of a former competitor to an
ally, or due to the advantages accruing to a firm now seeming to operate at a larger size.

After alliance completion, average fares of the Delta-United alliance grew at

about 5%,83 while the (average) fares of its rivals fell 2% (due to aggregate economy

wide factors). Thus the growth in the price gap between this alliance and its rival firms

 

8° Borenstein (1989) found that rival carriers did not benefit from higher markups of the dominant carriers,
while Kim and Singal (1993) had found that rival fare movements closely followed those of the merging
carriers.

8' Recall that while the Delta-United alliance was the last to be announced, it was quick to begin operations.
‘2 Similar results were found by Borenstein (1990) and by Kim and Singal (1993). The former found that
when Northwest and Republic Airlines merged in 1986, its fares were higher in routes to and from its
Minneapolis hub even before the merger. The latter formd that large increases in airfares in the
announcement period of firms that were merging but were not financially distressed.

‘3 This is calculated as 0.0751 -0.0247 for the Delta-United alliance in the alliance completion period.

54

had increased 7% between the fourth quarter of 1988 and the third quarter of 1999 and
this is noted in Table 3.4 as the parameter estimate of the Alliance Participation dummy

(ALLYm) in the alliance completion period.

3.1 1 Hub-Specific Evidence: First Difierence Estimates of the Impact of Alliance

Formation on Fares

Table 3.5 reports the hub-specific first difference estimates for the fare impact
from alliance participation. Pooled OLS hub-specific results are shown in Appendix E
for comparative purposes. Again, parameter estimates from the adoption of these
alternative methods of estimation consistently show robustness to this choice.

First, note the growth in the price gap between alliance and rival carriers at
Chicago O'Hare (0RD) and at Dallas Fort Worth (DFW). These two airports are 'double-
hubs' in that they are hub airports of both the Delta-United alliance“ and the us
Airways-American alliance.85

The growth in the price gap at Chicago O'Hare (0RD) due to the formation of
these two alliances was not statistically significant. At Dallas Fort Worth (DFW), while
the growth in the price gap was again not statistically significant due to the Delta-United
alliance at any of the three time periods, results for the US Airways-American alliance
were dramatically different there. In fact, parameter estimates of the US Airways-

American Alliance Participation dummy variable (ALLYin) were stronger there than at

 

8" Six of the nine Delta-United alliance hubs are considered concentrated. See US. General Accounting
Office, GAO/RCED-90-147 and US. General Accounting Office, GAO/RCED- 90-102.
85 In 1996, American and United owned 87% of the landing slots at Chicago O'Hare.

55

 

 

 

 

 

 

 

88888 85.8 888.8

8: 888 888- - 888 .888 65888850
58.8 82.8 888.8

8: 8888 828- - .188 8:8 A8388...

«ovum..— 23:50:23:
@888 888.8 888.8 9888

88 :88 888- .888 .888 .888 68886 3388
38.8 888.8 888.8 888.8

88 888 888- .838 .828 .8888 8,888.8er e8
8888 3.2.8 8888 28—8

888 8888 88:8- .888 .883 888 68088880
8888 888.8 5.88 E88

88 888 8.88 8.88 888 .828 88.888853
88888 888.8 888.8 888.8

88 888 8.88- 888 8:8 .828 558.289
888.8 888.8 888.8 888.8

888 8888 888- 8888 888 8888. 98883.5
888 888.8 888.8 888.8

88 888 888 .888- .888- .8888 2883.59
8888 888.8 288.8 888.8

88 888 288- 888 888 888 85888886
8888 888.8 888.8 888.8

88 888 :88- .888 .888 888 A8858?

team :8...

83.5-5.5...

8 5 >8... 8. 2.8 8. M2885 88:85

 

 

Ill

8.? 288 ms .8 8888.8 .858me 888 £8on £5 ”Wm 838.8

56

 

 

 

 

.888 88.8 88.8

88. 8888 2.88 88.8 - .8888 .98 a8....
.8888 .888 .888

88. 8888 888 888 - 888- .38.. 3......
.888 .888 888.8

.8... .888 8:8 :88- - 888 288.95..
88.8 88.8 .888

.8... 888 888- 3.88 - 8.88- 658888.26
.888 .8888 88.8

.8... 888 88.8- 88.8 - 588.8 .8833...

Baum— .3303an
.288 £888 .888

.8: 888 8:8- - 8.88 .888 .3886 8.388
.8888 .888 .838

88. 888 888- - 88.8 :88... 68888.88... :8
8.8.8 .8888 .888

88. 8888 888- - 8.88 828 .50888820
.8888 .888 .888

.8... 888 888 - 888 .328 83.8.8588
8.8.8 .888 ..8..8

88. 888 888- - 588.8 .888 688.2.
.888 .888 .888

88. 8888 888 - 888 888 .38.. 2......
.888 .888 .288

.8... 888 8:8 - 888- ......88 288.88..

2 “a .5880 no 8 5 >38. a. 28.8 me. "838...; 828.830

 

.888. 8 2.8

57

 

 

.Al-v.:: NJ» V. .MM

 
 

I’M

9‘53. -

 

 

 

 

 

 

 

 

.888 .888 .888

8: 2888 888 - 8888- 888 .5885856

week—H «fiasco—=59:
.388 .888 .888 .888

888 8888 288 828- .8888- :22...- 838.352
.888 .288 .888 .888

.28 888 8888 38.8 8888 .288 823.885.:
.888 .8888 .888 .288

M28 8888 8888 8888- 888- :88 3868.85:
.888 .888 .888 .8888

8.8 8.88 828 :82...- :~88. 82.8.8. .58 8388..
.288 .8888 .888 2.88

8: 8888 8888- 888- 8.88 8888- $8888..
.888 .888 .988 .288

M28 8888 .288 888 888 :88... .5885336

.85. :3.

”Baoflfiflth-uuéiehcz .m—
....88 .8888 .828

8v. 8888 88.8- 828 - .8888 .3886 8.38.8
.888 .288 28.8

82 2888 8888- :828 - 82... 68888.85... :8
.888 .8888 .828

82 888 888- .888 - 5.8 68.98885
:88 .888 .888

83 888 888 2888- - 2.....288 09.88.08.288

2 m. .888 8 8 s 5... .s. 28 8. 8.88.. .8888.

 

.888 E 0.88

58

 

i: v v =3»;

 

 

 

 

.888 .888 .888

82 2.88 288 :828- - :828- ...38 .8302
.888 .888 .888

88. 888 888- .828 - .888 .82888882
.888 .888 .888

8: 288 888- 288 - 2.88. 388882888:
.888 .888 .828

88. 888 888 :88- - .828- .58 8.88..
.888 .888 .888

88. 888 888- 888- - 888 88-88888.
.888 .888 .888

m8. 888 888 8.88 - 2.88.8 8.885.320

«ovum acre—mfiou
.888 .888 .8888

8888 888 - 5888- 888- .838 .857.
.888 .188 .888

82 8.88 888 - 8.88- .228 823.8888...
.888 - .288 .888

88. 8.88 888 888 2.5.8 82868288»:
.888 .888 .828

N8. 888 8:8 - 8.88- .888- .58 8.88..
.288 .888 .888

88. 888 8888- - 888- :8888- $88.88..

7. N8 .328 2. S s 58 .s a“... 8. 8.85., .5882.

 

.888 88 2.3

59

 

 

 

 

 

 

.888 .888 .288

88 888 828- - 888 :8828 .88 833032,.
.828 .828 .828

8. 888 828- - .2888- 888 5898820
588 .888 .888

882 888 888- - 888- 2.828 .82an3
.888 .888 .888

88 888 888- - 888 .8888 SEE Bag
.888 .888 .888

88 :88 828- - 888 .888 8888235

«ovum.— «nuance-5::
.888 .888 .888 .888

88 888 828- .3828 2:888 .828 .88 88388
.888 .888 .888 .888

88 888 8.28- 888 888 2.82.8 3828830328
.838 .888 .888 .888

88 888 828- 888 888- 8:28 5898820
.888 .888 .888 .888

88 888 :88- 888 $88- .828 23:83:
.888 .888 .2 28.8 .388

88 888 288- 888 888 .888 858883
.888 .888 .888 .888

88 288 :28- 888 888 .888 .5838?an

.88. :3.

Baotou—8738322 m: .U

2 N8 6388 8 8 s .58. .E 9388128885888

 

.828 2 23

60

.833 883.858 828882 83-... 8.: 88 8m .8 28 a .8888” 8.88858 H:- 85 a...

.8885 288 was: 28 8 83.88 a .8 3 888m 8 5820 88 88288 28:8 b29330 8.. «8 28-8 08 .8
.85 2858 882 883898238 83 .3

 

 

 

 

 

 

.888

88 888 828- :828 - :838 9885888
.888 .888 .888

88 888 828- 888 - .1828 .888 8328888
.288 .288 .888

82 888 888- 888 - 828 8898820
.888 .888 .888

882 888 828- 888 - :828 .528 2882
.288 .888 .888

88 888 828- :828 - .888 98,3833
.888 .888 .888

88 888 828- 888 - .8828 8888236

Baum 5328:30
.888 .888 .8888

88 888 828- - 2.8888 .888 $888888

2 mm 58280 .8 NO 5 >248. .5 28-: mo. 638.53 802539

 

8.88 2 288

61

any other hub airport.86 For instance, in the full effect period, the parameter estimate of
the Alliance Participation dummy variable shows that this alliance was associated with a
32% growth in the price gap between US Airways-American and its major rival firms on
some of the busiest routes in its network.

At the Northwest-Continental Minneapolis (MSP) hub, the growth in the price
gap due to the formation of this alliance was a strong 22%, indicating the dominance of
the increased market power effect there from the formation of this alliance.87 On routes
to and from Cleveland (CLE), formation of the Northwest-Continental alliance had

88 The only hub airport at which there was

resulted in a 7% growth in the price gap.
evidence of the realization of improved efficiency/cost synergy from alliance formation
was Newark (EWR).89 There, formation of the Northwest-Continental alliance had
resulted in a 13% drop in the price gap between this alliance and its major rival carriers
serving the same markets.

Thus overall, evidence in Table 3.5 shows that domestic alliance formation had a
positive and statistically significant fare impact at a majority of the alliance hub airports,

providing strong and clear evidence of the realization of increased market power from

these three alliances.90 The only exception was at Newark (EWR), a Northwest-

 

“ Except in the announcement period, when the US Airways-American Alliance Participation dummy
variable was slightly stronger at Charlotte than at Dallas Fort Worth.

‘7 Borenstein (1990) reported that by 1987, the merger between Northwest and Republic had resulted in
relative fares at Minneapolis to be 38% higher than industry average fares.

8‘ While low cost Southwest Airlines did create competitive pressures for the Northwest-Continental
alliance, it was only on a handful of markets with Cleveland or Detroit as one endpoint. See Boguslaski,
Ito and Lee (2002) for a detailed examination of the entry strategy of Southwest, along with predictions
regarding its future entry. For more evidence on the competitive effects from low cost/low-fare
competitors see US. General Accounting Office, GAO/01-518T.

’9 Pooled OLS results in Appendix E identify Houston (IAH) as the Northwest-Continental hub at which
improved efficiency was realized due to alliance formation. After the FD transformation though, the
Alliance Participation dummy variable, while still negative, was not statistically significant there.

9° Five of the six hub airports of the US Airways-American alliance were considered 'concentrated' by the
US. General Accounting Office. See US. General Accounting Office, GAO/RCED-90-l47 and US.

62

Continental hub where alliance participation was associated with the realization of

improved efficiency/cost synergies.91

3.12 Synopsis and Conclusion

The recent trend between the country’s most domith carriers of forming
alliances on domestic route networks seems to have replaced the mega-mergers that took
place immediately after deregulation. In this chapter we sought to answer the question of
whether the effect of each of these alliances been the realization of some efficiencies or
that of increased market power.

While pooled OLS estimation results provided strong and consistent evidence of
the overall dominance of increased market power from the formation of each of the three
major domestic alliances, the adoption of the first difference method provided results that
were largely quantitatively similar. That is, controlling for unobserved firm and route
heterogeneity had only slightly improved our insight into the relationship between
alliance formation and the dependent variable.

First difference estimation showed that on the busiest routes served by each of the
three major domestic alliances, alliance participation had a quantitatively important and
statistically significant impact on the (positive) growth in the price gap. These results not
only clearly demonstrate the dominance of increased market power from alliance

participation, but also that firms that had remained 'outside' such agreements had been

 

General Accounting Office, GAO/RCED- 90-102. See previous footnote for criteria employed for defining
an airport as 'concentrated‘.

9' Other than Memphis and Cleveland, all Northwest-Continental hub airports were already 'concentrated'
even before the alliance was formed. The US. General Accounting Office defines a concentrated airport as
one where one airline handled at least 60% of the enplaned passengers, or two carriers carried 85% of the

63

unable enjoy a fare 'umbrella'. The increased market power enjoyed by domestic alliance
forming firms could have been made possible either since this event allowed the
conversion of a former competitor to an ally (that is, there was a reduction in the number
of competitors by one), or due to the advantages accruing to the firm now operating at a
larger size.

The next obvious question is whether the noted trajectory in the growth of
alliance fares can be supported by any quality improvements taking place due to alliance
formation. Unfortunately we are unable to proceed in this direction, first since the O&D
data-base of the Department of Transportation does not provide this information on a
route specific basis. Second, while alternative sources of carrier service-data exist, for
instance the Bureau of Transportation Statistic's Air Carrier Statistics (Form 41 T raflic)-
T -1 00 Domestic Segment data-base, this data is incompatible with that of the O&D.92
These two constraints made this exercise beyond the scope of this paper and it proceeds
with an examination of the competitive impact of domestic alliance formation, armed
with an estimation method that allows improved control of unobservable influences on

the dependent variable.”

 

total enplaned passengers. See US. General Accounting Office, GAO/RCED-90-l47 and US. General
Accounting Office, GAO/RCED- 90-102.

92 This data base and the O&D data base are constructed under different criteria. Specifically, the former
provides information on the basis of non-stop flights while the later, on direct flights that may have more
than one stop.

93 The relationship between changes in quality and fare changes can also be difficult to interpret. For
instance, one common measm'e of quality in the airline industry is load factor, defined as the percentage of
seats filled. While on the one hand higher load factors can be expected to reduce fares due to lower costs
and may even signal lower quality due to greater crowding on the plane, on the other, high load factors may
be achieved during periods of high demand when fares should be higher. Similar ambiguity exists with a
measure of quality like circuity, defined as the ratio of the actual route distance to the distance flown on the
route. Thus some common measures of quality in this industry present us with the possibility of
controversial and ambiguous interpretations due to their multi-dimensional impact on changes in fares.

64

Chapter 4. Alliance Formation and Airport Dominance

In this chapter, the query of the competitive impact of domestic airline formation
is taken further. Specifically, this chapter aims to quantifying the impact of alliance

formation on airport dominance.

4.1 The Fare Impact of A irport Dominance

The strength and direction of the relationship between airport dominance and
fares has been well established in previous research on the airline industry. For instance
Evans and Kessides (1993) found that control of airport facilities confers the carrier
significant power over fares. Borenstein (1989) found that both route and airport level
dominance determine the degree of market power exercised by a carrier.94

This direct relationship between fares and airport dominance was reconfirmed on
our sample of routes and for the time periods corresponding to alliance formation: when a

95 was included as an explanatory variable in the price

proxy for airport dominance
equation 3.3, the magnitude of the Alliance Participation dummy variable (ALLYm,
defined earlier in Table 3.2) decreased. Specifically, the inclusion of a proxy for airport
dominance reduced the quantitative impact of the Delta-United Alliance Participation
dummy variable from 11% (noted earlier in Table 3.4) to 6%. The Northwest-
Continental Alliance Participation dummy variable fell flour 5% (in Table 3.4) to

negative 4%, and the US Airways-American Alliance Participation dummy variable fell

from 22% (in Table 3.4) to 3%. In all three cases however, it maintained statistical

 

9“ These papers were reviewed earlier in Chapter 2, Section 2.3.2.

65

significance at standard levels of significance. The dramatic decrease in the Alliance
Participation dummy variable (ALLYin) when a proxy for airport dominance was
included in the price equation shows that the dominant market power effect (ALLYm > 0)
found in the previous chapter, was power conveyed through the control of airport
facilities.96

Given this, it is useful to quantify the relationship between each alliance event and
the change in this (now familiar) source of market power. This is the basis of the
investigation in this chapter and is undertaken to provide greater clarity to the increased
market power result of Chapter 3.

Before proceeding, note that on the one hand, the direct relationship between a
carrier’s fares and its airport dominance may be explainable by the 'natural' benefits that
the dominant carrier enjoys due to its reputation. First, for instance, it may be seen to
offer better service, more fiequent flights etc.97 This advantage is enhanced by the
Frequent F lyer Program since travelers prefer to enroll in a program that will provide the
most destination choices at the time of reward redemption, along with a greater
possibility of having successfully accumulated the required mileage.98

On the other hand, airport dominance can create entry barriers: the dominant
carrier may enjoy bargaining power over airport authorities if it is an important source of
revenue for the airport and this may play a role in gate and slot allocation there.

Similarly, if the dominant carrier already owns the majority of gates/slots at an airport, it

 

9’ The proxy for airport dominance will be defined in Table 4.1.

9" In Chapter 3, the potential problem of the inconsistency of parameter estimates due to the omitted
variables problem is reduced by the adoption of the first difference estimation method.

97 Nako (1992) found that for business travelers, flight frequency had the largest impact on firm choice.
Toh and Hu (1988) report that these travelers value convenient schedules.

66

can refuse to sell or lease them to rivals wishing to expand and/or entrants wishing to

enter.”

4. 2 Explanatory Variables and Expected Signs

The (log of) Airport Market Share (AMin) is taken as the dependent variable and
it is defined as the carrier's average share of traffic100 on all markets it serves fi'om the
two endpoint airports of the route.101 Captured by Adm in the structural equation 2.4, it
proxies the carrier's airport-level dominance.

The vector of explanatory variables includes a constant plus:

1. The Alliance Participation Dummy variable, that indicates whether the firm is an
alliance member (ALLYin). It is defined as one when the firm is a partner of the

major domestic alliance in question, and zero otherwise. (Expected Sign: Positive).

2. T-l Time Dummy variables, to control for the effects of macroeconomic factors on

the dependent variable.

 

9“ The impact of the non-linear structure of Travel Agent Commission Override programs (TACO's) is also
important as it favors the dominant firm. Similarly, ownership of a Computer Reservation System (CRS)
firrthers the carrier's airport dominance advantage.

99 See US. General Accounting Office (GAO/RCED-90-l47) on entry deterrence by larger carriers due to
their ownership of gates and slots at important airports.

'00 Traffic is defined as the sum of all enplaned passengers, local and connecting.

‘0' This definition of airport market share is similar to that of Borenstein (1989). Borenstein (1991) and
Evans and Kessides (1993 and 1994) use a slightly different construction.

67

3. Route Market Share (MSm), defined as the percentage of all coach passengers
traveling on carrier 'i', route 'r', at time 't'.102 Given that the sample of firms is limited to
the ten largest domestic carriers (as justified earlier in Chapter 3, Section 3.4), Route
Market Share (MSm) is calculated as if these 10 carriers were the only contenders for
passengers on the sample of routes.

Note that to correctly assess the movements in the market share of the alliance
firm, alliance total post-alliance market share should be compared to their total pre-
alliance market share. This is since once two firms have entered into an alliance their
joint market share will necessarily be greater than it was of the individual pre-alliance
firms. For carriers remaining 'outside' the agreement, it is calculated as their simple

average market share (Expected Sign: Positive).

4. An explanatory variable to control for the impact of gate or slot unavailability
(Gate/Slob).

If carriers are constrained in their ability to expand or enter an airport, the changes
in their airport dominance there will be lower. For incumbent carriers, this can be due to
genuine limits on the number of gates or slots available for expansion, while for new
entrants, this could be a constraint if the incumbent carriers refirse to lease or sell
slots/gates in an attempt to limit rival firm expansion and entry at these airports. Both

these conditions create an expectation for a negative parameter estimate for this variable.

(Expected Sign: Negative).

 

'02 See for instance, Evans and Kessides (I993).

68

5. An explanatory variable to control for the geographical 'isolation' of at least one
endpoint airport (ARPT,). It is expected that such 'singular' airports will be associated
with lower airport dominance changes as carriers have less 'space' to expand: rival carries
and entrants interested in serving that city will all serve out of the same airport. Also,
such airports can have higher Passenger Facility Charges (PFCs) and/or higher delay
costs due to congestion problems. These conditions create an expectation for a negative
parameter estimate for this variable

Conversely, the lower expected competitive pressures due to the absence of such
a second facility could be associated with higher airport dominance. (Expected Sign:

Indeterminate).

4. 3 Method of Inquiry

As in the previous chapter, three specific quarters are selected: the quarter before
the alliance was announced in the press, the quarter after alliance agreement but before its
consummation, and the quarter nearly a year after the alliance had been operating. For
each major domestic alliance, Table 3.1 showed the actual quarter selected.

The timing of the change in airport dominance will be presented as the full,
announcement and completion effects. The announcement eflect captures the impact of
alliance formation on airport dominance between the first two quarters selected, that is,
between t = 1 and t = 2. Similarly, the completion effect captures the change in airport
dominance afforded by alliance participation between t = 2 and t = 3. The fill] efi’ect
considers the change in airport dominance due to the alliance in all three quarters, that is

fi'om before alliance announcement to a year into its full-fledged operations.

69

Dependent and independent variables used in this chapter are defined in Table 4.1

and variable summary statistics form Table 4.2.

Table 4.1: Variable Definitions ‘

Variable
AM in

ALLY in

MS in

Gate/ Slot ,

ARPTr

D2/D3

121ng in

RANKm

Definition
Carrier 'i"s average total market share from the two endpoint airports
of route 'r' at time 't', taken in logs.

Alliance participation dummy variable indicating whether the firm 'i'
is a member of a major domestic alliance at time 't'.

Total route market share of alliance member 'i' on route 'r' at time 't’.
Average route market share for rival carriers.

Dummy variable that indicates whether at least one endpoint of a route 'r'
is gate or slot constrained.”

Dummy variable that indicates whether at least one endpoint airport of a
route 'r' does not have second airport in the same metropolitan area.°

Time dummy variables for the second/third quarter.
The two period lagged MS in for carrier 'i' on route 'r' at time 't'.

The rank of the MS in of carrier 'i' on route 'r' at time 't', calculated in
descending order.

(a) See Appendix A for data description and screening.
(b) See Appendix C for list of airports within the sample that are gate or slot constrained.
(c) see Appendix D for list of airports within the sample that are not geographically 'isolated'.

70

Table 4.2: Variable Descriptive Statistics '

 

 

 

 

 

 

 

Variable Network Mean s.e. Min. Max
Full Effect
Fare in DL-UA 200.94 122.23 18 1757
NW-CO 212.56 125.96 19 2591
US-AA 217.68 118.77 15 1797
AM in (%) DL-UA 12.29 14.88 0.1 88.6
NW-CO 13.52 13.53 0.2 88.6
US-AA 14.35 15.91 0.2 96.7
MS in (%) DL-UA 19.09 27.1 1 0 100
NW—CO 19.10 27.53 0.1 100
US-AA 19.52 27.25 0.1 100
DISTr DL-UA 1 177.47 673 .20 30 2704
NW—CO 989.12 587.84 17 2565
US-AA 1021.65 655.15 67 2724
lagMS in (%) DL-UA 19.09 27.1 1 0.1 100
NW-CO 12.85 24.30 0 100
US-AA 12.27 22.99 0 100
Announcement Effect
Fare in DL-UA 202.40 122.37 17 1757
NW-CO 217.44 130.36 19 2591
US-AA 222.54 122.48 16 1797
AM in (%) DL-UA l 1.66 14.33 0.2 85.5
NW—CO 13.55 13.53 0.4 83.1
US-AA 14.31 16.1 1 0.4 84.1
MS in (%) DL-UA 19.45 27.54 0.1 100
NW-CO 19.05 27.56 0.1 100
US-AA 19.57 27.39 0.1 100
DISTr DL-UA 1 183.19 672.99 67 2704
NW-CO 989.12 587 .85 17 2565
US-AA 1021.65 655.16 67 2724
lagMS in (%) DL-UA 18.88 26.76 0.1 100
NW-CO 13.93 24.85 0 100
US-AA 14.2 24.98 0 100
Completion Effect
Fare in DL-UA 200.13 122.49 18 1757
NW-CO 210.35 121.19 19 1710
US-AA 210.19 114.77 15 1797
AM in (%) DL-UA 12.27 14.68 0.1 88.6
NW—CO 13.52 13.42 0.2 88.6
US-AA 14.3 15.63 0.2 96.7
MS in (%) DL-UA 19.46 27.51 0.1 100
NW-CO 19.34 27.72 0.1 100
US-AA 19.47 27.08 0.1 100

 

71

 

 

 

Table 4.2 (cont'd).
Variable Network Mean s.e. Min. Max.
DIST, DL-UA 1174.86 673.21 30 2704
NW-CO 989.12 587.85 17 2565
US-AA 1021.65 655.16 67 2724
lagMS in (%) DL—UA 19.46 27.55 0.1 100
NW-CO 1 1.1 1 23.09 0 100
US-AA 9.88 20.75 0 100
Number of Aliance Routes” DL-UA 691
NW-CO 476
US-AA 469
N (Number of Observations
on Network) DL-UA 18,657
NW-CO 12,852
US-AA 12,663

 

(a) See Chapter 3, Section 3.7 for time period description and Table 4.1 for Variable Definitions.

(b)Thedatasetisbalanced.

4. 4 Evidence: Pooled OLS Estimates of the Impact of A lliance Formation on Airport

Dominance

Table 4.3 bears the pooled OLS estimation results, where the impact of alliance

formation (ALLYm) on airport dominance (AMm) is noted after controlling for the

influences of gate constraints/slot controls (Gate/Slot,) and for the geographical 'isolation'

of at least one endpoint airport (ARPT,).

Parameter estimates in the first column of Table 4.3 show the dramatic positive

impact that alliance formation (ALLYm) had on changes in airport dominance (AMm) on

ma sample of some of the busiest airports in the country. '03 Specifically, and for instance

for the full effect of the Northwest-Continental alliance, participation in this alliance was

associated with a 52% increase in airport dominance.

 

'03 Appendix B lists the 45 busiest airports of the country.

72

 

 

.83.. 8:268 20.38%: .25 .8. 2a .8 .2. 2e a .5828. 2.8.8.2 u... e8 ....
.888. 28.6 was: 28.. 88:38 a é 5 88.8 .m .985 8m .2828... 23... 885.93 .8 3. .28. new A8
.55 325.. 8.8... 8833.232. 5.3 A3

 

 

 

A888 :88
88 $88 88.. .78.? . :8? .88.? .288 .888 7.7.8:
A888 A888 A888 A888 A888 A888
88 883 88.. .88.? - :88 :88? .888 8:2 8.32
A888 A888 A888 A888 A888 A888
88.: 888 88.. .288 - 88.? .28.? .888 .888 7.8.5
30cm mete—950
A888 A888 5.8.8 A888 A888 A888
88 .83 785 - .88.? 888 888 88.8 .888 7.7.8:
A888 A888 A888 A888 A888 A888
88 08.8 88.... - 88.? ...88.e .88.? .888 .88.... 8-32
A888 A888 A288 A888 A888 A888
8.: .83 e8: - .88.? 88.? .88.? .888 .888 75-3
«ovum wag—30:56::
A888 A888 A888 A888 A888 A888 A888
7.8.8 888 8.82 .88.? .88.? 8:8 88.? .888 .888 72-8.
A888 A888 A888 A888 A888 A888 A888
8.8 828 3.88 .88.? 88.? 2828 .88.? .888 .888 8.32
A888 A888 88.? A888 A388 A888 A888
8.8 888 8...: .88.? .88.? 3.8.? .88.? .888 .828 75.3 .85. 2....
819,302
2 m 8588 8 .277. .8285 e. 2 2......7. A... .28 8. 02.5., 2.886

  

    

  
  

 

 

...A.... .278 8.. .a 8.58 .80 38a .0 7 .28

73

 

 

 

 

 

Note that the magnitude of the Alliance Participation dummy variable (ALLYm)
noted in the full effect period was lead by its large changes in the period corresponding to
full fledged alliance operations (that is, in the completion effect) for all three alliances,
though the changes in airport dominance due to alliance announcement were also not
trivial by any standard.

The third column of Table 4.3 shows that the parameter estimates of the gate
constraint/slot control explanatory variable (Gate/Slot.) is consistently negative when
statistically significant, indicating that the scarcity of this physical resource had placed
limits on changes in airport dominance at such airports and over the time period
examined. 104

Parameter estimates of route market share (MS...) were unsurprisingly positive
and maintained statistical significance throughout, showing the strong and positive
association between route and airport level dominance. Parameter estimates of the
dummy variable controlling for the 'isolation' of at least one endpoint airport (ARPT,)
was positive when statistically significant, showing that when an airport enjoyed such
singularity in a given metropolitan area, that it was associated with greater airport

dominance.

4. 5 Some Econometric Issues

4. 5. I Collinearity Between Explanatory Variables
Collinearity between route market share (MSin) and the Alliance Participation

dummy variable (ALLYin) was suspected. In Appendix F, pooled OLS estimation results

 

'0‘ Evans and Kessides (1993) find that airport capacity constraints augment a carrier's market power.

74

of the impact of alliance formation on airport dominance (AM,,,) are reported if Route
Market Share (MSm) is excluded as an explanatory variable.

A comparison of the pooled OLS estimation results in Table 4.3 with those in
Appendix F shows that while the inclusion of Route Market Share (MSm) as an
explanatory variable in the former, reduced the quantitative impact of the Alliance
Participation dummy variable (ALLYm) and its standard error was larger, there was no
qualitative change: it remained both large in magnitude and significant statistically in its
association with airport dominance (AMin).

Pooled OLS parameter estimates of the Alliance Participation dummy variable
(ALLYin) in Table 4.3 range from 72% for the US Airways-American alliance to 51% for
the Delta-United alliance. A comparison of the magnitude of these parameter estimates
with those in Appendix F, indicates that changes in the dependent variable (AMm) are not
captured entirely by changes in route level dominance (MSin) but rather, the impact of

alliance participation on the dependent variable is important.

4. 5.2 The Omitted Variables Problem

Recall from Chapter 3 (Section 3.9) that pooled OLS estimation may result in
inconsistent parameter estimates due to the omitted variables problem. Specifically, this
estimation method ignores the impact of time constant unobserved effects on the
dependent variable. There may be some market specific and/or firm specific
characteristics that affect a canier's ability to dominate an airport, and these influences
have so far been relegated to the idiosyncratic error term (U in in equation 3.3).
Therefore, a method within the class of models of unobserved heterogeneity is

appropriate, where such unobserved effects are explicitly accounted for.

75

In terms of airport dominance changes, the basic model of unobserved
heterogeneity is:
ln AMirt = Xirt B + Uin+ Cir
(4.1)
Where Xi“ is the vector of regressors (listed earlier in Section 4.3), Ci, is the unobserved
heterogeneity that captures the qualitative unobserved and time invariant influences on

airport dominance. Um is the idiosyncratic error.

4. 5.3 Endogeneity of an Explanatory Variable

Endogeneity of Route Market Share (MSm) was suspected and confirmed on each
route network.105 Since the strict exogeneity condition on the explanatory variables fails
if Ci, (the unobserved heterogeneity term in equation 4.1) is correlated with any element
of the vector Xim, the first step is to adopt a more appropriate estimation method.

If the noted endogeneity of Route Market Share (MSin) is due to its correlation with Cir,
then this problem can be mitigated by the adoption of first differencing (FD), a method
within the class of models of unobserved heterogeneity. The FD method allows for
arbitrary correlation between Ci, and X“ and the FD transformation eliminates the

unobserved effect Cir.106

 

'°’ On the Delta-United alliance network, the estimate of the reduced form residual was 0.0248, with a t-
statistic of 87.03 and a p-value of 0. Endogeneity of route market share was similarly found on Northwest-
Continental and on US Airways-American alliance networks.

'°° Recall from Chapter 3 that the FD method falls within the class of models of unobserved heterogeneity
that allow us to control for the effects of unobservable firm level factors such as managerial quality, and
route level factors such as the 'mix' of passengers between leisure and business types, on the dependent
variable.

76

Recall fi'om Chapter 3 (Section 3.9) that the FD transformation lags the elements
of the dependent and independent variables and subtracts them. If two quarters are
considered, that is t = 2, then equation 4.1 becomes:

AlnAMin=62+AXmB+AUm
(4.2)

where A In AM,” = ln AMm - 1n AM“, H and 9 2 is the second period intercept.

4. 6 Preliminary Evidence: First Dijfirence Estimates of the Impact of A lliance

Formation on Airport Dominance

Table 4.4 shows parameter estimates after the FD transformation. Recall that
time invariant explanatory variables cannot be estimated as these are not distinguishable
from Ctr (in equation 4.1) and are dropped after the FD transformation.

First differencing parameter estimates in Table 4.4 re-confirm that alliance
participation (ALLYin) had a positive and large impact on the growth in the airport
dominance gap at some of the most important airports of the country. For instance, for
the Delta-United alliance, results in the first column of Table 4.4 show that in the full
effect period, the growth in the airport dominance gap between alliance participant and
non-participant firms was 31%.

While the sign of the parameter estimate of the Alliance Participation dummy
variable (ALLYin) fi'orn pooled OLS estimation (Table 4.3) is not different from that after
the FD transformation (Table 4.4), its magnitude is now much lower. For instance, for
the Northwest-Continental alliance in the full effect period, alliance participation was

associated with a 52% higher (average) airport dominance as compared to the 28%

77

growth in the airport dominance gap shown in Table 4.4. These large differences in the
magnitude of the parameter estimate of prime interest points to the importance of
controlling for unobserved heterogeneity at the route and firm level when examining this
relationship. The message though, remained unaltered: airline cooperation through the
formation of domestic alliances has lead to the widening of the gap in airport dominance
between firms entering these agreements and those remaining outside them at some of the
busiest airports of the country.

Table 4.4 also shows that that a positive and significant effect on the growth in the
airport dominance gap due to alliance participation (ALLYin) existed even before these
alliance had actually formed (that is, in the alliance announcement effect period).107 For
instance, the first column of Table 4.4 shows that just announcement of participation in
the Northwest-Continental alliance was associated with a 22% increase in the airport
dominance gap between this alliance and its rival carriers.”8 For US Airways-American,
the results are even stronger in magnitude: the announcement of this alliance was
associated with a 67% growth in its airport dominance gap.109

Comparing the pooled OLS parameter estimates for the explanatory variable that
is a proxy for the impact of route level dominance (MSin) in Table 4.3, with the first
differencing parameter estimates in Table 4.4 shows that its quantitative and qualitative

impact remained unchanged between the two estimation methods.

 

'07 Except for Delta-United, for which the Alliance Participation dummy variable was not statistically
significant in the announcement effect period.

' Recall from Chapter 3, Table 3.4 that the growth in the fare gap during the announcement effect period
for the Northwest-Continental alliance was not statistically significant.

"’9 These pre-alliance results are not surprising if passengers had responded to the 'benefits' offered by the
alliances even before they were up and running. One important way this may have been possible is though
the merger of the Frequent Flyer Programs: alliance forming carriers had declared their intentions to
acknowledge even previously accumulated mileage on their partner carrier’s flights. See "Marriages of
Convenience", Frequent Flyer, Page 16, July 1998.

78

.-

 

4. d 39/. z

r t

do: 8:353 reasoned .35. see one sen as; 2e a 28%? 88:83 N... one e...

.8985 2mm we mic—t 05 he agar—80¢ a .89 En sonoom .m cox—20 com .mcoumucou 033.9 bogus—axe com 3. 035. com :8

 

 

 

8.8.8 R8 8 88.8 88 8
$3 838 88.8- L58 - .888 .838 firms
£88 8388 88.8 38.8
88 82.8 $8.8- 88.8 - .888 882 8-32
38.8 88.8 8888 $8.88
88.: 888 88.8- .888 - .888 .828 5-3
30.5— note—aEeU
58.8 888.8 688.8 888.8
88 88.8 88.8- - 838 .388 .888 firms
2888 $28.8 88.8 58.8
88 882 83.8- - 8.8.8 .888 remade 8-32
8888 58.8 88.8 88.8
28.: 88.8 .888- - .888 .888 88.8- .355
«ooh—H «floaou==c==<
888.8 888.8 238.8 688.8 388.8
898 :88 828. .838 .828 .288 .888 firms
A888 888.8 388.8 A888 A888
88.2 28.8 :83. 38.8 88.8 .888 883 8-32
258.8 88.8 688.8 888.8 88.8
83: 8:8 82.8- .888 .828 E88 .888 STE .85.:
6102502
z 5388 8 8 e m: e. 5.2 As 23 mg Hosea; aoeeoaon

 

79

4 #
_.———- —__.. ._-......_.. -

def

me}
sir!
lnsfl

4. 7 The Persistent Endogeneity of the Explanatory Variable

The simple FD transformation provides consistent parameter estimates only if (the
changes in) Xi“ and Um (in equation 4.1) are orthogonal. But if the problem was
contemporaneous endogeneity, it will exist even after the FD transformation and the
parameter estimates in Table 4.4 will still not be consistent: strict exogeneity requires that
the explanatory variables in each time period he uncorrelated with the idiosyncratic error
(Um) in each time period.”0

The persistent endogeneity of Route Market Share (MSm) was established on each
alliance network.l ” Therefore, in order to mitigate this econometric problem, an
appropriate instrument is needed. A valid instrument (IV) is one that satisfies two
conditions: first, it should have some partial correlation with the endogenous explanatory
variable (MSm), given the other variables, and second, it should be orthogonal to the
idiosyncratic error (U in).

The two period lagged route market share (lag MS“. and defined earlier in Table
4.1) was selected as an appropriate instrument as it seemed reasonable to expect a
positive correlation between this and a firm's route market share (MSm) and of its
orthogonality to the time-varying error Um. I '2 The selection of this instrument indicates

the assumption that a two period lag is sufficient to control for any dynamic effects.

 

“° See Wooldridge 2002.

m For instance, on the Delta-United network, reduced form residual estimate was 0.02l7 with a t statistic
of 57 .44 and a P-value of 0. The test of endogeneity on the networks of the other two alliances yielded
qualitatively similar results. Note that a positive correlation between route market share and the time-
varying error will overestimate the impact of this explanatory variable on the dependent variable.

”2 An alternative instrument that can be employed for the endogenous route market share is its Rank,
defined earlier in Table 4.1. This is the instrument used by Evans and Kessides (1993) and is constructed
so that the firm with the largest route market share has a Rank of 1, while that with the smallest route
market share has a Rank of 8. Parameter estimates obtained with Rank as an instrument were qualitatively
similar to those in Table 4.4. Therefore only results using the two-period lagged route market slmre
instrument are shown and discussed.

8O

 

Estimation results after the FD transformation and the instrumentation of the

endogenous explanatory variable are discussed next.

4. 8 Evidence: First Diflerence—ZSLS Estimates of the Impact of A lliance Formation on

Airport Dominance

Given the lingering endogeneity of route market share (MSm) after the FD
transformation, the relation between Alliance participation (ALLYin) and airport
dominance (AMm) is examined with the two period lagged route market share (lag MSin)
as an instrument for the endogenous explanatory variable.

Table 4.5 shows these results on the three alliance networks, again as the
announcement, completion and full effects. A comparison of the results in Table 4.5 with
those afier the simple FD transformation in Table 4.4, shows that instrumentation of the
endogenous explanatory variable made a difference: parameter estimates of the
explanatory variable of primary interest (that is, ALLYirt) were smaller in magnitude in
Table 4.5. Standard errors after the simple FD transformation (in Table 4.4) were larger
than after pooled OLS estimation (Table 4.3) and larger still after the instrumentation of
the endogenous explanatory variable (Table 4.5). Between these three alternative
estimation methods however, parameter estimates of the Alliance Participation dummy
variable (ALLYm) remained both strong in magnitude and statistically significant at the

1 % level.

81

.83.. Ewe-e38 22,88? 85. .8. 8a .8 .8. as a 888:2... 8:88.858 a... o8 e...

.882... 2&8 mesa ea 8 88:88 a to. 2 888 .m .350 com 22:88 czar: baggage ea 2. 2.3 oom 28

 

  

- at... he...» 8.... Easiness-e5 e

 

 

 

 

8.288385 8 a... m3: wfl .6: Sci maze mafiafimm mAmN-ooaflo-EQ Em ”We 03$.

88.8
88 888 88.8- 288 - .288 8:18 firm:
5.8.8 2888 888.8 88.8
88 83 :38- 888 - .888 :338 8-32
888.8 88.8 88.8 88.8
88.: 8:8 88.8 .8888 - .888 .888 45-8
Baum note—nEeU
88.8 288.8 88.8 88.8
88 88.8 88.8- - .828 .888 e888 <88
88.8 58.8 888.8 88.8
88 88.8 88.8 - 88.8- .288 .838 8-32
888 A888 88.8 228.8
88.: 883 28.8- - .288 .888 :88? STE
«SCH uflOEOQEE¢=fi<
88.8 58.8 288.8 888.8 288.8
88.8 8:8 888. 888 8:8 .888 .828 5%:
288.8 58.8 838.8 8888 A888
8.2 88.8 .828- 88.8. 28.8- .:88 883 8-32
88.8 88.8 288.8 888.8
98.8 88.8 82.8- 888 .828 .888 .888 has-.5 .85 :5.—
Hug-5:502
2 am .588 B 8 s m: e. 5.2 a... 28 we. ”seem; “Becomes

82

The results in Table 4.5 again show that domestic alliance participation (ALLYm)
provided a direct and strong explanation for the growth in the gap in airport dominance
between alliance participant and non-participant firms, over the time periods considered
and on some of the busiest airports of the country. For instance, from the first column of
Table 4.5 and for the full effect of the Northwest-Continental alliance, results show that
participation in this alliance allowed the airport dominance gap between alliance and rival
firms to grow 14% between the three time periods examined.113

For the Delta-United and US Airways-American alliances, this parameter estimate
was even stronger in magnitude: the former alliance is associated with a 28% increase in
the airport dominance gap and the latter, with a 47% growth in this gap. Except for the
US Airways-American alliance, this result was driven by strong announcement effect
parameter estimates.

Next, Table 4.6 shows hub-specific F D-Two Stage Least Squares (ZSLS)
estimation results for the impact of alliance formation on the growth in the airport
dominance gap there. Parameter estimates of the alliance participation dummy variable

(ALLYm) were qualitatively similar to those discussed in reference to Table 4.5.l '4

 

”3 With respect to the statistically significant negative Alliance Participation dummy variable for the Delta-
United alliance in its announcement period, recall from Chapter 3 (section 3.6.1) that this is not
interpretable as the realization of improved efficiency, since in this period the alliance had not yet begun
o tions.

' ‘ At the Houston (IAH) hub of Northwest-Continental, alliance formation was associated with a decrease
in the airport dominance gap between alliance forming and rival firms. Recall that pooled OLS results in
Appendix E identified Houston (IAH) as the Northwest-Continental hub at which improved efficiency was
realized due to alliance formation.

83

.-L.

-..—.—-—-

 

 

 

 

 

 

 

 

988.8 82.8 98.8 88.8

8: 888 888 - 888 .888 :8388- 68:88:88
8888 822.8 888.8 282.8

8: 8838 888- - 8388 .388 .888. 9.8332

«ovum...— «Heaven-30::
88.8 288.8 88.8 888.8 88.8

8: 8888 888- 8288 3:8 .288 .888 62886 3388
5.88 88.8 28.8 588.8 988.8

82 82.8 888- ::88 5888 .888 .888 6888885 98
88.8 288.8 288.8 888.8 888.8

88 888 888- 8888- ....828 .888 :888 889888
8288 88.8 888.8 228.8 288.8

8: 888 888. 88 888 .888 .828 58888582
2888 8:8 8:8 888.8 83.8

82 888 8888- :8828 .888 .888 2.288 588:5
888.8 888.8 88.8 888.8 82.8

888 8828 888- 888 :828 2288 .888 23,58 3:3
888.8 58.8 88.8 888.8 882.8

8: 888 888- 8288- 828 .888 888- 2-28 recon
888.8 52.8 882.8 888.8 82.8

8: 8.88 888 888. 888 .888 888- 88888586
@888 88.8 888 888.8 288.8

88 8888 E88- 888. 888 .888 .8888- 3.58333.

seem—=8...

“women-23%

8 888 8 8 e m: e 5.2 8.. 28828285358588

 

 

“mesa-as at? 88 we 8?. we. 28 8:8 magnum .mmm~.ooeao85..eaae8o&m 88:88-82an

84

 

 

 

 

@388 $28 @888 £28

22 $28 $.38 $88 - .888 9.38 55:23
3888 92:8 @888 8828

82 $28 888. SS8 - .588 .838 05933
8888 $28 @888 3:8

3: 85.8 :88 $88- - .388 .388 2m8§89
@828 A328 5888 8888

3: $88 :88 388 - .388 .328 6R8 82586
8888 888 3888 82:8

8: 3:8 888. .888 - .888 .588 A8855?

«ovum 5:29.30
8888 888.8 @888 @888

3: :28 «~38. - 828 .288 :88- 68856 9388
8838 68.88 6:88 A288

NE :58 £88- - 3888 .8288 8888- 8.888.885 5m
@888 $388 $888 528

82 8&8 £88- - ~33 .388 $328- 58038820
838 8888 $888 $88

3: $28 $88- - 888 .888 288- 98.88.0353
2888 A888 2888 $.28

22 ~58 8888. - $38 .888 388. 85123
$88 888 2888 A828

82 228 :88- - £88 .8288 §88 SEE 8:5
9888 82:8 2888 £828

3: 3:8 888. - $88 .8888 3.8838- 2385on

z .8 2588 8 8 s m: s 5.2 2 28 we. azafi> 602085

 

.9888 3 use.

85

 

 

 

 

 

 

 

 

$88 8:88 8.88 $888

an: 388 $88- - 888- .288 8888 £535.35

«ovum «Hon-09:59::
@888 8888 98:8 8:888 8588

888 $888 $38. 288. 888 8888 55:8- 8388 83302
588 8388 €388 €888 8888

2:8 888 2:8- 8888 £88- .288 .888 538.88%:
38:8 @838 2888 8888 $28

«.28 228 8:8. «388 $888 .8888 £888 8282:2880:
€888 52.8 2888 8888 $3:

828 8888 828. 388 338 .888 2.2888 $28 8:88
$888 8888 $28 5888 38:8

88 83.8 88888- 8388. £88. .8888 8888 $3285
@888 8888 2888 88888 6888

mm: 85.8 8388- 8888. 888 .288 .3888 888885856

3.88.8.8

u—aaflcflflfiaUuunokfltcz .m—
8388 @2888 3888 88:8

3: 8:8 858- 888- - .8888 .888 68836 3388
E88 2888 588 88:8

22 38.8 888- 8.888 - .8888 .3888 888388888 Em
$88 @8888 2888 68888

882 89.88 8388- 8388 - .8288 .898 88988820
A888 €888 @838 $888

3: £888 888 £88 - .8288 .888 388.0828

2 N8 .836 8 8 s 82 8 >88 8 28 82 83.8; 88:85

 

8.8.8 3. use.

86

 

 

 

 

 

@8888 @8288 88888 8588

32 $88 38.8. 3.28. - $88 328. 958 .3302
58888 988888 88888 68:88

82: 8828 8828- 288 - .8288 .388 Emémzasqaz
$288 98288 88888 @2888

mm: EH8 3:8- 388 - .288 828 338.388.:
§ 288 @8888 $8888 2 £88

83: SR8 8888- £38 - .Nm88 .5838 $288.38:
83888 2:28 @8888 8388

83; 8838 N888- :38 - .288 $28 358885
38888 68888 58888 68.88

mm: 85.8 2888- 8888- - .388 588 58822086

Baum note—aficv
288888 5.828 38888 88:88

32 328 £88. - 8888 .588 838. Agmzaaoz
$8888 $28.88 58888 c.8888

mm: 8888 8:8- - 888- .288 .3838 523.8858:
98288 23:88 58888 @8288

mm: 3&8 $88- - 228 .588 .588 $86.23.:
E8888 $288 @8888 988888

8: 8838 :2? - 388. .288 8888 $388.82.
@3888 @288 @8888 $288

83: :58 888. - 828- .388 #88- 55882.5

2 .8 8.588 8 8 s m: s 5.2 As 23 88. us§> 888885

 

.9588 8.. 2.3

87

 

 

 

 

 

 

 

 

@3888 58888 8888888 @8888

882 $38 2:8 - .8288- .888 .888 3:88 «28.08.28
5888 33888 888888 28888

82 2888 228 - .288- .888 .8888 58988820
94888 88888 $88888 9.88888

82 $38 8:28 - .8288- .388 .3888 $28652
@3888 $8888 $8888 8: 888

ES :88 828 - 8888- .8888 N828 $388.39
62888 688888 8888.8 888.8

82 58 82:8 - .8288- :888 .38 P888236

Baum ufluaooflsaflfl<
@8888 888888 $88.88 388888 388$

SE @888 :28 .8288- .288- .588 .2888 982858.85
88388 @2888 $38.88 €88.88 @888

gm 8828 $28 $288. .8288- .3288 .388 Smafifiofizm
$58.8 $88.8 $8888 $88888 $88888

mom 3&8 :28 .8888- .888- .8888 .888 58988820
5888 $8888 @8888 €888.88 58888

a: 8:88 8.88 .8888- .8888- .588 .838 $2ng
9:888 $3888 $8888 38888 388888

888 838 $.28 :388 $888 .388 2288 8593.3
@8888 88888 88888 $88888 @8888

8%... $888 $28 .8888 .8888- .8288 .8838 @888235

.85 :8...

Baotou—<8??? m3 .0

Z NM 8:38.80 8 No 5 m2 5 >4»? 2 23 m2 6335/ 5888839

 

.9888 88 2.3

88

 

.88... 8.88.828 8.36088. 832 .82 85 .88 .8. 28 .8 2.82.838 2.888% N... 85 ....

.mowfifio 05 .«o MES: 05 mo 53:83 a he Em conoom .m .8520 8m 8.838%“. 033...? Deanna—axe 8.“ Nd 038,—. 8m 39
.895 28:38 $.52 303838.888»: 5:5 A3

 

 

 

 

 

 

 

88% 888888

882 8888 83:8 .8388- - .388 .8888 9828.385
5888 @8888 888888 88888

82 88:8 8828 .888- - .8888 .828 388882208228
88888 88888 88888 $8888

82 888 828 .8288- - .8888 .898 88988829
82888 88888 888888 @8888

82 £28 £88 .888- - .8888 .8888 Aszzsamz
68888 288.8 58888 $8888

:2 8888 $28 .388 - .8888 8888 9:833
@8888 @8888 888888 288888

82 888 828 .8288- - .888 .8838 $888235

Baum gum—$.50
28888 @8888 88888 388.8

82 8888 :28 - .8888- .:N88 .8888 Eafiaasa

z .8 .828 8. 8 s m: s :2 2 2s 82 82.58, 88585

 

.9888 88 22:.

89

4.9 Synopsis and Conclusion

Afier the satisfaction in Chapter 3 of the hypothesis of the attainment of
efficiency improvements versus the exercise of market power following the formation of
the three major domestic alliances, in this chapter the relationship between this event and
the changes in airport dominance that it had permitted was examined. The query had
proceeded in this direction since it was confirmed that the power over fare changes due to
alliance formation that was found in Chapter 3, was conveyed primarily through the
control of airport facilities. Specifically, it was noted that when a proxy for airport
dominance (AMm) was added as an explanatory variable in the price equation, the
magnitude of the Alliance Participation dummy variable (ALLYm) decreased.

In this chapter, parameter estimates from alternative methods of estimation
adopted to contend with some econometric issues, all continually conveyed the same
message: domestic alliance formation explained a dramatically increasing airport
dominance gap between firms participating in these arrangements and those remaining

outside them, at some of the most important airports of the country.

90

Chapter 5. Sub-Sample Analysis: Disentangling the Market Power

Results from the adoption of alternative methods of estimation in Chapter 3
yielded a consistent and strong answer to our primary hypothesis of increased efficiency
versus increased market power: on some of the most important routes in the country the
formation of domestic alliances, causing pairings between the six most dominant carriers,
had lead to these carriers enjoying the spoils of increased market power. The increased
market power realized by firms participating in domestic alliances could have been made
possible due to the alliance allowing a reduction in the number of rival firms by one, or
due to advantages accruing to a firm now operating at a larger size. This defines the

hypothesis of this chapter and the clarification of this obscurity is its main purpose.

5.1 The Price Equation

Recall that the price equation in Chapter 3 (section 3.1) was derived from the
model of profit maximization in Chapter 2 (section 2.1) and it relates the log of average
fares (lfarem) to alliance formation and to other variables that form the vector of
structural variables x...” That is,

In Pin=XinB+Uin
(5.1)
The explanatory variable of prime interest is the Alliance Participation dummy variable
(ALLYm), defined as one if the firm is a member of the domestic alliance in question and

zero otherwise.

 

”5 Dependent and explanatory variables were defined earlier Section 3.2 of Chapter 3 and in Table 3.2.

91

5. 2 Hypotheses Tests on Route Sub-Samples

The full route sample will be divided into overlapping and non-overlapping
routes. The former are defined simply as those routes on which both firms participating
in an alliance operated during each time period considered. Similarly, non-overlapping
routes are those on which at most one firm participating in the alliance operated at each
time period considered.

On our sample of routes between each hub airport and the 45 busiest airports of
the country, the majority of the routes were served by one or the other alliance partner:
only about 19% of the total sample of routes was served by both members of the Delta-
United alliance during the time periods corresponding to alliance announcement and
formation. For both Northwest-Continental and US Airways-American, overlapping

routes were only 3% of our total route sample.

5. 2. 1 Efliciency versus Market Power

First, it will be interesting to see whether any notable magnitude changes take
place in the Alliance Participation dummy variable (ALLYm) on these route sub-samples
from those noted earlier in Chapter 3 (section 3.10) for the full route sample. Thus our
primary hypothesis of efficiency versus market power (detailed earlier in Chapter 3) will
be re-examined here for the two most important route sub-samples. If the magnitude of
ALLY“. increases, for instance on alliance overlapping routes from that noted earlier in
Chapter 3 (section 3.10) for the full sample, it will show that domestic alliance
participation was more effective in increasing the growth of the price gap on their

overlapping routes than it was on the total sample.

92

It is important to note at this point that the potential for the realization of
improved efficiency/cost synergies or of increased market power exists on both
overlapping and on non-overlapping routes.

For instance, on their overlapping routes, alliance formation has a potential for the
creation of efficiencies through the reduction of redundancies between the partner firms.
The dominance of this effect (captured by a lower ch in terms of our structural model of
profit maximization in Chapter 2) should lower the growth of the post alliance price gap
(that is, ALLY,“ < 0) between alliance forming firms and their rivals.

On the other hand, it is on these overlapping routes that there has been a direct
reduction in the number of competing firms. Here, the power to raise fares after alliance
formation could have been derived by the conversion of a former competitor to an ally,
that is, due to the reduction in the number of rival firms by one. The two firms that
formerly competed directly may now have greater opportunity, incentive and power to
collude on their overlapping markets, especially if the services of the two alliance
member firms are now perfect (or near perfect) substitutes to an important (and lucrative)
segment of demand. Evidence of the dominance of these factors (captured by a higher
Adm in terms of our structural model of profit maximization in Chapter 2) will be a
positive Alliance Participation dummy variable (ALLYm > 0) showing a positive grth
in the price gap between alliance forming firms and their rival carriers.

On their non-overlapping routes, alliance membership advantages can accrue due
to the firm now operating at a larger size or through multi-market contact. There may
even be some scope for the realization of cost synergies to the extent that the distinct
routes are being served from common airports and where on-ground cost synergies can

be feasibly achieved.

93

 

5. 2. 2 Size versus Concentration

The test of the sub-sample hypothesis requires a comparison of the sign and
magnitude of the Alliance Participation dummy variable (ALLYm) on the two route sub-
samples. The parameter estimate of the Alliance Participation dummy variable (ALLYm)
may be greater in magnitude on non-overlapping routes than when the sub-sample is
restricted only to alliance overlapping routes. This will show that since domestic alliance
formation had a greater impact on the growth in the price gap on their complementary
route system, that the greater market power was derived due to an increase in the size of
the firm. l 16

Conversely, the parameter estimate of the Alliance Participation dummy variable
(ALLYm) on overlapping routes may be greater in magnitude than that on their non-
overlapping ones. This will show that since the increased market power was found on
routes that both alliance members served, it was due to the conversion of a former

competitor to an ally, that is, through a decrease in the number of firms by one.

5. 3 Method of Estimation

In Chapter 3 (section 3.10), the potential inconsistency of the pooled OLS
parameter estimates due to the omitted variables problem had lead to the estimation of the
price equation using first differencing (FD), a method under the class of models of
unobserved heterogeneity. Thus, this will also be the method of estimation adopted in

this chapter.

 

”6 Since the dominant effect found in Chapter 3 on the full sample was of increased market power, finding
that the Alliance Participation dummy variable on non-overlapping routes was greater in magnitude than
that on overlapping routes, translates into the dominance of the market power effect on the former.

94

Recall from Chapter 3 (section 3.9) that in term of our price equation (equation
5.1), the basic model of unobserved heterogeneity in which unobserved (route and firm)
effects are explicitly included is:
In Pin = Xirt [3 + Um+ Cir
(5.2)
First differencing (FD) lags the elements of the dependent and independent variables and
subtracts them. If two quarters are considered, that is t = 2, then
Alnpm=92+AXinB+AUirt
(5.3)

where A In Pm = ln Pin - 1n Pin (-1 and 0 2 is the second period intercept.

5. 4 Sub-Sample Evidence: First Difference Estimates of the Impact of A lliance Formation

on Fares

Variable definitions are the same as were noted earlier in Table 3.2. Variable
summary statistics by route sub-sample form Appendix G. This appendix shows that for
instance, for the time period corresponding to the event, the average fare of the Delta-
United alliance on their on overlapping routes was $213.94, while that on their non-
overlapping routes was $198.09. For this alliance, the average trip distance was 1585.91
on their overlapping routes, and 1098.02 miles on their non-overlapping routes.

Table 5.1 shows the estimation results after the FD transformation and when the
total sample is divided between alliance overlapping and non-overlapping routes. It

shows that the sign of the Alliance Participation dummy variable (ALLYm) consistently

95

 

 

 

 

 

 

888888 88288 @8288
88 388 8:88 - .888 828 .388 858.85
«ovum 2.0835552.
@2888 8.88888 @8888 628.8
.282 2888 8888 E88 .288- .888 .388 88885-82
8.8888 282.88 2.288 88288
2. 3.88 2.88 .828. 8288 2.28 .388 8.88.85
.888 :8...—
"3.88.80.82.82
@2888 3888 9.888
8288 8888 888 8.88. - .888 .388 8888.582
5888 8.888 @888
28.8 .288 288 .888. - .328 .388 888.85
.888 88.880
@2888 @2888 2.888
.8282 8288 8888 - 8:88 .8828 .388 88885-82
$888 9.8888 28.888
28.8 3888 $88 - :88 2.28 .388 888.85
«SCH «EOE-00:36::
£2888 @8888 @2888 92888
88.2 8.888 8.88 388 8888 8.28 .388 88885.82
3888 @2888 $888 @8888
28.. .288 88.88 8888 .288 .228 .388 888.35
8388 :8...
”885......—
z 8 38.80 8 8 s 5.2 8.. 388 8. 388> 88.89

g

.. 8888.88 388 :8 .8. 3.8 82.8 .3888 888.85 88 8.. 388.

96

 

 

 

 

 

97

88888 9.288 82288
88 888 8:88 - .888 828 .388 88885
vacuum “EOE-09:56::
@2888 58888 88888 @2888
.282 2888 88888 .888 .288- .888 .388 88885-82
28888 28288 28288 88888
E. 3.88 2.88 828- 8288 23:8 .388 88885
.888 .8...
"—5floéfloUuumoBAP—OZ
@2888 3888 $8888
8882 .888 888 8.88- - .888 .388 88885-82
5888 8.888 @888
88.8 2288 .888 888- - .828 .388 88885
Eur—H note—9.89
$2888 @2888 98888
88.2 8288 8888 - .888 .828 .388 88885-82
5888 98888 288.8
28.8 .888 8888 - :88 .828 .388 88885
«ovum afloaooflscflfi<
$2888 38888 $2888 32888
88.2 8888 8.88 888 8888 8.28 .388 88885-82
5888 88.888 @8888 68888
28.. 8288 888 8888 .288 .828 .388 88885
888......
"8.882-385
z 8 .888 8 8 5 >3... 1.3.: 8_ 388$ 8.8.85

{
. .moapambsm 3:3— 88 .68 28.3 8: Mo 88:55 8880th 8.5 ”mm 838,—-

indicated the dominance of the increased market power effect from the formation of each
of the three alliances on both route sub-samples.

Comparing sub-sample regression results in Table 5.1 with those for the full
sample in Table 3.4 shows which sub-sample lead the increased market power results of
Chapter 3 and that were noted in Table 3.4 for each of the three domestic alliances. For
instance, for the US Airways-American alliance and for its full effect, the market power
result of Chapter 3 was lead by the growth in the price gap on routes that were not jointly
served by both US Airways and American Airlines. For Delta-United, evidence of the
realization of increased market power that was noted earlier in Table 3.4, is now seen
through the division of the full route sample into sub-samples, to have been driven by the
growth in the price gap on routes that both Delta Airlines and United Airlines served.

Next, in examining the sub-sample results in terms of the timing of fare changes,
Table 5.1 shows that for the Delta-United alliance, it was in the announcement period and
on their non-overlapping routes that the alliance participation had a stronger impact on
the growth in the price gap. Specifically, the announcement of the Delta-United alliance
was associated with a 15% increase in the price gap. This quantifies the advantage
created just by alliance announcement for its member firms, due to this new 'firm' now
seeming to operate at a larger size, or through multi-market contact. In the post alliance
formation period though, the parameter estimate of the Alliance Participation dummy
variable (ALLYm) showed a stronger impact on the growth in the price gap on routes that
both member carriers did jointly serve. Overall, this effect dominated and allows the
conclusion that the market power derived by the Delta-United alliance afier alliance

announcement and a year into full fledged operations, was due to an increase in

98

concentration on these busy routes, that is, through the reduction in the number of rivals
by one.

Table 5.1 shows that for the US Airways-American alliance, the member firms
were also able to exercise greater market power even during alliance discussions and on
routes that both did not jointly serve. Specifically, immediately after the announcement
of the US Airways-American alliance, the price gap between alliance participants and
their major rivals grew by 22%. This result can be explained by the advantages created
for the firm now (securing to) operate at a larger size, or through the workings of multi-
market contact.

The overall results for the Northwest-Continental alliance are qualitatively similar
to those noted above for the US Airways-American alliance. Specifically, the
dominance of the increased market power effect shown earlier in Table 3.4 was lead by
the growth in the price gap on routes that both member carriers did not jointly serve. A
notable difference was seen between this and the US Airways-American alliance in terms
of the timing of the fare changes. Results in Table 5.] show that during the
announcement effect period, the growth in the price gap due to the announcement of this
alliance was not statistically significant: alliance announcement was not a significant
source of any growth in the price gap. Recall from Chapter 1 (section 1.4) that this
alliance was the first to be announced and that the generous degree of integration planned
between Northwest Airlines and Continental Airlines had attracted concern by both
regulators and journalists alike. Table 5.1 shows that it was only after the alliance began
operations that there was an increase in the price gap. Parameter estimates of the
Alliance Participation dummy variable (ALLYm) during the completion effect show that

the completion of this alliance increased the price gap by 7%.

99

That a dominant market power effect can take place even on routes that partner
carriers do not jointly serve is especially interesting since in the past, the permissibility of
these alliances has been judged by regulators on the basis of the degree of overlap
between the carriers.117 In this context, Table 5.1 reveals that the Delta-United and US
Airways-American alliances, the first two to begin operations, exercised their increased
market power even during alliance discussions on routes on which they did not both
serve.

Next, Appendix H shows pooled OLS estimation results of the fare impact
(lfarem) from alliance participation (ALLYin) on these route sub-samples. Pooled OLS
estimation allows us to examine the relation between alliance participation and fare
changes afier controlling for the scarcity of two important (time invariant) inputs, that is
the scarcity of gates/slots (Gate/Slot.) and the absence of a competitor airport in the same
metropolitan area (ARPTr).118 Estimation results in this appendix show the robustness of
our route sample-specific conclusions to the method of estimation: the Alliance
Participation dummy variable (ALLYm) retains its positive sign and dominant effect on
the same route sub-sample as was noted earlier in Table 5.1.“9

Note also that in Appendix H, the parameter estimate for the airport isolation
variable (ARPTr) variable was negative for the Delta-United and US Airways-American
alliances, showing that the absence of another airport nearby prevented these alliances
from segmenting demand by elasticity, a result dominant on their overlapping routes.

A comparison of the parameter estimates in Table 5.1 with those in Appendix H

shows that they are generally quantitatively comparable and qualitatively similar. Thus

 

”7 See for instance, US. General Accounting Office RCED-99-37-
”3 See Table 3.2 for definitions of explanatory variables.

100

t -

/—__‘

results show robustness to the method of estimation and the overall message is the same:
participation in one of the major domestic alliances allowed members with a fare growth
greater than those of firms remaining outside these agreements. Even on the route sub-
sample on which there was a large potential for the realization of cost efficiencies (that is,
on alliance overlapping routes) there is no evidence that such an effect was ever realized.
F D-Two Stage Least Squares estimates of the impact on the growth in airport

dominance on these sub-samples due to alliance formation, are shown in Appendix I.

5.5 Synopsis and Conclusion

This chapter began with the hypothesis that in examining the direction of the
growth in price gap of the firms participating in one of the three major domestic alliances,
if on the one hand greater market power dominance was noted on their overlapping
routes, then this will show that market power was lead by a change in the availability of
travel options. That is, due to the conversion of a former competitor to an ally or a
decrease in the number of firms.

On the other hand, evidence of a dominant growth in the price gap on alliance
non-overlapping routes suggests that demand complementarities (or multi-market
contact) between alliance forming carriers were more important. That is, that the market
power noted earlier in Chapter 3 was derived from the advantages accruing to a firm now

operating at a larger size.

 

”9 Except for the announcement effect for Delta-United in which pooled OLS estimation shows a dominant
alliance effect on alliance overlapping routes.

101

Results shown in this chapter indicate strong evidence in support of the market
power hypothesis on both sub-samples and for all three alliances. '20 Even on routes
where there was the greatest potential for the realization of cost synergies that is, on
overlapping routes, neither of the three alliances exhibited its realization at any period
fi'om before alliance announcement to a year into its operations. In fact, an overall
dominant growth in the price gap on overlapping routes for the Delta-United and
Northwest-Continental showed that their increased market power was derived from a
reduction in the travel options available to travelers. While for the US Airways-
American alliance, overall, demand complementarities or multi-market contact were

more important for the realization of increased market power.

 

'20 Recall that the Department of Transportation has previously approved airline alliances on me grounds
that the allying carriers serve largely non-overlapping and therefore unrelated markets.

102

Chapter 6. Summary and Conclusions

This paper primarily investigated (within the general framework of Bresnahan,
1989) the price behavior of the three major domestic airline alliances that were
announced in early 1998, causing pairings among the six most dominant carriers in the
US. airline industry. The primary hypothesis examined was whether these agreements
had resulted in improvements in efficiency or in the realization of increased market
power.

In Chapter 1, some of the structural and partnership differences between
international and domestic airline alliances were highlighted. These differences would be
the basis for understanding the empirical results of previous research examining
international alliances in the post deregulation airline industry. For instance, in Chapter 2
it was discussed that consumer benefits are realizable from code-sharing between
international carriers since their itineraries generally have an interline feature. Also,
international alliance member networks are more complementary in nature and some
post-alliance network realignment may have eliminated redundant routes among the
partners. 12' This discussion formed the basis for understanding the distinctly different
results when examining domestic airline alliances.

Chapter 1 also discussed that the major domestic alliances were announced at a
time when many of the industry’s major players were both struggling financially and

122

competing aggressively with other domestic carriers. This dichotomy is an important

 

'2' The Northwest-KLM international alliance, formed in 1991, had this feature with Northwest abandoning
some European routes and adding some routes fi-om non-hub American cities to Amsterdam.

'22 Anti-competitive concerns continue to be part of the analysis of this industry, as major carriers are
inclined to exert discipline when threatened with rival entry. For instance, in May I999 and under Section
2 of the Sherman Act, the Department of Justice sued American Airlines from attempting to monopoliu
through predation, service to and from its Dallas-Fort Worth airport.

103

 

aspect of the industry. A number of empirical studies on the domestic airline industry
reviewed in Chapter 2, provided evidence of the increased market power that had already
resulted from carrier mergers or though high airport dominance, and the financial
difficulties of the industry were only exacerbated in late 2001 when it became obvious
that this cyclical industry suffered not only from cost inefficiencies but was also now
vulnerable to a set of international political motivations.

Regression results in Chapter 3 and Chapter 4 provided a quantification of the
anticompetitive impact of domestic alliance formation that had taken place in this
environment. Our results provide strong and consistent evidence that increased market
power dominated each domestic alliance event. Specifically, the parameter estimate of
the Alliance Participation dummy variable ranged from 4% for the Northwest-
Continental alliance to 22% for the US Airways-American alliance, results that were
statistically significant at the 1% level and that provide evidence that each domestic
alliance had resulted in the increase in the price gap between alliance carriers and their
major rivals. Results from an alternative method of estimation were qualitatively similar.
The public policy implications of these results are in general against the permissibility of
cooperative arrangements between two firms that were once each other's biggest rivals,
and in particular, against the recent trend of domestic alliance formation that has created
an additional layer of legitimate anti-trust and regulatory concern.

Next, given that previous literature on the airline industry has established the
strong positive association between airport dominance and fares, a quantification of the
association between the alliance event and the change in airport dominance was sought.
This formed Chapter 4 and there, alternative methods of estimation again provided

qualitatively similar and quantitatively strong results. These results were shown in Table

104

4.5 where due to the persistent endogeneity of an explanatory variable, the first
difi‘erence-ZSLS estimation method was employed. Parameter estimates ranged fi'om
14% for the Northwest-Continental alliance to 47% for the US Airways-American
alliance, providing evidence that alliance formation caused a strong growth in the airport
dominance gap between alliance participants and the firms remaining outside these
agreements.

Then in Chapter 5 (Table 5.1), a dominant Alliance Participation dummy variable
was noted for the Delta-United alliance on overlapping routes, indicating that its market
power effect was lead by a change in the availability of travel options, that is, due to the
conversion of a former competitor to an ally or due to a decrease in the number of firms.
On the other hand, a dominant Alliance Participation dummy variable on non-
overlapping routes shows that demand complementarities (or multi-market contact)
between the carriers were important and that market power was derived from the
advantages accruing to a firm now operating at a larger size. This result was found for
the US Airways-American and Northwest-Continental alliances.

The results shown in this paper provide the answer to a question that has
confounded regulators concerned with the anticompetitive impact fi'om the increasing

123 While regulatory agencies have been in what seems like a

'cooperation' in the industry.
constant state of inquiry of the industry's conduct, to the best of our knowledge there is
no other detailed empirical research addressing this issue. An interesting extension of

this work would be to examine the fare impact of domestic alliance formation on

 

‘23 An interesting extension of this work would be to examine the fare impact of domestic alliance
formation on different consumer 'types', that is, on leisure versus business travelers. To the extent that
these alliances were formed to better retain the demand of this lucrative segment, this extension is an
important one. The analysis could be also be extended to include a larger set of airports. Further, a
clarification of why different sub-sample results were seen across the three domestic alliances is needed.

105

different consumer ‘types’, that is, on leisure versus business travelers. To the extent that
these alliances were formed to better retain the demand of this lucrative segment, this
extension is an important one. Further, a clarification of why differences in sub-samples
results were found is also needed.

While the timing of their fare responses highlighted some differences between the
individual alliances, the uniformity of the increased market power result supports the
generalization of arguments against each of these arrangements. However, the results of
this paper and the arguments raised in it need not be in conflict with those raised to
improve the financial strength of the industry. Instead, these results not only underscore
the need to enhance and preserve competition in this industry, but also for sustaining and
extending conditions that force improvements in firm level efficiency. While
partnerships do reduce actual and potential competition by their very nature, it is

important that they also create additional efficienciesm

That is, cooperation need not be
in conflict with the ideals of competition and firms should be permitted cooperation with
competing firms if the aim and result is an improvement in consumer welfare. However,
there is a need to distinguish between these measures and those that allow firms the
power to achieve and exercise market power through various antitrust loopholes. The

former refers to measures such as the airline financial stabilization package that rushed

through Congress in the weeks following September 1 1, 2001 and the latter, to the recent

 

'2‘ See Kroszrrer, Mullin, Jaffe and Alexander (2002) for a discussion of various organizational forms with
motivations and effects similar to those of outright mergers.

106

 

decision by the Department of Transportation to allow further 'consolidation' of this

industry through even wider domestic alliances.125

 

’25 See Atlanta Journal-Constitution, "Delta to Network with Rivals. Carrier Expected to Joint Northwest,
Continental Airlines" August 23, 2002. Also, The New York Times, "DOT Approves United-US Air
Code-sharing' , October 3, 2002.

107

APPENDICES

108

Appendix A: Data Base Construction.

A record in the Origin and Destination Survey (DBlA) is an observation with the
carrier, time, origin and destination airport, itinerary and fare. A number of restrictions
are applied to this large database. We consider only tickets representing trips outbound
from the base airport. Additionally, the following standard screens are used to remove

records from the database:

1. Records with non-credible fares that suggest reporting errors. This screen is based on
yields for which mileage categories are the same as were used by the Civil
Aeronautics Board and the General Accounting Office.

2. Open jaw tickets that are neither one-way, nor round trips.

3. Interline tickets, that is trips that involve travel on more than one carrier.

4. Non-coach fares, that is all business and first class tickets, except for Southwest,
which reports all tickets as first class.

5. Zero fare tickets.

6. Tickets that are not direct or have more than one stop between the trip origin and
destination.

7. Routes with at least one endpoint outside the continental United States.

109

Appendix B: The 45 Busiest Airports.
The following are the 45 busiest airports of the country, ranked by 1999 passenger

enplanements.

1. Atlanta

2. Chicago- 0 Hare
3. Los Angeles

4. Dallas/Ft. Worth
5. San Francisco

6. Denver

7. Minneapolis

8. Detroit

9. Miami

10. Newark

11. Las Vegas

12. Phoenix

13. Houston

14. New York- JFK

15. St Louis

16. Orlando

17. Seattle

18. Boston

19. Philadelphia

20. New York-La Guardia

110

21. Cincinnati

22. Charlotte

23. Salt Lake City

24. Washington (Dulles)
25. Pittsburgh

26. Baltimore

27. Tampa

28. Washington (Ronald Reagan)
29. Fort Lauderdale
30. Portland

31. Chicago - Midway
32. Cleveland

33. San Jose

34. Memphis

35. Oakland

36. New Orleans

37. Raleigh

38. Houston

39. Nashville

40. Indianapolis

41. San Antonio

42. Dallas

43. Austin

44. Columbus

111

45. Albuquerque

Source: ACI Traffic Data: World airports ranking by total passengers (1999).

http://www.airports.org/traffic/td_passengers

112

Appendix C: Gate Controlled and Slot Constrained Airports

AirporttCode)

l.

2.

>0

Charlotte (CLT)

Chicago O'Hare (ORD)126
Cincinnati (CVG)

Detroit (DTW)
Minneapolis (MSP)
Newark (EWR)

New York (JFK)

New York (LaGuardia)

Pittsburgh (PIT)

10. Washington-Reagan (DCA)

Type of Con_s£arn'_t
Gate
Slot
Gate
Gate
Gate
Gate
Slot
Slot
Gate

Slot

Source: US. General Accounting Office, GAO/T-RCED-98-112.

 

'26 The High Density Rule was lifted at Chicago, O'Hare in July 2002.

113

Appendix D: Cities Served by More Than One Airport”

1. Chicago (O'Hare and Midway)

2. Dallas ((Ft. Worth and Love Field)

3. Detroit (Metro and City)

4. Houston (Bush and Hobby)

5. Los Angeles (Los Angeles International and Burbank)

6. Newark (JFK and LaGuardia)

7. New York (LaGuardia and Newark)

8. San Francisco (San Francisco International and Oakland)

9. San Jose (San Jose International and San Francisco)

10. Washington (Dulles and Baltimore)

* Secondary airport must fall within the top 45 busiest airports of the country to qualify

consideration in our sample of routes.

Source: OAG Desktop Guide, June 1999.

114

 

 

 

 

 

 

828 388 A288 2888

83 888 8888 - - .828 - 888. .828 5522.885
:88 2888 9.88 6888 2:88

8: 828 8.8.. - e888 .8888 - 888- .8888 288888...

week—H ufluaouflflcfl:
2888 2888 8288 22.888 2888 888

828 .828 3.... - .8828 .288 8.88. .8888 .888 6.2885 32.-28.8
2888 2888 8.88 22888 2888 2888 2888

82 .888 88.2 .888- .888 .828 .8888- 288 .828 5888.88... 88
2828 22.888 2288 22.88 2888 22.888

888 888 288.2. .8888- - .8328 R88 .888. 2288 88588.85
28.28 2888 8.88 2888 2888 2888 2888

82m 888 .82 .8228- .2828 .888 8888- 8888 .2888 288.882.8592
2:888 2288 2888 E88 2.88 28.88 88288

82 8:8 282.. :88- .8228 .888 8888- .288 .828 58.325
2828 2888 62.88 2888 28888 2888 A888

888 2828 288.8 8888 .888 .888 888- .8288- .828 2388 822.5
@888 8888 $88 2888 c888 2888 2888

82m 888 E... 828 8.88- .8828 .8888- ..888- .8828 2288885
2828 2888 8288 2888 2888

828 8288 8.8.. - - .828 8288- 888- .2888 65888885
23828 2888 2888 2888 22888 2888

828 828 888.. - 888 .8888 .888- 8.88- .238 3.3382...-

888 2......

"8.85-3.5...

z 8 8888 aha. 8283.0. 852 8 8. 25.2 2... 28 8. .382...> 8.8.85

 

 

 

.8. 3.8 828 .3888 So .8288. .888 an: .m 8888.

   

115

 

 

 

 

 

888888 88888 88888 28888 88888 $888..

82 888 88.8 8288- .288 .8888 2.8888- - .8828 ES. 8:5
28888 288.8 28888 58.8 5888 288888

82 2828 88888 8888- .888 .8888 8888 - .8888 SEE 3:5
282.888 2888. 888.8 28888 28888 282888

8: :88 88... :88. 3:88 .828 2.888. - .8828 Ema 52.5
28288 888.8 888.8 28888

8: 3.88 828.8 - - .2328 2:888. - .888 8582.88.86
83:88 28888 28888 288.8 28888

2.8: 8828 28.8 - .888 .8888 .8888. - .8828 2883522

«our—m gee—aficu
888888 @2888 88882 882888 88888

8: 888 88.8 - 58:8 8828 - .5288 .8828 6.2886 823.38
c.8888 c.8888 @8888 28888 828.8 @3888

8: 8828 888 2.888- .888 .288 - 2.888 58888 2888.285 88
52.8 288.8 28888 28888 28888

82 888 888 .8888 - .8828 - 88.88- 2888 292988020
6888. 288.8 28888 588. 288.8 88888

8: 8288 288.2 .8828- ...828 .8888 - 888 .8288 25.88.0853
88888 28.-888 288888 288888 28882 28888

882 8828 25.8 288- .8828 .8888 - 8288 2.2288 238.285
@8888 288.8 E88 28.888 E88 28888

32 8888 888.8 888 2.8828 .8288 - 8.88.? .8228 23.-28 3.8
5888 2:882 28888 2:888 @8888 288888

8: :88 8:8 888 8:88- .8828 - 52888. .888 28.28.3882

2 .8 .588 ...r2< 2.2838 . .55. 8 8 s 5.2 8. 2.8 82 ”czar; Eunice

 

88.2.88 m 888%..

116

 

 

 

 

 

 

 

€88 $8.8 $88 888 A888

8: 88.8 88.8 5.88.8 - .828 - :88 88.8- 64885.35

soot”.— «583.52.:
$38 @888 888 E88 888 @888

38 88.8 88+ 2.888- - 88.8 28.8. 38.8.. 88.8- Eamzasoz
$28 88.8 58.8 6888 A888 §88

8: 8:8 888 - 83.3 :83 2.8888- 38.8. .888 $888358:
83.8 88.8 388 A888 8888 $88 88.8

8: 88.8 883 2.88.8 83.3 883. .828- 888 :888 Egfiqeoz
8828 A888 888 9:88 3888 E88

88 88.8 888 - 888 888 8.88. .8888 .3888- 585.2%
$8.8 $8.8 $8.8 $8.8 $8.88

88 88.8 88.8. - - .828 .238- 88.8- :88... Epemesn
5:8 588 88.8 58.8 88.8 888.8

88 888 888 2.88.8- - 82.8 88.? 2.28.8 38.8 64885056

.85 :5.

Sag-outflcnvénskataz .m—
om88 A888 888 $88 8888

8: 88.8 883 - 883 .888 .838- - :83 64836 3383
A888 $8.8 £88 588 $8.8 $88

82 2:8 883 2.888-5888 88.8 3.28. - .828 688.285 :8
23:8 :88 888.8 88.8 38.8

82 88.8 3.3 88.8- - .888 88.8- - :88 8888830
2828 A888 6888 38.8 $8.8 $88

8: 88.8 3: .828- 888.8 .888 .288- - .888 admozmémfi

2 mm 8528 .. . E”? n. . 85an _ 29 B 8 s 5.2 15388128538235

 

8.258 m 3:83.

117

 

 

 

 

 

882.8 888.8 38.8 888.8 838.8

82 8.88 8.8... .888- . 88:8 .5888- - 288. 2388.332
8888 A888 A888 A288 288.8

8: 888 3.88 - .888 .888 :888. - .828 8828888982
8888 58.8 888.8 58.8 288.8 888.8

8: 388 2.88 2.888 8888 28:8 .8828. - 2.288 228888802
288.8 58.8 388.8 $88.8 $88.8

8.3 888 888 - 8:88 .888 5888. - 888. 5.88288:
8888 58.8 888.8 228.8

8.: 888 R88 - - .888 .82.? - .828 9858.888
828 $88 @888 88.8 828.8

8: 8888 888 888- - .828 2.588- - .8888 23885356

«ooh—m .532950
58.8 58.8 8888 8:88 €8.88

82 828 888 8888- - .888 - 288- 888- Eamiaaoz
88.8 828.8 $88.8 $88.8 828.8

82: 8888 3.88 . .888 .888 - .888- 2.888 8828888852
$88.88 88.8 288.8 888.8 $88.8 888.8

8: 8888 83: 2.8888 .888 .888 - 8888 888 cam—8 828802
988.8 888.8 888.8 28.8 838.8

8.: 888 888 - 888 .8888 - 2.888 1.888. $28 82832
288.8 288.8 688.8 888.8

83: 888 888... - - 8:8 - 8888 888- 5.89888

2 "m .8588 2.832 9.82828 . SE. 8 S s 5.2 8.. 28 mo. 63858, 88:35

 

8.258 m 55.2%.

118

 

 

 

 

 

 

 

 

 

888.8 888.8 58.8 988.8 888.8

82 888 882 - .828- .888 - .828- 288- 3:8 8220328
2388 888.8 888.8 288.8

82 $88 888 - - .828 - .858- _28.8 88808820
988.8 :88 5.8.8 $8.8 5.8.8

82 .888 828 - :888- 58.88 - 5888. 8888 ASEMSE
288.8 888.8 888.8 588 88.8 288.8

:2 888 53 £88- .828 2.888 - .82.? 5888 23,-28 3:8
288.8 8:88 8888 888.8

82 888 882 - - 2388 - .2888- :88- 8888235

Baum Eon—355:5.-
8888 3.8.8 2888 $8.8 838.8

«28 888 23.8 - - .888 .888- ...828- :88 E8 8588;
888.8 888.8 8288 E88 2288 88.8

~28 888 888 - .888- ...888 .228- .838- :888 3:8 829032..
58.8 288.8 882.8 88.8 82.8

88. 8888 888 - - .828 .888- .888- 2.2.8828 295808020
58.8 888.8 828.8 58.8 888.8 888.8

88 858 882 - .888- 5888 .8888- :888- 2.288 328 was:
828.8 888.8 888.8 838.8 8888 88.8 288.8

88 888 888 2.58.? .838 .2888 :88. 8:88. 8228 SEQ-8:5
88.8 838.8 88.8 888.8 888.8

88 288 82-8 - - 8:88 .828- .8:8- .888 2588235

.85 :5.

”agto§-nhaE_< m: .0

z “8 2388 .852 8.82826 . .59 8 8 5 >82 30388128535888

 

.8888 m 888%.

119

 

 

.88... 33.28 8.288888 .35. $2 85 .8 .8. 28 a 5.8ch 2.888% r: Ea :.._
.8985 88.8 «o mafia 05 mo 5.2—£83 a 8.“ En cozoow .m 5885 new 2525.8 032.8.» Dogma—axe Ea Eon—Snow .28 NA 033- 8m

68202. b.82a8woow 8: Pa 85 atom-:8 no» a 52583 98 29:9. :8 555 3.59% 3552.8 :09? Ba 88» no 8: Eu U 585%? com A8

fies 28:8 :32 8833.853: .83 3

 

 

 

 

82 888 883 - 8888 888- - .888 E8 8.5885
888.8 888.8 2828.8 888.8 88.8

82 8888 8:8 .828- .888 888- - .828 SE8 8280328
8228 88.8 282.8 82.8

82 888 823. - .888 8888- - 2.888 58888320
8888 58.8 828.8 888.8 688.8

82 888 883 888- 8288 2.888- - 2.888 288282:
88.8 88.8 888.8 828.8 888.8 28.8

:2 888 8828 888- .888 .888 8888- - .828 $883.85
858.8 58.8 288.8 828.8

82 8288 888 - 8288 5288- - .828 858382.20

woo-an.— groan—:9
E88 8288 888.8 888.8

82 888 8888 - .888 - .828- $88 E8 8588:

2 N8 8528 88.82 9.82825 :88: 8 8 s 83. 20388128538885

 

.8888 m 8888.

120

 

 

 

.93.. 8.353.. 82.688. .26. .8. 85 2. a... 2.. 3 .§....&.. 8.8.5.38 r: us. a...
.885... a... .o 8...... 2.. 8 8.8.5.... a a. S. 8.68 .m .336 8m .2688... 2.8.2, 8.2.5....» 5. .... 2...... gm .8

.82... 185.. .28... 8.2.3.383... 883......

 

 

121

89......
«.3... R2... 888 88... .8888 .82...- - .88.. 5%:
.88.... 888.8 .88.... 58.8 58.8
88 :2... 883 .88.... 28.... .88.... - .82... 8-32
8.8.8 .88.... .88.... .888 .88....
8...: 88m... 28.. .88...- .88.... .82.... - .88.. <22...
«comm coco—aficu
58.8 .28.... .888 .388 2.8.8
8.... 82... 888 :88... .88.... - .88...- .88.. firm:
.288 .88.... .28.... .888 .888
88 82... 8.88 .88.... 88... - 82.... .88.. 0932
.88.... .88.... 8.8.8 .28.... .88....
88... 8...... 88.... 88.... .88.... - 88m..- 88.... 3.25
Sauna «EOE-0055::
.88.... .88.... .888 888.8 .888 82......
88.8 8.8... 3.88 :83... .28.... 82...- .88...- .88... $.15
.88.... .88.... 82...... .88.... .28.... .888
88.8 .88... 288 :88... 88.... 82...- .88...- .88.. 8-32
.888 .888 .888 .888 .88.... .28....
88.... 8...... 28.... 88...- .88.... .88... .88.... .88.. «5.5 .85 .8...
6102502
2 M. 8528 . Em... €8.38 8 8 s 5.2 a... 28 8. fiat... .588...

 

.8823... 298 .98.: 38.. A... 28 8.8 .835... 30 3.8.. u... 888......

 

 

Appendix G: Variable Descriptive Statistics by Sub-Sample.a

 

 

 

 

 

 

 

 

 

Variable Route Sub-sample Network Mean s.e. Min. Max
Full Effect
Fare in Overlapping DL-UA 213.94 128.71 1 8 1757
NW—CO 199.48 102.65 24 982
US-AA 210.44 110.59 55 1437
Non-Overlapping DL-UA 198.09 120.58 16 1710
NW-CO 213.08 126.73 19 2591
US-AA 217.95 1 19.06 15 1797
AM in (%) Overlapping DL-UA 1 1.86 14.59 0.2 77.6
NW-CO 13.74 13.20 0.8 67.3
US-AA 15.03 17.88 0.5 76.8
Non-Overlapping DL-UA 12.38 14.93 0.1 88.6
NW-CO 13.52 13.55 0.2 88.6
US-AA 14.33 15.84 0.2 96.7
MS in (%) Overlapping DL-UA 17.08 23.45 0.0 100
NW-CO 19.52 30.16 0.1 100
US-AA 16.88 28.73 0 98.7
Non-Overlapping DL-UA 20.03 28.26 0 100
NW-CO 19.10 27.47 0.1 100
US-AA 19.63 27.18 0.1 100
DISTr Overlapping DL-UA 1585.91 617.58 228 2583
NW-CO 1026.82 599.95 81 2565
US-AA 1094.14 405.68 541 2243
Non-Overlapping DL-UA 1098.02 654.54 30 2704
NW-CO 989.02 587.84 17 2565
US-AA 1019.23 661.80 67 2724
lagMS in (%) Overlapping DL-UA 16.87 23.69 0 100
NW-CO 12.89 24.69 0 99.9
US-AA 10.23 22.97 0 92.6
Non-Overlapping DL-UA 19.59 27.81 0.1 100
NW-CO 12.86 24.3 0 100
US-AA 12.34 22.99 0 100
Announcement Efiect
Fare in Overlapping DL-UA 219.54 138.96 17 1 7 57
NW-CO 203.08 105.00 24 982
US-AA 216.19 123.55 55 1437

 

122

 

Appendix G (cont'd).

 

 

 

 

 

 

 

 

 

Variable Route Sub-sample Network Mean s.e. Min. Max.
F are in Non-Overlapping DL-UA 198.65 1 18.1 1 17 1390
NW-CO 218.01 131.18 19 2591
US-AA 222.76 122.46 16 1797
AM in (%) Overlapping DL-UA 1 1.12 13.55 0.3 77.6
NW-CO 13.75 13.52 0.8 67.3
US-AA 14.19 17.62 1.7 76.8
Non-Overlapping DL-UA 1 1.78 14.48 0.2 85.5
NW-CO 13.55 13.54 0.4 83.1
US-AA 14.31 16.07 0.4 84.1
MS in (%) Overlapping DL-UA 17.19 23.88 0.1 100
NW-CO 18.93 29.77 0.1 98
US-AA 16.30 28.56 0.1 98.7
Non-Overlapping DL-UA 19.97 28.29 0.1 100
NW-CO 19.08 27.52 0.1 100
US-AA 19.69 27.35 0.1 100
DIST, Overlapping DL-UA 1596.83 623.39 228 25 83
NW-CO 1026.07 619.25 105 2565
US-AA 1091.82 377.95 541 1917
Non-Overlapping DL-UA 1101.10 651.94 67 2704
NW-CO 988.84 587.31 17 2565
US-AA 1019.50 661.72 67 2724
lagMS in (%) Overlapping DL-UA 16.89 23.65 0.1 100
NW-CO 13.63 25.16 0.1 97.7
US-AA 12.72 26.13 0 92.6
Non-Overlapping DL-UA 19.34 27.4 0 100
NW-CO 13.96 24.86 0 100
US-AA 14.25 24.94 0 100
Completion Effect
Fare in Overlapping DL-UA 212.15 128.01 1 5 1757
NW-CO 201.81 1 10.33 24 982
US-AA 209.68 121.72 57 1437
Non-Overlapping DL-UA 197.41 121 .05 1 8 1710
NW-CO 210.76 121.62 19 1710
US-AA 210.21 1 14.49 15 1797
AM in (%) Overlapping DL—UA 1 1.86 14.1 1 0.2 75.7
NW-CO 13.4 12.96 1.3 67.3
US-AA 15.33 18.44 0.5 76.8

 

123

 

 

Appendix G (cont'd).

 

 

 

 

 

Variable Route Sub-sample Network Mean s.e. Min. Max.
AM in (%) Non-Overlapping DL-UA 12.35 14.79 0.1 88.6
NW-CO 13.52 13.44 0.2 88.6

US-AA 14.26 15.52 0.2 96.7

MS ,,, (%) Overlapping DL-UA 17.04 23.27 0.1 100
NW-CO 20.6 30.97 0.1 100

US-AA 16.79 29 0.1 98.7

Non-Overlapping DL-UA 20.03 28.4 0.1 100

NW-CO 19.31 27.63 0.1 100

US-AA 19.59 26.99 0.1 100

DIST, Overlapping DL-UA 1571.7 612.45 264 2583
NW-CO 1015.75 588.79 81 2554

US-AA 1074.18 41 1 .41 541 2243

Non-Overlapping DL-UA 1094.3 1 656.24 30 2704

NW-CO 989.76 588.14 17 2565

US-AA 1019.68 662.5 67 2724

lagMS in (%) Overlapping DL-UA 17.06 23.58 0.1 100
NW-CO 1 1.38 23.09 0.1 99.9

US-AA 8.42 21.1 0 92.6

Non-Overlapping DL-UA 20.05 28.38 0.1 100

NW-CO 11.1 23.09 0 100

US-AA 9.94 20.7 0 100

m
(a) See Chapter 3, Section 3.7 for time period description and Table 4.1 for variable definitions.

124

 

 

 

 

 

 

@5558 8:58 858.8 @528 58.8 32.8
58 535 :85 - 285 .355 ...§:5 .825 8:5 8355 555306
80h”.— «non—3:35
258.8 225.8 8258 A838 .888 8588 $88
3.2 225 885 .35.? 88.? .385 5.8.? .335 .885 853. m.._85_s.n..,o.82
@848 8:58 8.88 885.8 52.8 653.8 835.8
E. $3.5 $9.5 88.? 235 .5355 $85 .5525 .385 833— 5553.5
.85 :5...
“—defioflcncUauuéBA—tez
588 2:58 @588 925.8 588.8 9.588
3...: 8.25 85mm .38.? - .835 .83.? .825 .335 .385 55538582
528 858.8 958.8 588 $88 6588
255 8:5 853 .28.? - .255 .335- :335 .3555 8:5 5553.5
80am secs—9.59
835.8 58.8 825.8 58.8 225.8 925.8
a}: 3:5 285 - N85? .885 .88.? .8525 .2555 8:3 555355.82
858.8 885.8 258.8 855.8 258.8 38.8
255 825 88.... - $85. .335 .82.? :85 .4525 .385 5359.35
“903nm Havana-=55:
858.8 225.8 225.8 858.8 825.8 858.8 8.588
53.2 825 $85 .23.? 585.? .5535 .235- .2:5 .385 833. 535.3582
$2.8 358.8 58.8 A288 3588 5.88 858.8
255 825 :52 .88.? 55.? .2255 .555- .4585 .335 .328 55552.5
.85 :5.
"3:52.30:
2 x .5528 8 8 .55. . E... 5826 s 5.5.. 2 2.8 we. ".533, .8288

 

.33555 .53. .8 v.1. 8.8 52% mega 30 5285 "m 5559..

125

 

. .5.

 

 

 

 

 

 

 

@228. 5.88 885.8 8858 A888 2288
58 :25 82.8 .8855- - .885 .885- .825 .825 .253. 8885
«noun..— neg—afieu
5.8.8 @258 23.8 88.8 .888 58.8
825 $25 253“ - 885 .885 885- .285 .825 8332 88.85-82
888 82.88 885.8 858.8 38.8
58 225 88.8 - 885 .885 - :825 .825 .253. 8582.5
«Baum anon-0256::
A288 8258 825.8 88.8 225.8 288.8 825.8
58.2 825 853 .82.? .825- .885 285 .385 .825 8532 882.3082
28.8 888.8 288.8 88.8 88.8 .888 58.8
2.. 225 82.8 :88? 88.? .385 .88.? .825 .825 .253. 8282.5
.88. :5.
Baotou-«£8.52 m3
@858 88.8 855.8 825.8 858.8 825.8
585 885 888 285.? - .885 8855- .885 .885 .2532 5588.35-82
858.8 $.88 8.88 222.8 288.8 882.8
58 225 28.8 88.? - .885 8.85 .825 385 853. 558.35
Sou...— gate—1:30
885.8 825.8 925.8 825.8 825.8 858.8
585 225 88.8 - 885 .285 8.8.? .285 88.? 833. 8.288582
2 m. 8588 B 8 €29 . .52 .5228 s 5% A... 2.8 m2 ”285, .8888

 

.9888 m 58:25....

126

 

N2.»
2

"hm—d $3.9m

N“

88280

53.2.?
Ma

63... 8586.5 2.2.88.2 .38. :2 9a :5 ex: .5 a .5053. 2.852% u... 2. ....
82.58.. 29.5.53 .2 N5 8:03 5 saga 8m £82522. 03.5. 935:... 3.. 3. 2.3 8m @

- .5955 5855 .5255 .385 .385 555%...5eoz
8 .29 . Em... .5528 s .55... A... 2.8 52 ”03.5,, .8883
.9385 m .8859...

 

127

 

 

 

 

 

 

.885. 8855.. 42.85. .885. .885.
8.8. 5585 885- 8.55. 2.85 .285 .58.... 8.8.. 8.8.85-8...
88.5. .885. .885. 6855. €8.85
2.. 885 .8~5 ...85..5- 885- .285 2.5 8.2... 8.8.85
.85 :5.
”Eggnog-394835.562
E85. 8.85. @585. .885.
~85. 885 885 .885- - .885 .285 .88.. 8........8>o..87.
.885. .885. 885.8 .885.
2.8 2.85 885 885- - .885 .285 8.8.. .8888...
.88. 5.8.38
.885. .885. .885. .285.
~85. :85 :85. - .825 .885 .885 .38.. 8.8.85.8...
.885. .885. .285. .885.
2.8 ~55 $85- - .8525 .885 558.5 8.8.. 8.8.85
vacuum «EOE—09:25:
.885. .885. .885. 2855.
85.2 885 85.5- .88... .825 .885 .885 8.8.. 8.88.8587.
@855. 8.85. .885. .2555. .885.
255 285 .855- 3.85 2.825 .885 £2.85 8.8.. 8.9.2.5
.82. :5.
8.82-3.2.
s m... s 5...... .5 ES 8. 8......9 .8888

 

«8.3.8.8 8.8.. 8. .5. 28 8...... 888...... 38.8.88... .8“... W588...

 

   

 

 

128

 

 

 

 

.885. 985. .8555. .885.
8.... 885 885- - .825 .885 .885 8.8.. 88885-82
.825. 88.5. .8555. .885.
8~ 885 85.....- - 885 .885 ...885 8.8.. 8.88.85
«ugh—H «floaouflficflfld.
.8855. .5855. .885. .8555. .5855.
58.2 885 885- 885 ..-.5 .5855 .885 8.8.. m_.a........85.8.,.
.825. .885. .825. .5855. .885.
2.. 885 885- .855- 885 .885 :88... 8.8.. 8.88.85
.85.. :5.
Eac...»E<-9?E_< m3
.885. .885. .8555. .885.
58... 885 :25- 8555 - .885 :83... 8.8.. 888.8582
.285. .885. .2255. .82...
8~ 58.5 825- 825- - .....85 885- 8.8.1.8888...
30am actoEEoU
.~.85. .885. .8555. .5855.
58... 885 825- - 8555- .285 .825 8.8.. 8.88.8582
.825. .885. .8855. .285.
8~ 885 8~5 - 885- .5855 85.5 8.8.. 8.88.05
«ovum “Gui—00:56::
z N... .388 8. 8. s 8,. s .3... .s 2... 8. 8.8.8.. .8885

 

.5280. . 882......

129

58.-. Biézv boéooame ._26_ $2 9a .xom <1 06 a 2.85:»: beefing H... Ea z...
82.5% 038853 .e an 883 a 3:36 8m .8358 22%: 8022.85 as. .5238 5. Z. 0:3 8m g
.55 Bang 252 £2.33égfi as, 3

 

 

$3.8 $8.8 8898 $8.8
N2.» 393 898- ES. - ES... 553 839m 3&25552
52.8 53.8 8898 sate
EN $86 $3.? 38.0- - .385 :88... 830m wsaqazzv
89b”.— =88...an
z 5588 B 8 5m: s 5,2 22$ we. a§§>§2§a

mm

 

.auaoov _ xmvaommao‘

130

 

u..

 

BIBLIOGRAPHY

131

BIBLIOGRAPHY

Abramowitz, Amy D. and Stephen M. Brown (1993). Market Share and Price
Determination in the Contemporary Airline Industry. Review of Industrial Organization:
Volume 8, 419-433.

ACI Traffic Data: World Airports Ranking by Total Passengers, 1999.
http://www.airports.org/traflic/td_passengers.

Alam, Ila M. Semenick, Leola B. Ross and Robin C. Sickles. (2001). Time Series
Analysis of Strategic Pricing Behavior in the US Airline Industry. Journal of Productivity
Analysis: Volume 16, 49-62.

Bamberger, E.Gustavo, Dennis W. Carlton and Lynette R. Neumann. (March, 2001). An
Empirical Investigation of the Competitive Effects of Domestic Airline Alliances.
National Bureau of Economic Research: Working Paper Number 8197.

Banerjee, Abhijit and Lawrence Summers. (September, 1987). On Frequent Flyer
Programs and Other Loyalty-Inducing Economic Arrangements. Harvard Institute of
Economic Research: Discussion Paper Number 1337.

Bender, Alan R. and Frederick J. Stephenson. (1998). Contemporary Issues Affecting the
Demand for Business Air Travel in the United States. Journal of Air Transport
Management: Volume 4, 99-109.

Berry, Steven T. (1990). Airport Presence as Product Differentiation. American
Economic Review: Volume 80, Number 2, 394-399.

Berry, Steven, Michael Camall and Pablo T. Spiller. (May 1996). Airline Hubs: Costs,
Markups and the Implications of Customer Heterogeneity. National Bureau of Economic
Research: Working Paper 5561.

Beutel, Phillip A. and Mark E. McBride. (January 1992). Market Power and the
Northwest-Republic Airline Merger: A Residual Demand Approach. Southern Economic
Journal: Volume 58, Number 3, 709-720.

Boguslaski, Charles, Harumi Ito and Darin Lee. Entry Patterns in the Southwest Airlines
Route System. The International Industrial Organization Conference, Boston,
Massachusetts. April 5-6 2003.

Borenstein, Severin. (November 1991). The Domith Firm Advantage in Multiproduct
Industries: Evidence from the US. Airlines. The Quarterly Journal of Economics:1237-
1265.

--(May 1990). Airline Mergers, Airport Dominance and Market Power. American
Economic Association Papers and Proceedings: Volume 80, Number 2, 400-404.

132

«(Auturnn, 1989). Hubs and High Fares: Dominance and Market Power in the
US. Airlines Industry. RAND Journal of Economics: Volume 20, Number 3, 344-
365.

Borenstein, Severin, and Nancy Rose. (1994). Competition and Price Dispersion in the
US. Airlines Industry. Journal of Political Economy: Volume 102, Number 4, 653-678.

Boyer, Kenneth D. (1992). Mergers That Harm Competitors. Review of Industrial
Organization: Volume 7, 191-202.

Bradley Michael, Anand Desai and Kim B. Han. (April 1983). The Rationale Behind
Inter-firm Tender Offers: Information or Synergy? Journal of Financial Economics:
Volume 11, Number 1-4, 183- 206.

Brander, James A. and Anming Zhang. (Winter, 1990) Market Conduct in the Airline
Industry: An Empirical Investigation. RAND Journal of Economics: Volume 21, Number
4, 567-583.

Bresnahan, T.F. Empirical Studies of Industries with Market Power. New York, North
Holland, 1989. Chapter 5 of The Handbook of Industrial Organization. Ed. R.
Schmalensee and R. D. Willig. 1011-1058.

Brueckner, Jan K., International Airfares in the Age of Alliances. (February 2003). The
Review of Economics and Statistics: Volume 85, Number 1, 105-118.

- (December 2002). Airport Congestion when Carriers have Market Power, The
American Economic Review: Volume 92, Number 5, 1357-1375.

«(2001). The Economics of International Code-sharing: An Analysis of Airline
Alliances. International Journal of Industrial Organization: Volume 19, 1476-
1498.

Brueckner, Jan K., Nichola J. Dyer and Pablo T. Spiller. (1992). Fare Determination in
Airline Hub-And-Spoke Networks. Rand Journal of Economics: XXIII, 309-333.

Burton, John and Pat Hanlon. (1994). Airline Alliances: Cooperating to Compete?
Journal of Air Transport Management: Volume 1 Number 4, 209-227.

Cairns, Robert D. and John W. Galbraith. (November, 1990). Artificial Compatibility,
Barriers to Entry and Frequent Flyer Programs. Canadian Journal of Economics: Volume
XXIII, Number 4, 807-816.

Call, Gregory D. and Theodore E. Keeler. Airline Deregulation, Fares and Market

Behavior: Some Evidence. Ed. Andrew F. Daugherty. Cambridge: Cambridge University
Press, 1985. In Analytical Studies in Transport Economics.

133

Caves, Douglas W., Laurits R. Christensen and Michael W. Tretheway. (Winter, 1984).
Economics of Density Versus Economies of Scale: Why Trunk and Local Service Airline
Costs Differ. RAND Journal of Economics: Volume 15, Number 4, 471-489.

Clougherty, Joseph A. (January 2002). A Political Economic Approach to the Domestic
Airline Merger Phenomenon. Journal of Transport Economics and Policy: Volume 36,
Part 1, 27-48.

Davies, S. W. and Paul A. Geroski. (1997). Changes in Concentration, Turbulence and
the Dynamics of Market Shares. The Review of Economics and Statistics: 383-391.

Dempsey, Paul Stephen and Andrew R. Goetz. Airline Deregulation and Laissez-Faire
Mythology. Quorum Books, 1992.

Douglas G. W. and J. C. Miller. (1974). Quality Competition, Industrial Equilibrium and
Efficiency in the Price-Constrained Airline Market. American Economic Review:
Volume 64, Number 4, 657-669.

Dresner, Martin, Robert Windle and Yuliang Yao. (September 2002). Airport Barriers to
Entry in the US. Journal of Transport Economics and Policy: Volume 36, Part 2, 389-
405.

Dresner, Martin and Michael Tretheway. (1992). Modeling and Testing the Effect of
Market Structure on Price. Journal of Transport Economics and Policy.

Evans, William and Ioannis Kessides. (May 1994). Living by the "Golden Rule":
Multimarket Contact in the US. Airline Industry. The Quarterly Journal of Economics:
Volume 109, Number 2, 341-366.

—(1993). Localized Market Power in the US. Airline Industry. Review of
Economics and Statistics: 66-75.

Farrell, Joseph and Carl Shapiro. (March 1990). Horizontal Mergers: An Equilibrium
Analysis. American Economic Review: Volume 80, Number 1, 107-126.

Graham, David R., Daniel P. Kaplan and David S. Sibley. (Spring, 1983). Efficiency and
Competition in the Airline Industry. Bell Journal of Economics: Volmne 14, 118-138.

Hu, Michael Y., Rex S. Toh and Stephen Strand. (July-August, 1988). Frequent Flyer
Programs: Problems and Pitfalls, Business Horizons: 52-57.

Hurdle, Gloria J., Richard L. Johnson, Andrew S. Joskow, Gregory J. Werden and

Michael A. Williams. (December 1989). Concentration, Potential Entry and Performance
in the Airline Industry. The Journal of Industrial Economics: Volume 38, Number 2.

134

 

J oskow, Andrew 8., Gregory J. Werden and Richard L. Johnson. (1994). Entry, Exit and
Performance in Airline Markets. International Journal of Industrial Organization:
Volume 12, 457-471.

Kim, E. Han and Vijay Singal. (June 1993). Mergers and Market Power: Evidence from
the Airline Industry. The American Economic Review: 549-569.

Kroszner, Randall 8., Wallace P. Mullin, Judson L. Jaffe and Cindy R. Alexander.
(Spring 2002). President's Council of Economic Advisors, Economic Organization and
Competition Policy. Yale Journal on Regulation: Volume 19, Number 2, 541-597.

Lee, Darin and Maria Jose Luengo Prado. The Impact of Passenger Mix on Reported
"Hub Premiums" in the US. Airline Industry. The International Industrial Organization
Conference, Boston, Massachusetts. April 2003.

Levine, Michael E. (1987). Airline Competition in Deregulated Markets: Theory, Firm
Strategy, and Public Policy. Yale Journal on Regulation: Volume 4.

Liu, Qihong. The Effects of Market Share on Price Dispersion: An Analysis of the US.

Airline Industry. The International Industrial Organization Conference, Boston,
Massachusetts. April 2003.

McAffe, Preston R., Joseph J. Simons and Michael A. Williams. (1992). Horizontal

Mergers in Spatially Differentiated Noncooperative Markets. Journal of Industrial
Economics: Volume XL, Number 4, 349-358.

Meyer, John R. and John S. Strong. (1992). From Closed Set to Open Set Deregulation:

An Assessment of the US Airline Industry. Logistics and Transportation Review: Volume
28, Number 1.

Morrison, Steven A. and Clifford Winston. The Remaining Role of Government Policy in
the Deregulated Airline Industry. Washington DC: AEI-Brookings Joint Center for

Regulatory Studies. 2000. In Deregulation of Network Industries: What's Next? Ed. S.
Peltzrnan and C. Whinston. 1-40.

--(1996). Causes and Consequences of Airline Fare Wars. Brookings Papers:
Microeconomics, 85-123.

--(1990). The Dynamics of Airline Pricing and Competition. The Brookings
Institution AEA Papers and Proceedings: Volume 80, Number 2, 389-393.

Nako, Steven M. (1992). Frequent Flyer Programs and Business Travelers: An Empirical
Investigation. Logistics and Transportation Review: Volume 28, Number 4, 395-414.

135

OfficialAirlineGuide. (1999). Worldwide Edition, Reed Travel Group.

Oum, Tae Hoon and Jong-Hun Park. (1997). Airline Alliances: Current Status, Policy
Issues and Future Directions. Journal of Air Transport Management: Volume 3, Number
3, 133-144.

Park, Jong-Hun and Anmig Zhang. (1998). Airline Alliances and Partner Firms' Output.
Logistics and Transportation Review: Volume 34, Number 4, 245-255.

Rhoades, Dawna L. and Heather Lush. (1997). A Typology of Strategic Alliances in the
Airline Industry: Propositions for Stability and Duration. Journal of Air Transport
Management: Volume 3, Number 3.

Singal, Vijay. (1996). Airline Mergers and Multimarket Contact. Managerial and
Decision Economics: Volume 17, 559-574.

Stephenson, Frederick and Richard Fox. (Fall,1987). Corporate Attitudes Toward
F requent-F lyer Programs. Transportation Journal: 10-22.

Tirole, Jean. The Theory of Industrial Organization. Cambridge, Massachusetts: The MIT
Press, 1994.

Toh, Rex S. and Michael Y. Hu. (1988). Frequent Flyer Programs: Passenger Attributes
and Attitudes. Transportation Journal: Volume 28, Number 2, 9-22.

US. Department of Transportation. American Travel Survey. Bureau of Transportation
Statistics: 1988.

US. General Accounting Office, Aviation Competition: Challenges in Enhancing
Competition in Dominated Markets. GAO/01-518T. Washington DC: Government
Printing Office, March 2001.

«Aviation Competition: Effects on Consumers from Domestic Alliances Vary.
GAO/RCED-99-37. Washington DC: Government Printing Office, January
1999.

--Proposed Domestic Airline Alliances Raise Serious Issues. GAO/T-RCED-98-
215. Washington DC: Government Printing Office, June 1998.

«Airline Deregulation: Barriers to Entry Continue to Limit Competition in
Several Key Domestic Markets. GAO/RCED-97—4. Washington DC:
Government Printing Office, October 1996.

--Airline Competition: Higher Fares and Less Competition Continue at

Concentrated Airports. GAO/RCED-93-1 71. Washington DC: Government
Printing Office, July 1993.

136

«Fares and Service at Major Airports. GAO/RCED-90-102. Washington DC.
Government Printing Office, 1990.

«Airline Competition: Industry Operating and Marketing Practices Limit Market
Entry. GAO/RCED-90-147. Washington DC. Government Printing Office,
August 1990.

Wooldridge, Jeffery M. Econometric Analysis of Cross Section and Panel Data.
Cambridge, Massachusetts: The MIT Press, 2002.

137

  
  

 

\lljlfilljfijllflljfilfil1111111111

504 98