University of Southern California Dissertations and Theses

Essays on competition and antitrust issues in the airline industry

(USC Thesis Other)

Essays on competition and antitrust issues in the airline industry

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Essays on Competition and Antitrust Issues
in the Airline Industry
by
Hae Yeun Park
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ECONOMICS)
August 2022
Copyright 2022 Hae Yeun Park
Acknowledgements
I am grateful to a long list of people who have supported and encouraged me to complete my
journey. I would first like to express my sincere appreciation to my advisor, Geert Ridder. He
has been consistently supportive and has guided me in a better direction. I am also grateful to my
dissertation committee members, Guofu Tan and Erik Hovenkamp. The feedback and encourage-
ment they provided motivated me and helped me improve my research. The invaluable support and
patience from the committee chair and members enabled me to complete my dissertation and pre-
pared me to be a researcher. I also thank Fanny Camara, Yu-Wei Hsieh, and Leon Zhu for serving
on my qualifying committee and giving me helpful feedback. I also would like to thank Gior-
gio Coricelli, Cheng Hsiao, Hyungsik Roger Moon, Jeff Nugent, John Strauss, and other faculty
members for their generous support.
I would like to express my gratitude to Jinkook Lee for giving me the opportunity to be part of
the team. Working as a research assistant at the Center for Economic and Social Research (CESR)
was one of my greatest experiences during my PhD program, and it inspired me to redefine my
attitude as a researcher.
I appreciate the help I received from the Economics Department administrative staff, Young
Miller, Morgan Ponder, Alexander Karnazes, and Annie Le. I am also grateful to the USC Libraries
for supporting me in gaining access to necessary data through the Data Research Grant Program.
The first chapter of this dissertation used these data directly, and the Grant Program motivated me
to expand my research topics.
I thank my colleagues and friends at USC as well. Among many others, special thanks to
Jeehyun Ko, who helped me enormously whenever I needed assistance. I am also grateful to
ii
Seohee Ahn who helped me to think constructively, and Rihyun Park, a co-author of one of my
projects. Additionally, I am immensely thankful for the love and support from my old friends,
Jaewon Shin and Hyojin Park, who have been by my side. My PhD life would have been much
more arduous if I had not had them in my life.
I want to express my appreciation to my uncle and aunt, Bruce and Emily Howard. Their
support and encouragement helped me to overcome hard times. I will also never forget the love
I received from my grandmother, Myung Soo Park, who passed away about a month before my
defense.
Finally, I am genuinely grateful to my family for their unconditional and endless love and
support. Consistent love and support from my parents and brother kept me true to myself and
allowed me to be a better person. I am deeply grateful to be a daughter to my parents, Young Im
Lee and Hyo Dal Park, and a sister to my brother, Seong Ho Park. I will always remember their
greatest love.
iii
Table of Contents
Acknowledgements ii
List of Tables vii
List of Figures ix
Abstract xi
Chapter 1: Merger Effects of Low-Cost Carriers’ Merger 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Southwest Airlines’ Acquisition of AirTran Airways . . . . . . . . . . . . . . . . 7
1.4 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4.1 Sources of Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4.2 Sample Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4.2.1 Price Effects Sample . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4.2.2 Quality Effects Sample . . . . . . . . . . . . . . . . . . . . . . 11
1.4.2.3 Overlap Market Definitions . . . . . . . . . . . . . . . . . . . . 12
1.5 Empirical Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.5.1 Two-Way Fixed Effects DID regressions . . . . . . . . . . . . . . . . . . . 15
1.6 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.6.1 Market Level Descriptive Analysis . . . . . . . . . . . . . . . . . . . . . . 16
1.6.2 Price Effects of the Merger . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.6.3 Quality Effects of the Merger . . . . . . . . . . . . . . . . . . . . . . . . 20
1.6.3.1 Capacity Effects of the Merger . . . . . . . . . . . . . . . . . . 21
1.6.3.2 Price Effects for Nonstop Products . . . . . . . . . . . . . . . . 22
1.6.4 The Parallel Trends Assumption . . . . . . . . . . . . . . . . . . . . . . . 23
1.7 Conclusion and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Chapter 2: Effects of Codeshare Exit:
Evidence from the Merger of Alaska and Virgin America 47
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.3 Codeshare Alliance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.3.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
iv
2.3.2 Alaska (AS) and American (AA) Codeshare Agreement . . . . . . . . . . 53
2.4 Alaska Airlines’ acquisition of Virgin America . . . . . . . . . . . . . . . . . . . 55
2.4.1 Merger Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.4.2 Remedies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.5 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.5.1 Data Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.5.2 Sample Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.5.2.1 Codeshare Exit Effects Sample . . . . . . . . . . . . . . . . . . 60
2.5.2.2 Demand Estimation Sample . . . . . . . . . . . . . . . . . . . . 61
2.6 Descriptive Merger Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.6.1 Market Level Fare Changes . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.6.2 Fare Changes in Overlap and Either markets . . . . . . . . . . . . . . . . 64
2.7 Codeshare Exit Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.7.1 Empirical Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.7.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.7.3 Parallel Trends Assumption . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.8 Demand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.8.1 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
2.8.2 Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2.8.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
2.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Chapter 3: Determinants of Low-Cost Carriers’ Operational Performance 89
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.3 Data and Summary Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.3.1 Data Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.3.2 Sample Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.3.3 Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.3.3.1 On-time Performance Variables . . . . . . . . . . . . . . . . . . 95
3.3.3.2 Explanatory Variables . . . . . . . . . . . . . . . . . . . . . . . 96
3.4 Empirical Specification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
3.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
3.5.1 Departure On-time Performance . . . . . . . . . . . . . . . . . . . . . . . 101
3.5.1.1 Departure Delay Duration . . . . . . . . . . . . . . . . . . . . . 101
3.5.1.2 Probability of Being Delayed by 15 Minutes or More . . . . . . 103
3.5.2 Arrival On-time Performance . . . . . . . . . . . . . . . . . . . . . . . . 104
3.5.2.1 Arrival Delay Duration . . . . . . . . . . . . . . . . . . . . . . 104
3.5.2.2 Probability of Being Late by 15 Minutes or More . . . . . . . . 105
3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
References 124
v
Appendix A
Appendix to Chapter 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
A.1 Codeshare Remedy Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
A.2 AS-AA Codeshare Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
vi
List of Tables
1.1 LCCs Market Share: RPM (%) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.2 Summary Statistics: Price Effects Sample (Balanced Panel) . . . . . . . . . . . . 30
1.3 Summary Statistics: Price Effects Sample (Unbalanced Panel) . . . . . . . . . . . 31
1.4 Summary Statistics: Quality Effects Sample (Balanced Panel) . . . . . . . . . . . 32
1.5 Summary Statistics: Quality Effects Sample (Unbalanced Panel) . . . . . . . . . . 33
1.6 Pre & Post Differences: Market Level Fares . . . . . . . . . . . . . . . . . . . . . 34
1.7 Pre & Post Differences: Market Level Flight Frequency . . . . . . . . . . . . . . 35
1.8 Pre & Post Differences: Market-Carrier Level Flight Frequency . . . . . . . . . . 36
1.9 The Price Effects Results (Balanced Panel) . . . . . . . . . . . . . . . . . . . . . 37
1.10 The Price Effects Results (Unbalanced Panel) . . . . . . . . . . . . . . . . . . . . 38
1.11 The Quality Effects Results (Balanced Panel) . . . . . . . . . . . . . . . . . . . . 39
1.12 The Quality Effects Results (Unbalanced Panel) . . . . . . . . . . . . . . . . . . 40
1.13 The Capacity Effects Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
1.14 The Price Effects for Nonstop Flights . . . . . . . . . . . . . . . . . . . . . . . . 42
1.15 Parallel Trends Assumption Tests: Price Effects . . . . . . . . . . . . . . . . . . . 43
1.16 Parallel Trends Assumption Tests: Quality Effects . . . . . . . . . . . . . . . . . 44
2.1 Remedy Relevant Markets with Codeshare Products . . . . . . . . . . . . . . . . . 79
2.2 Regional Airlines affiliated to Major Airlines . . . . . . . . . . . . . . . . . . . . 80
2.3 Codeshare Exit Analysis Sample Summary Statistics . . . . . . . . . . . . . . . . 80
2.4 Demand Estimation Sample Summary Statistics . . . . . . . . . . . . . . . . . . 81
2.5 Pre & Post Differences: Market Level Fares . . . . . . . . . . . . . . . . . . . . . 81
vii
2.6 Merger Effects Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
2.7 The Codeshare Exit Effects Results (Definition A) . . . . . . . . . . . . . . . . . 83
2.8 The Codeshare Exit Effects Results (Definition B) . . . . . . . . . . . . . . . . . 84
2.9 Parallel Trends Assumption Tests . . . . . . . . . . . . . . . . . . . . . . . . . . 86
2.10 Demand Estimation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.1 Domestic Non-segment Market Share, 2000-2015 . . . . . . . . . . . . . . . . . . 109
3.2 On-Time Performance Data Reporting LCCs . . . . . . . . . . . . . . . . . . . . 110
3.3 List of Airports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.4 Statistics of dep delay 15 and arr delay 15: Day of the Week . . . . . . . . . . . . 112
3.5 Statistics of dep delay 15 and arr delay 15: Scheduled Time . . . . . . . . . . . . 113
3.6 Distance Group Summary Statistic . . . . . . . . . . . . . . . . . . . . . . . . . . 114
3.7 Variable Definitions and Summary Statistics . . . . . . . . . . . . . . . . . . . . . 115
3.8 Determinants of Departure Delays: Main Results . . . . . . . . . . . . . . . . . . 116
3.9 Determinants of Departure Delays: Alternative Specification . . . . . . . . . . . . 117
3.10 Origin & Destination Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . 118
3.11 Determinants of Departure Delay Probability . . . . . . . . . . . . . . . . . . . . 119
3.12 Determinants of Arrival Delays: Main Results . . . . . . . . . . . . . . . . . . . . 120
3.13 Determinants of Arrival Delays: Alternative Specification . . . . . . . . . . . . . . 121
3.14 Determinants of Arrival Delay Probability . . . . . . . . . . . . . . . . . . . . . . 122
A1 Virgin/American Domestic U.S. Overlap Routes . . . . . . . . . . . . . . . . . . 131
A2 Alaska/American Domestic U.S. Overlap Routes . . . . . . . . . . . . . . . . . . 132
viii
List of Figures
1.1 U.S. Domestic Market Share . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.2 Average Fare on Overlap Routes . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.3 Average Flight Frequency on Overlap Routes . . . . . . . . . . . . . . . . . . . . 29
1.4 Price Effects: Overlap Definition A . . . . . . . . . . . . . . . . . . . . . . . . . 45
1.5 Price Effects: Overlap Definition B . . . . . . . . . . . . . . . . . . . . . . . . . 45
1.6 Quality Effects: Overlap Definition A . . . . . . . . . . . . . . . . . . . . . . . . 46
1.7 Quality Effects: Overlap Definition B . . . . . . . . . . . . . . . . . . . . . . . . 46
2.1 US Airlines’ Market Share in 2016 . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.2 Three Types of Airlines Products . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.3 AS-AA Codeshare Markets (Traditional) . . . . . . . . . . . . . . . . . . . . . . . 77
2.4 AS-AA Codeshare Markets (Virtual) . . . . . . . . . . . . . . . . . . . . . . . . . 77
2.5 AS-AA Codeshare Total Passnegers (Traditional) . . . . . . . . . . . . . . . . . . 78
2.6 AS-AA Codeshare Total Passnegers (Virtual) . . . . . . . . . . . . . . . . . . . . 78
2.7 VX Presence on AS-AA Codeshare Routes . . . . . . . . . . . . . . . . . . . . . 85
2.8 The Nesting Structure 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
2.9 The Nesting Structure 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.1 U.S. Domestic Market Shares . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.2 Conditional Marginal Effects of org hub . . . . . . . . . . . . . . . . . . . . . . . 123
3.3 Conditional Marginal Effects of des hub . . . . . . . . . . . . . . . . . . . . . . . 123
A1 No. of Document Remedy 1 Markets and Those with Actual Codeshares . . . . . . 133
A2 No. of Document Remedy 2 Markets and Those with Actual Codeshares . . . . . . 133
ix
A3 No. of Document Remedy 3 Markets With AS and Those with Actual Codeshares . 134
A4 No. of Document Remedy 3 Markets With AA and Those with Actual Codeshares 134
A5 No. of Document Remedy 4 Markets and Those with Actual Codeshares . . . . . . 135
A6 No. of All Document Remedy Markets and Those with Actual Codeshares . . . . . 135
A7 AS-AA Codshare Market Types . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
A8 AS/AA Codeshare Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
A9 AA/AS Codeshare Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
x
Abstract
This dissertation comprises three essays on the competition and antitrust issues in the airline
industry, focusing on the competitive effects of airline consolidation and the behavior of low-cost
carriers (LCCs).
The first chapter investigates the effects of the merger between two low-cost carriers, South-
west Airlines and AirTran Airways, and focuses on overlapping routes. To control factors that
could affect the outcomes, two-way fixed effects difference-in-differences regressions are used on
outcomes at the market-carrier-level. The results suggest that changes in price are about 5-6%
higher for merging airlines than non-merging airlines, and there is no evidence of a relative in-
crease in the quality of the merging airlines on the overlapping routes. The main takeaway from
this paper is that, unlike the small price effects of mergers between legacy carriers, this merger be-
tween low-cost carriers appears to have significant anti-competitive effects on overlapping routes.
The second chapter
1
analyzes the competitive effects of the merger between Alaska Airlines
and Virgin America, and the impact of the Alaska-American codeshare discontinuation that re-
sulted mainly from remedies by the Department of Justice (DOJ). The results from the descriptive
merger analysis show that the merger was neither anti-competitive nor pro-competitive. Regard-
ing codeshare exit effects, the findings from a two-way fixed difference-in-differences analysis
and demand estimation indicate that the cessation of the codeshare agreement between Alaska
and American worked to reduce prices on the relevant markets; meanwhile, the demand in these
markets decreased.
1
This is joint work, coauthored with Rihyun Park.
xi
In the last chapter, I examine the determinants of low-cost airlines’ service quality, focusing on
on-time performance (OTP), using recently available low-cost carriers’ (LCCs) OTP data as well
as other data for control variables. Although LCCs initiated their operations focusing on short-
and medium-haul routes under a point-to-point system, the rapid expansion of these carriers has
allowed them to operate under a hub-and-spoke system and to serve more connecting passengers
through focus cities. The focus of this paper is whether departing from or arriving at a core airport
affects an LCC’s OTP, as is the case for full-service carriers. The findings show that departing
from or arriving at a hub or a focus city airport tends to reduce departure and arrival delays of
LCCs after controlling for several factors such as weather conditions, ages of aircraft, within-week
trends, and scheduled times. This relationship can be understood as an internalization of delay
externality because a carrier can suffer more from a delay on routes from or to a hub or a focus city
airport.
xii
Chapter 1
Merger Effects of Low-Cost Carriers’ Merger
1.1 Introduction
Since airline deregulation in 1978, the United States airline industry has experienced various
changes in market structure as well as in other features.
1
Many new airlines began to operate,
whereas some airline carriers ceased to operate. Some airlines were merged/integrated into other
airlines, and the market structure changed over time. In addition to entry and exit, flight fares
dropped due to competition from deregulation and service quality was also affected.
Among the various changes in the industry, one of the most notable ones is the growth of
low-cost carriers (LCCs).
2
The increase in the LCCs’ market share can easily be seen in Figure
1.1 and in Table 1.1. Figure 1.1 shows changes in the U.S. domestic air travel market shares of
legacy airlines, LCCs, and the two merged airlines, Southwest Airlines (WN) and AirTran Airways
(FL).
3 4
For shares of both revenue passenger miles (RPMs) and available seat miles (ASMs), the
market share owned by LCCs has increased consistently, and it has almost doubled over the last
1
Before the Airline Deregulation Act, there were mainly two types of regulations. One was for entry restrictions,
and the other was for fare controls. For a comprehensive description of the evolution of the U.S. domestic airline
market before and after the deregulation, see Borenstein and Rose (2014).
2
Passenger airlines can be largely categorized into two groups according to cost structure: legacy carriers (full-
service carriers) and LCCs. Legacy carriers are traditional airlines that were established before the deregulation,
whereas LCCs are relatively recent airlines offering lower fares and fewer services.
3
Market shares are calculated by using revenue passenger miles (RPMs) and available seat miles (ASMs), respec-
tively. RPMs is a measure of air passenger traffic, and ASMs is a measure of an airline’s capacity. The dashed grey
line indicates when the merger between Southwest and AirTran was announced, and the solid grey line indicates when
the merger was closed.
4
Market share of LCCs includes other small LCCs, as well as the merging airlines.
1
fifteen years. Among the LCCs, Southwest is the largest in the world. The consistent growth of
LCCs seems to come from both the sustained success of the largest one (Southwest) and the rapid
expansion of relatively smaller carriers such as Spirit Airlines (NK) and JetBlue (B6), as shown in
Table 1.1.
Another noteworthy trend in the airline industry is the consolidation of large airlines over the
last fifteen years, the so-called ‘mega mergers.’ Beginning with the merger between Delta Air
Lines (DL) and Northwest Airlines (NW) in 2008, mergers between legacy carriers have reduced
six large carriers to three airlines. There have also been mergers that involved LCCs. One was
the merger between Southwest Airlines (WN) and AirTran Airways (FL), and the other was the
merger between Alaska Airlines and Virgin America (VX). Due to the consolidation, the market
structure has become more concentrated.
While there have been many studies of legacy carriers on various topics, including the effects of
mergers, research that focuses on LCCs is much more limited. To fill this gap, this study examines
the impact of Southwest’s acquisition of AirTran Airways on the U.S. domestic air travel market.
5
A separate study of the merger between two significant LCCs is worthwhile because such a merger
could have different features than mergers of legacy carriers, as mentioned in Carlton et al. (2019).
A better understanding of LCCs would lead to a better understanding of the U.S. airline industry,
given their significant role in the industry and their difference from legacy airlines.
Theoretically, a horizontal merger can have two opposite effects on outcome variables like
price and quality. On one hand, anti-competitive effects result from a reduction in competition, as a
merger would result in a smaller number of airlines. In this channel, a merger increases the market
power of the merging firms, so that they can offer products with higher prices or lower quality.
On the other hand, a merger would result in efficiency gains. For example, in the airline industry
context, merging carriers may take advantage of an enlarged network (economies of scope) or
economies of traffic density (Brueckner and Spiller, 1994). If any efficiency gains are passed
on to consumers, a merger would result in lower prices and higher quality. Because theoretical
5
In most studies on a merger, including this one, the terms ‘merger’ and ‘acquisition’ are not distinguished, as the
difference between them is likely to be irrelevant for the purpose of analysis.
2
predictions depend on these two opposite forces, the net effects of a merger would depend on which
force is stronger. Empirical merger effects are quite case-specific, and they depend on industrial
environment and context, so a case study is required to examine the impact of a particular merger.
To examine the merger effects, I begin with a market-level descriptive analysis, where I com-
pare changes in market-level outcomes for different types of markets. The market-level descriptive
analysis provides a broad overview of how the airline industry changed before and after the merger.
Next, both price effects and quality effects are estimated with a focus on overlapping routes, which
are the most relevant type of market in terms of potential anti-competitive effects. The findings
show that overlap markets have, indeed, experienced the most price increases, and the merging
airlines increased their fares more than the other non-merging airlines on the overlapping routes
following the merger. When it comes to quality effects, there is no evidence of quality improvement
for the merging airlines. The results suggests that the merger is associated with anti-competitive
price and quality effects, especially for the overlap markets.
This study contributes to two strands of literature on merger retrospectives and the U.S. airline
industry. First, this study adds empirical evidence (case study results) to the growing literature
on merger retrospective studies. Recently, many studies have tried to evaluate merger approval
decisions. Whereas prospective merger analysis, such as merger simulations with demand esti-
mation, can provide useful information for the ex ante approval decision, ex post merger analysis
(a retrospective study of a realized merger) can help to evaluate the effectiveness of merger poli-
cies. As stated in the note provided by the Delegation of the United States
6
, implementing a case
study on a merger requires careful consideration from various perspectives, even though the re-
search questions seem straightforward. Even if generalization from each case study is not trivial,
additional case studies provide information for future research and help to improve the effective-
ness of merger policy as well as the methods used for prospective merger analysis.
7
In partic-
ular, the findings of this study can be applied to mergers in somewhat segmented—or vertically
6
This note can be found athttps://www.justice.gov/sites/default/files/atr/legacy/2011/08/01/
273463.pdf (last accessed in March 2022).
7
For example, Peters (2006) and Weinberg and Hosken (2013) conduct both a merger simulation and retrospective
analysis, and then compare the results.
3
differentiated—markets. The main distinction of the merger that is analyzed in this study is that it
is a merger between airlines that generally target price-sensitive passengers. The merger effects in
segmented markets may be different from the merger effects in less-segmented markets.
8 9
In this
sense, this study can provide useful findings regarding a significant merger in a segmented market.
This study also contributes to the airline industry literature, especially regarding LCCs. As
noted above, research on the U.S. airline industry has been focused on legacy airlines. This ap-
pears to be reasonable, as legacy airlines have longer histories and larger market shares than LCCs.
However, given that LCCs have been growing consistently—not only the dominant one (South-
west) but also the smaller ones such as JetBlue and Spirit have, understanding the impact of LCCs
has become more crucial to understanding the air travel market.
10
Specifically, because very lim-
ited analysis of mergers between LCCs has been done, this study provides useful insights regarding
this issue.
The remainder of the paper proceeds as follows. The next section is a literature review of
merger retrospective studies and the U.S. airline industry that focuses on the intersection of the two
strands of literature, i.e., retrospective analyses of airline mergers. Section 1.3 briefly documents
the Southwest (WN)/AirTran (FL) merger. In Section 1.4, I present the data used, the construction
of the sample, and the summary statistics. In Section 1.5, I describe the empirical strategy of the
study, including specification. Section 1.6 presents the results, and I discuss and draw conclusions
in the last section.
8
Another example of a segmented market is the hotel industry, where luxury hotels generally target ‘luxury seek-
ers’ whereas budget hotels are likely to target price-sensitive customers. In this industry, a merger between luxury
hotels and a merger between budget hotels could have different impacts on the industry.
9
Kwoka et al. (2016) also emphasized the segmented nature of competition in the airline industry. They found
that the pricing of legacy airlines only affects other legacy airlines, whereas LCC pricing affects the pricing of both
legacy airlines and other LCCs.
10
Small LCCs have made an effort to expand to compete with larger airlines in various ways. For example, JetBlue
(B6) started a partnership with American Airlines (AA) in 2020, whereas more recently, Frontier (F9) and JetBlue
(B6) made offers to acquire another LCC, Spirit (NK), in 2022.
4
1.2 Literature Review
The U.S. airline industry is one of the most examined industries, as there are many interesting
features that need to be investigated. In addition, detailed data from this industry are publicly
available.
Most of the studies on the impact of earlier airline mergers find that mergers are associated
with anti-competitive effects. For example, Borenstein (1990) finds increased airport dominance
in the Northwest/Republic merger, and that fares increased significantly on routes where these two
airlines were in direct competition prior to the merger, whereas the merger between TWA and
Ozark was not associated with strong anti-competitive effects. Werden et al. (1991) find similar
results that the TWA/Ozark merger was associated with a slight increase in fares, whereas the
Northwest/Republic merger caused a significant increase in fares. Kim and Singal (1993) also find
significant price increases on overlapping routes for multiple mergers in the 1980s relative to a
control group of routes that were unaffected by the merger.
Whereas most of the retrospective studies on mergers focus on the effects of eliminating actual
competitors, Kwoka and Shumilkina (2010) investigate the gain in pricing power from merging
with a potential competitor. They compared changes in fares on routes where a potential competitor
was eliminated due to the merger with changes in fares on other routes. Their results show that
prices rose by 5 to 6 percent on the routes where one carrier served and the other was a potential
entrant. Although the increases in prices were larger in markets where the two airlines had been
direct competitors, the elimination of a potential competitor seemed to have a significant impact
on market prices as well.
More recently, Luo (2014), Carlton et al. (2019), and Das (2019) examine the effects of legacy
airline mergers. Luo (2014) focuses on the price effects of the Delta (DL) and Northwest (NW)
merger announced in 2008. In this case study, two market types, nonstop markets and connecting
markets, are analyzed separately. The results show that the fares on the pre-merger overlapping
routes did not increase by much following the merger. Although the merger had a small price ef-
fect on connecting overlapping routes, there was no significant price effect on nonstop overlapping
5
routes. She argues that this is mainly because the competitive effects of LCCs are large, whereas
the competitive effects of legacy airlines are small. Accordingly, the DL/NW merger had little
effect, despite their large market shares, as both of the airlines were legacy carriers. Carlton et al.
(2019) examine the competitive effects of three recent mergers between legacy airlines: the merger
between Delta (DL) and Northwest (NW) in 2008, between United (UA) and Continental (CO) in
2010, and between American (AA) and US Airways (US) in 2013. After conducting a comprehen-
sive investigation of the price and output effects using difference-in-differences (DID) regression
analysis, they conclude that the recent legacy carrier mergers had an overall pro-competitive effect.
In a separate analysis of nonstop and connecting markets, the findings show that nonstop overlap
markets experienced significant output increases and fare decreases, where connecting markets did
not experience any anti-competitive effects. Das (2019) also conducts a case study of the American
(AA)/US Airways (US) merger. He concludes that prices decreased overall, but the price effects
varied by market size, which is defined as the number of passengers. Specifically, the prices de-
creased in larger markets, whereas smaller markets experienced an increase in prices. He also
investigates the quality effects of the merger using various measures of quality and finds that the
merger is associated with delay increases and cancellation decreases.
Compared with the price effects of a merger, the quality effects of a merger have been less
investigated. Bilotkach (2011) examines the effect of the merger between US Airways (US) and
America West Airlines (HP) on multimarket contact (MMC) and flight frequency. He finds that a
higher MMC level was correlated with a lower flight frequency, and concludes that just as MMC
can affect prices, it can also have an impact on non-price characteristics. Two other recent studies
examine the quality effect of mergers in the airline industry, using different quality measures.
Prince and Simon (2017) investigate the quality effects of five U.S. airline mergers, using on-
time performance (OTP) as a measure of service quality.
11
Their findings show that the airline
mergers did not have negative effects on OTP in the long run. Although they find some limited
11
The five airline mergers included are AA’s acquisition of TWA (2001), America West’s acquisition of U.S. Air-
ways (2005), Delta’s acquisition of Northwest in 2008, United’s acquisition of Continental in 2010, and Southwest’s
acquisition of AirTran in 2011.
6
OTP worsening for the relatively short post-merger periods (0-2 years), they conclude that airline
mergers are likely to result in long-run OTP improvements. Chen and Gayle (2019) examine the
quality effects of two mergers between legacy carriers, and they use routing quality as a quality
measure.
12 13
In this study, they investigate not only how the airline mergers affected quality, but
also whether the effects varied by the pre-merger competition intensity. Their findings show that
routing quality improved for routes where the merging airlines were not in competition prior to the
mergers but worsened for routes where they had directly competed.
1.3 Southwest Airlines’ Acquisition of AirTran Airways
In this section, I briefly describe the Southwest Airlines (WN) acquisition of AirTran Air-
ways (FL). In September 2010, Southwest and AirTran announced their agreement to Southwest’s
acquisition of AirTran. Even though they had some overlapping routes where they were direct
competitors of each other, the Department of Justice (DOJ) approved the merger, expecting it to
lead to an increase in consumer welfare through the offering of new services.
14
The merger was
officially realized in May of 2011, and the merging parties received a single operating certificate in
March 2012. The two airlines had integrated their system gradually, and the last flight of AirTran
operated on December 31, 2014.
Prior to the merger, both airlines were well-established LCCs. That said, neither of them had
critical financial problems, and this merger was not due to bankruptcy. While their market shares
were significantly different, Southwest being the dominant one, AirTran also had significant market
share at the time of merger. Their market shares are shown in Table 1.1. Among the LCCs, AirTran
12
The two airline mergers included in this study are the merger between Delta and Northwest (2008), and the
merger between United and Continental (2010).
13
Routing quality is the percentage ratio of nonstop flight distance to the product’s itinerary flight distance that is
used to get passengers from the origin to the destination, where a product is defined as a combination of a route and
an operating carrier.
14
A statement of the DOJ’s decision on the WN/FL merger can be found at https://www.justice.gov/opa/
pr/statement-department-justice-antitrust-division-its-decision-close-its-investigation
(last accessed in March 2022).
7
was the third-largest LCC, following the second-largest LCC, JetBlue. Therefore, Southwest was
able to increase its market share significantly by acquiring AirTran.
Even though the merger was approved by the DOJ without major challenges, there were con-
cerns regarding this merger. The first one was a general concern about the reduction in competition.
Because presence and growth of LCCs have helped to lower market fares over time, eliminating
one LCC from a market was expected to weaken the downward pricing pressure of the LCCs even
more than eliminating one legacy carrier. Another concern was the process of integration. The two
carriers had quite different operating systems in several ways. For example, Southwest was well-
known for its point-to-point system and for operation with only one type of aircraft. In contrast,
AirTran operated a hub-and-spoke system (Atlanta being the major hub) and used several aircraft
models.
15
The integration of employees was also an issue. For these reasons, the integration of
the two carriers was expected to be slow and to lead to operational inefficiencies. In fact, the inte-
gration process spanned more than two years, from the announcement in the third quarter in 2010
to the last AirTran flight in the fourth quarter of 2014. In spite of these concerns, the two airlines
succeeded in integrating and became stable as the new Southwest Airlines (WN).
1.4 Data
In this section, I present the data used in this study, as well as how the sample is constructed.
1.4.1 Sources of Data
One of the main datasets used in this study is the Airlines Origin and Destination Survey
(DB1B) data from the Bureau of Transportation Statistics (BTS). It is a quarterly 10 percent sample
of domestic airline tickets, as reported by the reporting carriers Each observation is an itinerary
with an itinerary fare. The data contain detailed information on each itinerary, such as itinerary
15
Southwest used Boeing 737 planes only, whereas AirTran used both Boeing 717 and 737 planes. After acquiring
AirTran, Southwest made a deal with Delta, and Delta leased the 717 planes from Southwest because Southwest
wanted to only use Boeing 737 planes.
8
fares, number of passengers who purchased the ticket at the price, origin and destination cities, and
origin/connecting/destination airports. Even though these data do not include passenger-specific
information or some of the restriction information on tickets, they have been used frequently in
studies of the U.S. airline industry.
16
Another main source of data used to construct quarterly flight frequencies is the U.S. domestic
flight schedules data from the Official Airline Guide (OAG). These data contain detailed informa-
tion on flights operated in the U.S, such as flight numbers, scheduled departure/arrival dates and
times, aircraft types, and total number of seats in the aircraft. The OAG schedule data is similar to
the OTP data, which is provided by BTS, but it has an important advantage in terms of the research
purpose of this study. Unlike the OTP data, where schedules are reported for reporting carriers,
schedules in the OAG data are reported for ticketing carriers. A ticketing carrier is the airline that
sells the ticket under its airline code, whereas an operating carrier is the airline whose aircraft is
used to operate the flight.
17
A reporting carrier is the airline that reports to the BTS and is usually
the operating airline who operated the first segment of the itinerary. Because ticketing carriers are
the airlines that interact with the consumers, merger effects need to be examined in terms of the
behavior of ticketing carriers. In other words, the OAG data are more suitable for investigating
merger effects than the OPT data.
18
I supplement the DB1B and the OAG data with a couple of other datasets to construct control
variables. Population data are included in the analysis, and is from the U.S. Census Bureau.
19
Although population does not change dramatically over time, especially given that it is annual
data, it is included to capture market size as well as potential economies of traffic density. In order
to capture changes in economic conditions, unemployment rate data from the U.S. Bureau of Labor
Statistics (BLS) are used.
16
Some of the data includes fare class information. However, this information is not recommended for use in any
analysis, as fare classes are likely to be different across different airlines.
17
A ticketing carrier and an operating carrier of a flight trip may differ. For example, a ticket may be sold by
American Airlines under its code AA, but the flight can be operated by Alaska Airlines (AS). This type of product is
called a ‘codeshare’ product.
18
Marketing carrier OTP data is available from January 2018.
19
Annual Estimates of the Resident Population: April 1, 2010 to July 1, 2016 Source: U.S. Census Bureau, Popu-
lation Division
9
1.4.2 Sample Construction
Two separate samples are constructed to estimate price effects and quality effects, respectively.
For both the price effects and quality effects, a market is defined as a directional pair of origin
and destination airports.
20
For example, a trip from Los Angeles International Airport (LAX) to
John F. Kennedy International Airport (JFK) is treated as a different market than a trip from JFK
to LAX. Only airports that are in the 48 contiguous states are included. As a result, any particular
characteristic of non-contiguous states are irrelevant to this study. To focus on the markets that are
significant, only those markets covered by the data for all sample periods are included.
21
Only mainline airlines such as Southwest (WN), AirTran (FL), and Delta (DL) are included.
22
Although there may be some other regional airlines in the markets, most of them have little traffic.
Because the market shares of the included airlines are more than 90% for most of the markets,
excluding the other airlines is unlikely to have a critical impact on the results.
23
1.4.2.1 Price Effects Sample
The price effects sample is constructed from the DB1B data. A basic observation unit—before
collapsing—is an itinerary on a market. Several restrictions are applied to itinerary-level observa-
tions. First, only roundtrip itineraries are included.
24
Any itinerary that is either a bulk fare or is
suspected to be an incredible fare is excluded.
25
Itineraries that include ground transportation, have
multiple ticketing carriers, or have more than two trip breaks (multi-directional trips per one way)
20
In the empirical literature regarding the airline industry, the two most common ways to define a market are to
use routes either between city pairs or between airport pairs. Either definition can be suitable, depending on the
research questions and assumptions of the study. In this study, I use airport pairs because it is more reasonable and
straightforward to calculate flight frequency using an airport pair market definition.
21
The markets that were not included in the data for all sample periods are likely to be small or seasonal markets.
22
The legacy airlines included are American (AA), Alaska (AS), Continental (CO), Delta (DL), Hawaiian (HA),
Northwest (NW), United (UA), US Airway US, and the LCCs are JetBlue (B6), Frontier (F9), AirTran (FL), Allegiant
(G4), Spirit (NK), Sun Country (SY), ATA (TZ), Virgin America (VX), and Southwest (WN).
23
I also estimate the sample with all airlines included, and the results are qualitatively the same.
24
Another often-used selection is to include both roundtrip and one-way itineraries, and to treat roundtrip itineraries
as two separate one-way trips. Under this approach, ticket fares for the half-roundtrip one-way tickets are assumed
to be half of the roundtrip fare. One advantage of including only roundtrip itineraries is that fare information can be
used as the raw data. Moreover, roundtrip tickets enable the identification of real origin airports, so that the potential
difference in impact of origin and destination characteristics can be examined.
25
The DB1B data specifies whether the ticket information is credible or not.
10
were dropped from the sample. To exclude rewards tickets and anomalies, tickets with too-low or
too-high fares are not included.
26
Any ticket with more than two coupons is also excluded. In other
words, the tickets included in the sample can have at most one stop for each directional trip. Most
of these restrictions are common in studies of the airline industry. After applying the restrictions,
I collapse the data into market-time-ticketing carrier-level observations. In other words, itineraries
by a ticketing carrier on a market at a time (quarter) are averaged to be one observation unit. For
the fare variable, a passenger-weighted average fare is used. For the price effects, I restrict sample
markets to those with at least 1,800 passengers per quarter. This is because small markets may
have different features than do large markets, and some of the markets having little traffic may be
seasonal markets.
1.4.2.2 Quality Effects Sample
For a quality measure, the market-carrier level flight frequency is used. Flight frequencies are
considered an important air travel quality measure because high frequencies enable consumers to
have more flexible schedule adjustments. Moreover, high frequencies make it easier for travelers to
find a flight that is close to his/her preferred schedule (Brueckner and Flores-Fillol, 2007; Berry and
Jia, 2010; Bilotkach, 2011; Brueckner and Luo, 2014). The quality effects sample is constructed
from the OAG flight schedule data. Although the raw data is at the flight level, the quarterly
flight frequencies are calculated to be consistent with the price data. Compared with the price
effects sample, the quality effects sample is much smaller. This is mainly because markets are
only defined for nonstop flights, and airlines are included only if they operate nonstop flights on
the market. I collapse the data into market-time-ticketing carrier-level observations, and for the
flight frequency variable, total quarterly frequencies are used.
26
The lower bound is $30 and the upper bound is $1,500.
11
1.4.2.3 Overlap Market Definitions
While a market is defined as a directional origin and destination airport pair, defining an overlap
market can be tricky, especially given that the DB1B data is a 10% sample of all tickets sold. Not
only did airlines enter and exit certain routes during the sample period, but not all flights that were
operated on the routes are included in the dataset. For this reason, I use two different definitions for
overlap markets. Definition A includes routes where both of the merging airlines operated at least
one quarter prior to the merger, and at least one of them operated through the pre- and post-merger
periods. Definition B includes routes where the merging airlines operated during all the sample
pre-merger period, and at least one of them operated through the pre- and post-merger periods.
Definition B is stricter, whereas Definition A includes more observations.
I focus my analysis on the balanced sample, which is balanced in terms of market-carrier
presence. As mentioned above, there are two main reasons for airlines to be in or out of the data.
One possibility is that the airlines did, in fact, enter or exit the market. The other reason is that
the DB1B data are a 10% sample. In other words, even if an airline was on the market during
the full sample period, it is possible that the airline is shown in the data only during the subset of
the sample period. Given these two main cases, the airlines included in the balanced sample are
more likely to be regular and significant players on the markets. Although similar results from the
balanced panel and the unbalanced panel do not fully exclude the possibility of a selection bias,
the unbalanced sample results are also reported for comparison and robustness.
Figure 1.2 and 1.3 provide time trends of the averages of the ticketing airline-level average fare
and quarterly total flight frequency on Overlap routes. Note that in these figures, overlap markets
are defined based on the merging airlines’ presence in each time period. For all four graphs, the
dashed vertical lines indicate the merger announcement date, and the solid vertical lines indicate
the merger closing date. For both of the figures, the left graphs show separately the averages
calculated for the merging airlines and the non-merging airlines, and the non-merging averages are
divided into the average for the legacy airlines and the average for the other small LCCs in the
right graphs.
12
While the average fares of both the merging airlines and the non-merging airlines were gen-
erally increasing until early 2012, the gap between the two averages seem to decrease after the
merger announcement. This may be related to the findings of previous airline merger studies that
say that merging airlines exert their increased market power right after the merger initiation, but
efficiency gains take more time to realize. In other words, merger effects can be anti-competitive
in the short run but can become pro-competitive in the long run. The right graph shows that the
average fare of the merging airlines increased rapidly after the announcement, especially compared
with the average fare of other LCCs. Because prices are strategic complements, the non-merging
airlines might have been affected by the increased pricing of the merging airlines. The legacy
airlines may have taken advantage of the merging airlines’ pricing, whereas the other LCCs did
not seem to follow suit. Another interpretation could be that the newly merged airline was able
to increase its market power as a stronger LCC, so that its prices were raised to the level of other
LCC rivals’ prices.
For the average flight frequency (Figure 1.3), the graphs are likely to be less informative be-
cause the sample size, especially that of the non-merging airlines, is small. Whereas the average
flight frequency of the merging airlines is consistent across time, the average of the non-merging
airlines increased substantially after the consummation of the merger. The right graph shows that
this increase was mainly from the legacy airlines, whereas the flight frequency of the small LCCs
is relatively stable. The volatility of the averages of the non-merging airlines in the right graph
is likely due to the small sample size. Although the two figures display the general time trends
of average fare and flight frequency in overlap markets, a more detailed estimation is needed to
examine the merger effects, because potentially significant factors are not considered.
The final sample includes three quarters of 2010 as the pre-merger period, and three quarters
of 2012 as the post-merger period. The pre-merger period is chosen to be prior to the merger
announcement time, and the post-merger period is after a single operating certificate was given.
Although the merger impact is expected to be different according to the time period (e.g., the short
run impact vs. the long run impact), 2012 is chosen to capture the potential increase in market
13
power and efficiency.
27
Table 1.2, 1.3, 1.4, and 1.5 show the summary statistics of the sample
used.
1.5 Empirical Strategy
One simple way to estimate merger effects is to compare pre-merger outcomes with post-
merger outcomes. However, this comparison would accurately capture the effects only if there
is no change in other factors that affect outcomes over time. Even though these comparisons
could give a simple look at the changes in outcomes, this would not be sufficient to capture the
effects of the merger, as the underlying assumption is highly unlikely. Thus, demand and cost
conditions that might have affected the outcomes should be controlled to determine the merger
impact. Another approach is to directly estimate the relationship between outcomes and all (or
most) of the relevant factors that affect demand and supply. Although this type of approach can
identify the merger effects by controlling the relevant determinants, identifying and including all
the relevant market determinants could be difficult. Thus, I use a version of the difference-in-
differences (DID) method in this study. DID is a standard econometric method based on comparing
changes in outcomes of affected groups to changes in outcomes of unaffected groups without
identifying the underlying demand and supply determinants. Although this method has often been
used in investigating treatment effects in various studies, I briefly describe how it works in the
context of merger effects in the airline industry.
For the airline industry, two main ways have been used to define an affected group (treatment
group) and an unaffected group (control group). One is at the market level. According to this
approach, markets that involve merging airlines are considered to be the affected group. For ex-
ample, a treatment group could be markets that were served by both of the merging parties prior to
the merger—overlapping routes. These markets are considered to be most affected by the merger,
27
It is conventional to exclude some periods between the pre- and post-merger periods because the integration
process takes time, and market power and efficiency gain are likely to be realized considerably later than the merger
transaction.
14
because competition between two airlines was removed right after the merger. This is why an-
titrust authorities are most concerned about these markets when it comes to evaluating a proposed
merger. On the contrary, a control group should be those who were not affected by the merger,
but that capture changes in other factors such as fuel price changes or common demand shocks.
Markets where neither of the merging parties operated prior to the merger (and after the merger)
are usually used as a control group. Defining groups at the market level is more likely to be done
to examine price effects (Carlton et al., 2019; Das, 2019; Capodaglio, 2019).
The other way to define an affected group and an unaffected group is at the airline level. With
this definition, the treatment group includes the merging airlines, whereas the control group in-
cludes the other airlines on relevant markets. The airline level definitions are more likely to be
used to examine quality effects (Prince and Simon, 2017; Chen and Gayle, 2019). Because the
main focus of this study is the merger effects on the merging airlines in relevant markets, I use
the non-merging airlines as a control group to understand how the behavior of the merging airlines
was changed, focusing on overlapping routes.
1.5.1 Two-Way Fixed Effects DID regressions
In order to estimate the merger effects, outcome equations are specified as follows:
lnOutcome
imt
=b
0
+b
1
WN=FL
im
Post
t
+Controls+g
im
+l
t
+e
imt
;
(1.1)
where i is an airline, m is a market (an airport pair), and t is a time (quarter). lnOutcome
imt
is either
lnFare
imt
or lnFF
imt
. lnFare
imt
is the natural log of the passenger-weighted average fare of airline
i on market m at time t. In a similar way, lnFF
imt
is the natural log of the quarterly flight frequency
of airline i on market m at time t. WN=FL is a dummy variable that indicates the merging airlines—
Southwest and AirTran. Post is a dummy variable indicating the post-merger periods. Controls
are control variables included to capture the effects of potential time-varying factors. Population,
number of airlines in the market, and unemployment rates are included as controls. Although
15
some studies use average population as a population measure,
28
I include origin and destination
population separately to observe if the outcomes are more sensitive to the origin characteristics.
The unemployment rates are used at the Metropolitan/Micropolitan statistical area (MSA) level,
and if the MSA-level information is not available, state level rates are used instead. g
im
is the
market-airline fixed effects, whereas l
t
is the time fixed effects. Lastly, e
imt
is the idiosyncratic
shock at the market-airline-time level.
The main variable of interest is WN=FL Post. Because the outcomes are in natural log
form, the parameter b
1
gives the percentage difference between the changes in the outcomes of
the merging airlines and those of the non-merging airlines on the relevant routes. In order words,
by using the DID model with the non-merging airlines as a control, b
1
shows the average merger
effects on the merging airlines on the overlapping routes.
1.6 Results
1.6.1 Market Level Descriptive Analysis
Although the focus of this study is to examine changes in the outcomes of the merging air-
lines compared with those of the non-merging airlines on overlapping routes, examining changes
in market-level outcomes resembles treatment/control definitions at the market level. These com-
parisons can provide a broad view of how the airline industry changed between before and after
the merger. To compare market-level changes, the markets are categorized into five (six) exclusive
groups: Overlap, Either, Exit, Enter, and Neither markets. Overlap markets are routes where both
Southwest and AirTran operated during the pre-merger period and at least one of them operated
after the merger.
29 30
Either markets are routes where only one of the merging airlines operated
28
For example, Berry and Jia (2010) uses a geometric mean of population of origin and destination as a market
population.
29
For the descriptive analysis, Overlap Definition A is used.
30
Although the merger was closed in 2011 and the merging airlines received a single operating certificate in 2012,
there were flights under the AirTran code (FL) operated until 2014. As a result, not all flights of the new Southwest
were operated under the Southwest code WN.
16
during the pre-merger period, without the other, and where at least one of them operated after
the merger. To see whether the routes with only Southwest were different from the routes with
only AirTran, Either markets are further divided into WN markets and FL markets.
31
Exit markets
are routes where at least one of the merging airlines operated prior to the merger, but neither of
the merging airlines were on the market during the post-merger period. Enter markets are routes
where at least one of the merging airlines operated during the post-merger period, but neither of
them operated during the pre-merger period. Enter markets are related to the network expansion
following the merger, whereas Overlap markets are the most relevant markets when it comes to
ex-ante merger evaluation. Lastly, Neither markets are routes where neither of the merging airline
operated during the sample period. These markets are considered to be the markets least affected
by the merger because the merging airlines were never directly on these markets.
For the different types of markets, market-level passenger-weighted average fares and total
number of flights are calculated for the pre- and the post-merger periods, respectively, and the
differences between the pre- and the post-merger periods are compared with those of Neither mar-
kets.
32
Given that a subset of Neither markets is usually used as a control group to estimate market-
level merger effects, these comparisons can provide a bigger picture of how the routes involving
the merging airlines behaved compared with the routes that were not (or were less) affected by the
merger.
Table 1.6 reports the comparison results of market-level average fares. To focus on routes with
significant traffic, only markets with at least 1,800 roundtrip passengers are included. Columns
(1) and (2) report market-level passenger-weighted average fares in the pre and post periods, re-
spectively, and Column (3) shows their differences. In the last column, the difference in Neither
markets is subtracted from each market’s difference. The values in the last column are essentially
simple 2-by-2 DID calculations, using Neither markets as the control group. Compared with the
fare changes in Neither markets, Overlap markets experienced larger fare changes. Similarly, fare
31
Either markets are almost clearly divided into WN markets and FL markets. In other words, there are very few
markets where one of the merging airlines operated prior to the merger, and the other operated after the merger.
32
To see the changes in outcomes level, the level variables are used rather than the log values.
17
changes are larger in Exit markets. On the contrary, Enter markets experienced a lower price in-
crease than did Neither markets. This seems intuitive, given the well-known Southwest effect, i.e.,
Southwest’s fare-lowering impact (Morrison, 2001). For Either markets, no significant difference
from Neither markets was found.
Although these results show the market-level price changes between before and after the merger,
there is one issue that should be addressed when it comes to interpreting the results. As described
earlier, the values in Column (4) can be understood as DID estimates, using the Neither markets as
the control. In this setting, each market type indicator is the treatment dummy. Because these treat-
ment dummies are defined based on the data, which is a 10% sample, these dummies are subject to
a measurement error problem. For example, a market that is defined as an Either market, based on
the presence of Southwest on the market, might have been an Overlap market, where AirTran (FL)
was too small to be included in the data. All of the data-based market definitions are subject to
this problem, except Overlap markets.
33
As has been documented in many studies of misclassified
binary variables (e.g., Mahajan 2006; Lewbel 2007; Nguimkeu et al. 2021), the estimates in this
case are likely to be biased downward.
34
Table 1.7 shows the comparison results of market-level total flight frequencies. To exclude
markets that are too small, only markets with at least 90 flights per quarter are included. Unlike
the results from the fare comparisons, Overlap markets did not experience any significant impact,
compared with Neither markets. On the contrary, the change in Exit markets was smaller, whereas
the change in Enter markets was larger. In fact, whereas most of the market categories experienced
decreases in total flight frequencies on average, the total flight frequency increased from 322.25 to
462.88 in Enter markets.
Although the market-level frequency can provide a useful overview of how each type of market
changed, it is not likely to be a good measure for a pre-post comparison because the market-level
total numbers are more sensitive to the number of airlines in the market. To address this issue, I
compare the averages of market-carrier-level flight frequencies. The results are reported in Table
33
If the two merging airlines were present on a market, the market could not have been any other type.
34
The related literature is well-documented in a recent work, Nguimkeu et al. (2021).
18
1.8. The resulting sign of Exit markets and Enter markets are opposite, compared with the signs of
those markets from the market-level results (Table 1.7). Even without comparing with the change
in Neither markets, the airlines in the Exit markets had increased flight frequency, whereas the
airlines in the Enter markets had decreased flight frequency from 252.98 to 220.92. Because the
signs of the results of the market level and the market-carrier level are different, it is likely that the
market-level results are due to changes in the number of airlines in the markets. Other than the
Exit and Enter markets, the result of the Overlap markets is not significant, and the Either markets
showed a larger decrease in flight frequency than did the Neither markets.
The results of the descriptive analysis show that if the market-level merger effects had been
examined using Overlap/Either markets as a treatment group and Neither markets as control
group, no pro-competitive effects would have been found. The results show the merger to be anti-
competitive. However, the pro-competitive effect of network expansion could have been detected,
as the DOJ’s decision had anticipated.
1.6.2 Price Effects of the Merger
Table 1.9 shows the price effects results for the balanced panel case. The first three columns
are the results from the Definition A sample, and the the last three columns are the results from the
Definition B sample. In Columns (1) and (4), I report the results of OLS regressions, which include
the treatment dummy and the post-period dummy instead of the fixed effects. These OLS results
are the same as simple 22 DID calculations. The main reason for including the OLS results is to
see the coefficients for WN=FL and Post, which are informative of general differences between the
merging airlines and the others, and between pre- and post-merger periods, respectively. Compared
with the other non-merging airlines (most of which are legacy carriers), Southwest and AirTran’s
prices are lower as expected. The post-merger period average fares are higher than the average
pre-merger period fares.
Columns (2) and (5) are the fixed effects results, which were estimated without control vari-
ables, whereas Columns (3) and (6) are the results using the population and unemployment rates as
19
control variables. Across the various specifications and two alternative overlap market definitions,
the price effects are consistently significant. After the merger, the merging airlines’ fares increased
by about 5.5% on the overlapping routes. In other words, the changes in the prices of the merging
airlines were about 5.5% higher than those of the non-merging airlines on the overlap markets.
Given that the overlap markets experienced the largest fare increases, compared with the other
types of markets (Table 1.6), the price increases by the merging airlines on the overlap markets
could be significant when all markets are considered.
Although it is not the main focus of this study, the influence of the control variables is worth
mentioning. First, the unemployment rate is negatively associated with fare increase, as expected.
Because higher unemployment rates tend to mean less demand for air travel, this negative relation-
ship was anticipated. For the population variables, origin population size had a negative effect on
fares. Because I use roundtrip itineraries, origin market characteristics may have a stronger impact
than destination characteristics. This is partly supported by the results that show that destination
population size is not significant. The negative relationship may be due to economies of traffic
density (Brueckner and Spiller, 1994), but more detailed analysis would be required to identify the
exact mechanism because large population could also imply high demand in general.
Other specifications include number of airlines, number of legacy airlines, and number of other
LCCs on the market as control variables. However, the results are not reported, mainly because
the price effects are consistent and similar to ones that are reported. Even if these variables are
important market competition measures, they are likely to be endogenous, and in most cases, the
results show that they are insignificant.
Table 1.10 shows the price effect results for the unbalanced panel case, and the results are
qualitatively similar to those for the balanced panel.
1.6.3 Quality Effects of the Merger
Table 1.11 shows the quality effects results for the balanced panel case. Similar to the tables
containing the price effects results, the first three columns are from the Definition A sample, and
20
the last three columns are the results from the Definition B sample. In the case of the Definition
A sample, there is no evidence of quality effects for the merger, whereas in the last two columns,
anti-competitive effects are detected. According to Columns (5) and (6), the merging airlines’
quarterly flight frequency was about 20-30% lower after the merger.
Even though the results show somewhat anti-competitive effects in quality, these results should
be interpreted with caution, especially given that the sample size is fairly small. Because the
overlap markets are markets that already included Southwest and AirTran, it is likely that there
are few other airlines that operate nonstop flights in these markets. As a result, the number of
control group airlines is small, especially in the balanced panel case. Nevertheless, the results do
not show pro-competitive quality effects. Given both market level and market-airline level flight
frequency decreased—not only on an absolute level, but also compared with the change of the
Neither markets (Table 1.7 and 1.8), none of the results provide a pro-competitive effect of the
merger on quality, which might have offset the anti-competitive price effects.
Table 1.12 shows the quality effects results for the unbalanced panel case, and the merger
effects results are qualitatively similar to those using the balanced panel.
1.6.3.1 Capacity Effects of the Merger
Airlines may use a large aircraft with low frequency instead of a small aircraft with high fre-
quency (Berry and Jia, 2010; Pai, 2010). If this is the case, low frequency itself should not be
understood as simple quality deterioration, even though consumers may find that it is more dif-
ficult to find a flight schedule that suits his/her preference. To determine if the merging airlines
compensated for relatively lower frequency by increasing the number of seats provided, I estimate
the capacity effects of the merger.
Table 1.13 shows the estimation results for all four panels. The results from the balanced panel
are reported in the upper part, and the results for the unbalanced panel are reported in the lower
part. All specifications include the fixed effects, but only Columns (2) and (4) include the control
variables. Only the coefficients of interest are reported. The results show that there is no evidence
21
of relative increases in the number of seats provided by the merging airlines on the overlapping
routes. In fact, the number of seats of the merging airlines was about 15-20% lower after the
merger in the Definition A case (Columns (1) and (2)). The anti-competitive effects seem to be
stronger when I use the stricter overlap definition (Columns (3) and (4)). These results show that
the merging airlines also decreased number of seats provided. In fact, Southwest did not change
aircraft used (Boeing 737) following the merger, which means the decrease in capacity was likely
to be from the decrease in flight frequency.
1.6.3.2 Price Effects for Nonstop Products
One limitation of the price effects analysis in Section 1.6.2 is that it is not straightforward
enough to complement quality effects. Specifically, the flight frequency of a connecting flight is
not easy to construct given the available data. This means that while we can conclude that the
changes in the fares of the merging airlines are larger than the ones of the non-merging airlines, we
cannot determine whether this was due to quality improvement or the exertion of increased market
power. In a similar way, a relative decrease (or no change) in flight frequency is more likely to
be understood as quality worsening if it can be supported by anti-competitive price effects results.
To address this issue, I match the flight frequency sample with the DB1B data and estimate if the
products included in the quality effects analysis experienced any price effects. Unlike the price
effects analysis in Section 1.6.2, this sample includes only direct flights.
The results are reported in Table 1.14 for all four panels. The results from the balanced panel
are reported in the upper section, and the results for the unbalanced panel are reported in the lower
section. All reported specifications include the fixed effects and the control variables. Columns
(1) and (3) are from Table 1.11 and 1.12 (i.e., the quality effects results), and Columns (2) and (4)
are the results for the price effects for nonstop flights. As shown by the numbers of observations,
most of the products included in the quality analysis are matched by the DB1B data. Whereas
the frequency results show relative decreases for the merging airlines, the price effects results
show that the relative price changes are larger for these flights. In fact, compared with the results
22
of the price effects in Section 1.6.2, which indicate about 5.5% increase in prices, the results
for nonstop flights show a higher percentage increase of 8%. In other words, if I restrict the
sample to nonstop flights, the merging airlines’ fares increased by about 8% after the merger on
the nonstop overlapping routes. These results show that the relative decrease in flight frequency
was not accompanied by a relative decrease in prices. Therefore, the new merged airline seems to
have restricted supply of direct flights to raise fares.
1.6.4 The Parallel Trends Assumption
As stated in Goodman-Bacon (2021) and many other studies using two-way fixed effects re-
gression models (e.g., De Chaisemartin and d’Haultfoeuille, 2020; Imai and Kim, 2021), a causal
interpretation of two-way fixed effects DID estimates depends on a parallel trend assumption and
constant treatment effects. In order to test for the parallel trend assumption, I estimate the merger
effects using an event study approach. A standard event study pre-trends test was conducted in Au-
tor (2003), and more recently, other studies used similar methods to check that treatment groups
and control groups did not show statistically different trends prior to the events (e.g., He and
Wang, 2017; Russell, 2021). Pre-treatment trends can be tested by estimating the following equa-
tion (equation 1.2) and testing whether the coefficients on the leads are statistically different from
zero (Roth et al., 2022):
lnOutcomes
imt
=a
0
+b
tå
t
WN=FL D
t
+Controls+g
im
+l
t
+e
imt
=a
0
+b
3
WN=FL D
3
+b
2
WN=FL D
2
+b
0
WN=FL D
0
+b
1
WN=FL D
1
+b
2
WN=FL D
2
+Controls+g
im
+l
t
+e
imt
;
(1.2)
23
where WN=FL is a dummy variable to represent the merging airlines, as in Section 1.5.1. D
t
is a
time dummy. The last pre-merger period is excluded so it can be used as a base.
35
The results are reported in Tables 1.15 and 1.16. Table 1.15 shows the results of the price ef-
fects pre-trends test for the balanced panel. Columns (1) and (3) are from Table 1.9. In Columns
(2) and (4), the base time is Pre 1, and the coefficients on Pre t and Post+ t show whether
the differences in the outcomes of the merging airlines and the non-merging airlines are statisti-
cally different compared with the differences in Pre 1. As shown in Columns (2) and (4), the
differences between the merging airlines and the non-merging airlines are significant for the post
periods, whereas there is no evidence of significant pre-trends in the pre-merger periods. Table
1.16 also shows that there is no evidence of pre-trends for the quality effects sample, while the
post effects are also weakly significant.
Figures 1.4, 1.5, 1.6, and 1.7 show the estimated coefficients and their 95% confidence inter-
vals. Figures 1.4 and 1.5 show that the merging airlines’ average fares were statistically higher
than the non-merging airlines fares during the post period. On the contrary, most of the post period
coefficients for the quality effects are not statistically significant at the 95% level. Even though
they are less significant, the negative post period coefficients support that there is no evidence of
pro-competitive effects.
1.7 Conclusion and Discussion
This study investigates the effects of the merger between two LCCs, Southwest Airlines (WN)
and AirTran Airways (FL). The price effects and the quality effects are estimated. The findings
show that changes in the average fares of the merging airlines are larger than those of the non-
merging airlines on routes where the merging airlines competed with each other directly prior to
the merger. Given that these types of markets experienced larger price increases compared with the
other markets (i.e., routes where only one of the merged airlines operated or routes where neither
35
Sun and Abraham (2021) document that most recent studies that conduct event study specification exclude rela-
tive periods close to the initial treatment.
24
of them operated prior to the merger), the results suggest that the merger is associated with price
increases by the merging airlines.
The quality effects of the merger are found to be either insignificant or somewhat anti-competitive,
depending on the sample used. The results show no evidence of pro-competitive quality effects and
the capacity effects, and the anti-competitive price effects associated with the nonstop flights are
found to be significant, which means the relative decreases in quality were not accompanied by
relative increases in capacity or the relative decreases in prices.
There are caveats regarding this interpretation. First, welfare implications should not be di-
rectly drawn from the price effects results without information about demand or associated prod-
uct quality, although the price effects appear to be significant. As noted in the DOJ’s decision
statement, this merger was expected to offer new products and improve consumers’ welfare, on
average. Thus, further research on other types of markets would complement the findings of this
study. That said, even if the merger had led to a larger network, the anti-competitive effects on
overlapping routes should be considered when evaluating the merger and the effectiveness of the
policy decision, because the effects seem to be substantial in the case of this merger, unlike the
other mergers involving legacy carriers.
The second caveat is one that is mentioned in Subsection 1.6.1: measurement error. Other
than the misclassified treatment indicator problem, which is already mentioned, the fact that the
DB1B data are a 10% sample is likely to hinder accurate estimation. Nevertheless, the findings
can be understood as general results for the airlines that were significant on the market. For future
research, these issues can be addressed using different estimation methods.
As with most merger retrospective studies, the results of this study cannot be simply gener-
alized and applied to other industries. However, this case study has important implications for
mergers of firms with significant roles in differentiated (and somewhat segmented) markets. In a
segmented market, a merger in one segment could have a different impact than a merger in another
segment, depending on the basis of the segmentation. Thus, more disaggregated examinations
would be beneficial.
25
Because price increase effects were detected, the results will help to evaluate the effectiveness
of the policy decision on this merger with further research on the extended network. For future re-
search, more flexible methods for constructing counterfactual outcomes (e.g., Arkhangelsky et al.,
2021; Athey et al., 2021) could help estimate a wider range of effects.
26
Tables and Figures
Figure 1.1: U.S. Domestic Market Share
27
Table 1.1: LCCs Market Share: RPM (%)
Year Southwest (WN) AirTran (FL)
a
JetBlue (B6) Frontier (F9) ATA (TZ)
b
2000 8.31 0.81 0.20 0.53 1.46
2001 9.26 0.94 0.68 0.56 1.72
2002 9.52 1.17 1.43 0.70 1.92
2003 9.58 1.43 2.30 0.91 2.27
2004 9.67 1.54 2.82 1.08 2.17
2005 10.39 1.94 3.46 1.17 1.11
2006 11.56 2.35 3.98 1.31 0.65
2007 11.97 2.85 4.26 1.48 0.73
2008 12.67 3.23 4.29 1.57 0.15
2009 13.56 3.34 4.21 1.46 -
2010 13.89 3.33 4.31 1.50 -
2011 14.64 3.31 4.57 1.69 -
2012 14.81 2.72 4.90 1.62 -
2013 15.34 2.05 5.08 1.47 -
2014 16.60 0.86 5.03 1.63 -
2015 17.92 - 5.18 1.88 -
2016 18.12 - 5.44 2.22 -
2017 18.03 - 5.41 2.53 -
2018 17.66 - 5.46 2.63 -
2019 16.69 - 5.44 2.92 -
Year Spirit (NK) Sun County (SY) Allegiant (G4) Virgin America (VX)
c
Total
2000 0.54 0.49 0.01 - 12.33
2001 0.69 0.40 0.00 - 14.26
2002 0.86 0.08 0.01 - 15.69
2003 0.91 0.19 0.04 - 17.64
2004 0.89 0.24 0.09 - 18.50
2005 0.73 0.31 0.18 - 19.29
2006 0.69 0.30 0.34 - 21.19
2007 0.92 0.32 0.47 0.10 23.10
2008 0.92 0.29 0.60 0.59 24.29
2009 0.89 0.23 0.81 0.99 25.49
2010 0.97 0.22 0.93 1.11 26.25
2011 1.17 0.25 0.93 1.36 27.92
2012 1.44 0.27 1.08 1.66 28.50
2013 1.81 0.32 1.19 1.62 28.88
2014 2.09 0.34 1.28 1.62 29.46
2015 2.54 0.38 1.39 1.59 30.87
2016 2.94 0.38 1.53 1.78 32.41
2017 3.29 0.40 1.59 1.83 33.07
2018 3.84 0.41 1.68 0.41 32.07
2019 4.15 0.52 1.73 - 31.45
a
AirTran Airways was acquired by Southwest (WN) in 2011.
b
ATA Airlines (TZ) ceased operations in 2008, after filing for Chapter 11 bankruptcy protection.
c
Virgin America (VX) started operations in 2007, and was acquired by Alaska Airlines (AS) in 2016.
28
Figure 1.2: Average Fare on Overlap Routes
Figure 1.3: Average Flight Frequency on Overlap Routes
29
Table 1.2: Summary Statistics: Price Effects Sample (Balanced Panel)
Variable Mean Std. Dev. Min Max
Overlap Def. A (observations: 19,638)
WN/FL .308 .462 0 1
Post .5 .5 0 1
lnFare 5.931 .266 3.664 7.248
Fare 389.897 104.899 39 1405
Population: origin (mil.) 4.249 3.858 .62 19.15
Population: destination (mil.) 4.212 3.806 .62 19.15
Unemployment rate: origin 8.756 1.87 4.333 15.467
Unemployment rate: destination 8.856 1.942 4.333 15.467
No. of Airlines 7.006 1.548 2 13
No. of Legacy Airlines 4.313 1.106 0 7
No. of Other LCCs .812 .831 0 5
Overlap Def. B (observations: 16,962)
WN/FL .325 .469 0 1
Post .5 .5 0 1
lnFare 5.919 .267 3.664 7.248
Fare 385.284 104.012 39 1405
Population: origin (mil.) 4.344 3.919 .62 19.15
Population: destination (mil.) 4.274 3.911 .62 19.15
Unemployment rate: origin 8.754 1.855 4.333 15.467
Unemployment rate: destination 8.89 1.947 4.333 15.467
No. of Airlines 7.033 1.557 2 13
No. of Legacy Airlines 4.296 1.103 0 7
No. of Other LCCs .789 .845 0 5
a
Overlap Definition A is that an overlap markets is defined as a directional airport pair
where both of the merged airlines operated at least one quarter prior to the merger, and at
least one of them operated through pre and post merger periods. Overlap Definition B is that
an overlap market is defined as a directional airport pair where both of the merged airlines
operated during the sample pre-merger period, and at least one of them operated through
pre and post merger periods.
30
Table 1.3: Summary Statistics: Price Effects Sample (Unbalanced Panel)
Variable Mean Std. Dev. Min Max
Overlap Def. A (observations: 24,244)
WN/FL .28 .449 0 1
Post .463 .499 0 1
lnFare 5.913 .293 3.664 7.281
Fare 385.672 113.509 39 1452
Population: origin (mil.) 4.202 3.797 .62 19.15
Population: destination (mil.) 4.181 3.763 .62 19.15
Unemployment rate: origin 8.802 1.89 4.333 15.467
Unemployment rate: destination 8.911 1.961 4.333 15.467
No. of Airlines 7.095 1.55 2 13
No. of Legacy Airlines 4.382 1.113 0 7
No. of Other LCCs .825 .834 0 5
Overlap Def. B (observations: 20,776)
WN/FL .29 .454 0 1
Post .464 .499 0 1
lnFare 5.903 .289 3.664 7.281
Fare 381.558 112.153 39 1452
Population: origin (mil.) 4.294 3.84 .62 19.15
Population: destination (mil.) 4.252 3.872 .62 19.15
Unemployment rate: origin 8.799 1.875 4.333 15.467
Unemployment rate: destination 8.945 1.97 4.333 15.467
No. of Airlines 7.12 1.562 2 13
No. of Legacy Airlines 4.363 1.112 0 7
No. of Other LCCs .804 .849 0 5
a
Overlap Definition A is that an overlap markets is defined as a directional airport pair
where both of the merged airlines operated at least one quarter prior to the merger, and
at least one of them operated through pre and post merger periods. Overlap Definition B
is that an overlap market is defined as a directional airport pair where both of the merged
airlines operated during the sample pre-merger period, and at least one of them operated
through pre and post merger periods.
31
Table 1.4: Summary Statistics: Quality Effects Sample (Balanced Panel)
Variable Mean Std. Dev. Min Max
Overlap Def. A (observations: 660)
WN/FL .873 .334 0 1
Post .5 .5 0 1
lnFF 5.371 .696 .693 6.709
FF 262.976 163.729 2 820
lnSeat 10.257 .727 5.613 11.747
No. of Seats 35679.85 24721.818 274 126427
Population: origin (mil.) 3.415 2.267 .62 13.013
Population: destination (mil.) 3.415 2.267 .62 13.013
Unemployment rate: origin 8.801 2.002 5.733 15.467
Unemployment rate: destination 8.801 2.002 5.733 15.467
No. of Airlines 2.306 .743 1 5
No. of Legacy Airlines .218 .449 0 2
No. of Other LCCs .221 .415 0 1
Overlap Def. B (observations: 504)
WN/FL .929 .258 0 1
Post .5 .5 0 1
lnFF 5.395 .758 .693 6.709
FF 276.248 174.71 2 820
lnSeat 10.261 .785 5.613 11.747
No. of Seats 36685.772 25309.538 274 126427
Population: origin (mil.) 3.356 2.218 .62 9.528
Population: destination (mil.) 3.356 2.218 .62 9.528
Unemployment rate: origin 8.719 1.894 5.733 14
Unemployment rate: destination 8.719 1.894 5.733 14
No. of Airlines 2.29 .488 1 3
No. of Legacy Airlines .115 .319 0 1
No. of Other LCCs .202 .402 0 1
a
Overlap Definition A is that an overlap markets is defined as a directional airport pair where
both of the merged airlines operated at least one quarter prior to the merger, and at least one of
them operated through pre and post merger periods. Overlap Definition B is that an overlap
market is defined as a directional airport pair where both of the merged airlines operated
during the sample pre-merger period, and at least one of them operated through pre and post
merger periods.
32
Table 1.5: Summary Statistics: Quality Effects Sample (Unbalanced Panel)
Variable Mean Std. Dev. Min Max
Overlap Def. A (observations: 794)
WN/FL .839 .368 0 1
Post .491 .5 0 1
lnFF 5.033 1.168 0 6.709
FF 228.684 168.943 1 820
lnSeat 9.915 1.198 4.248 11.747
No. of Seats 30984.023 25019.215 0 126427
Population: origin (mil.) 3.474 2.409 .62 13.013
Population: destination (mil.) 3.474 2.407 .62 13.013
Unemployment rate: origin 8.844 2.036 5.733 15.467
Unemployment rate: destination 8.846 2.033 5.733 15.467
No. of Airlines 2.395 .788 1 5
No. of Legacy Airlines .288 .513 0 2
No. of Other LCCs .23 .421 0 1
Overlap Def. B (observations: 557)
WN/FL .88 .326 0 1
Post .508 .5 0 1
lnFF 5.187 1.09 0 6.709
FF 255.273 179.091 1 820
lnSeat 10.047 1.131 4.248 11.747
No. of Seats 33827.417 25713.48 0 126427
Population: origin (mil.) 3.284 2.157 .62 9.528
Population: destination (mil.) 3.279 2.155 .62 9.528
Unemployment rate: origin 8.719 1.873 5.733 14
Unemployment rate: destination 8.724 1.87 5.733 14
No. of Airlines 2.314 .495 1 3
No. of Legacy Airlines .135 .342 0 1
No. of Other LCCs .215 .411 0 1
a
Overlap Definition A is that an overlap markets is defined as a directional airport pair where
both of the merged airlines operated at least one quarter prior to the merger, and at least one of
them operated through pre and post merger periods. Overlap Definition B is that an overlap
market is defined as a directional airport pair where both of the merged airlines operated
during the sample pre-merger period, and at least one of them operated through pre and post
merger periods.
33
Table 1.6: Pre & Post Differences: Market Level Fares
Market Type No. of Markets Pre Post Differences Diff-in-Diff
[(2) - (1)] [(3) - Neither]
(1) (2) (3) (4)
Overlap 604 356.30 420.78 64.48** 22.48**
(1.89) (2.09) (2.81) (4.43)
Either 1,413 343.01 386.81 43.81** 1.80
(1.37) (1.60) (2.10) (4.02)
WN 1,103 342.58 385.12 42.54** 0.54
(1.55) (1.82) (2.39) (4.17)
FL 310 344.53 392.83 48.3** 6.30
(2.95) (3.34) (4.45) (5.61)
Exit 94 364.04 433.21 69.17** 27.16**
(4.80) (7.03) (8.51) (9.17)
Enter 127 453.45 470.5 17.05** -24.95**
(5.27) (5.37) (7.53) (8.27)
Neither 858 366.42 408.42 42.00**
(2.22) (2.60) (3.42)
a
Standard errors are in parentheses.
b
Column (4) difference-in-differences use Neither market as a control group.
c
** : 1% significance, * : 5% significance,y: 10% significance.
34
Table 1.7: Pre & Post Differences: Market Level Flight Frequency
Market Type No. of Markets Pre Post Differences Diff-in-Diff
[(2) - (1)] [(3) - Neither]
(1) (2) (3) (4)
Overlap 62 500.42 475.78 -24.64 -18.47
(27.79) (23.81) (36.59) (37.03)
Either 911 566.87 564.16 -2.72 3.45
(9.63) (10.00) (13.89) (15.00)
WN 754 523.39 517.85 -5.53 0.63
(9.89) (10.29) (14.27) (15.35)
FL 156 775.59 785.96 10.37 16.54
(27.72) (28.5) (39.75) (40.15)
Exit 25 798.16 610.91 -187.25
y
-181.09
y
(69.98) (64.61) (95.24) (95.41)
Enter 28 322.25 462.88 140.63** 146.8**
(18.38) (24.09) (30.30) (30.82)
Neither 2,786 443.76 437.6 -6.17
(3.96) (4.06) (5.67)
a
Market level flight frequency is total number of flights in a market.
b
Standard errors are in parentheses.
c
Column (4) difference-in-differences use Neither market as a control group.
d
** : 1% significance, * : 5% significance,y: 10% significance.
35
Table 1.8: Pre & Post Differences: Market-Carrier Level Flight Frequency
Market Type No. of Markets Pre Post Differences Diff-in-Diff
[(2) - (1)] [(3) - Neither]
(1) (2) (3) (4)
Overlap 62 230.39 226.91 -3.48 -2.27
(8.62) (8.33) (11.99) (12.38)
[404] [390]
Either 911 359.71 341.34 -18.37** -17.16**
(3.98) (3.75) (5.47) (6.29)
[4,307] [4,517]
WN 754 342.37 327.94 -14.43** -13.22*
(1.55) (1.82) (2.39) (4.17)
[3,458] [3,572]
FL 156 432.11 392.98 -39.13** -37.92*
(10.98) (3.34) (4.45) (5.61)
[840] [936]
Exit 25 338.2 420.35 82.15* 83.36*
(21.62) (30.38) (37.29) (37.42)
[177] [109]
Enter 28 252.98 220.92 -32.06 * -30.85
y
(11.81) (10.23) (15.63) (15.93)
[107] [176]
Neither 2,786 336.11 334.9 -1.21
(2.19) (2.20) (3.11)
[11,035] [10,921]
a
Market-carrier level flight frequency is total number of flights operated by an airline in a market.
b
Standard errors are in parentheses.
c
Number of observations are in square brackets.
d
Column (4) difference-in-differences use Neither market as a control group.
e
** : 1% significance, * : 5% significance,y: 10% significance.
36
Table 1.9: The Price Effects Results (Balanced Panel)
Overlap Definition A Overlap Definition B
OLS FE FE OLS FE FE
(1) (2) (3) (4) (5) (6)
WN/FL -0.149** -0.152**
(0.0054) (0.0056)
Post 0.144** 0.146**
(0.0044) (0.0048)
WN/FL Post 0.0549** 0.0549** 0.0547** 0.0568** 0.0565** 0.0565**
(0.0073) (0.0054) (0.0054) (0.0077) (0.0057) (0.0057)
Unemployment: Org -0.00939** -0.00885**
(0.0025) (0.0027)
Unemployment: Dest -0.00852** -0.00599*
(0.0026) (0.0028)
Population: Org -0.0969* -0.141**
(0.0420) (0.0431)
Population: Dest -0.0402 -0.0700
(0.0446) (0.0469)
Constant 5.896** 5.864** 6.611** 5.886** 5.846** 6.892**
(0.0031) (0.0024) (0.262) (0.0034) (0.0026) (0.277)
Market-Carrier FE No Yes Yes No Yes Yes
Quarter FE No Yes Yes No Yes Yes
Controls No No Yes No No Yes
No. of Markets 604 604 604 516 516 516
No. of Merging Carriers 1,008 1,008 1,008 920 920 920
No. of Non-merging Carriers 2,265 2,265 2,265 1,907 1,907 1,907
Observations 19,638 19,638 19,638 16,962 16,962 16,962
Adjusted R
2
0.138 0.286 0.288 0.144 0.293 0.295
a
The dependent variable is the natural log of passenger weighted average ticket fare for the ticketing carrier on the
market.
b
Overlap Definition A is that an overlap markets is defined as a directional airport pair where both of the merged airlines
operated at least one quarter prior to the merger, and at least one of them operated through pre and post merger periods.
Overlap Definition B is that an overlap market is defined as a directional airport pair where both of the merged airlines
operated during the sample pre-merger period, and at least one of them operated through pre and post merger periods.
c
Standard errors are in parentheses and clustered at the market-airline level.
d
** : 1% significance, * : 5% significance,y: 10% significance.
37
Table 1.10: The Price Effects Results (Unbalanced Panel)
Overlap Definition A Overlap Definition B
OLS FE FE OLS FE FE
(1) (2) (3) (4) (5) (6)
WN/FL -0.141** -0.137**
(0.0057) (0.00565)
Post 0.157** 0.157**
(0.0043) (0.0047)
WN/FL Post 0.0578** 0.0595** 0.0596** 0.0546** 0.0595** 0.0596**
(0.0078) (0.0057) (0.0057) (0.0077) (0.0059) (0.0059)
Unemployment: Org -0.0117** -0.0114**
(0.0027) (0.0028)
Unemployment: Dest -0.0104** -0.00696*
(0.0027) (0.0029)
Population: Org -0.0695 -0.115*
(0.0468) (0.0482)
Population: Dest -0.0717 -0.113*
(0.0464) (0.0489)
Constant 5.872** 5.851** 6.648** 5.862** 5.836** 6.976**
(0.0029) (0.0026) (0.276) (0.0032) (0.0027) (0.292)
Market-Carrier FE No Yes Yes Yes Yes Yes
Quarter FE No Yes Yes Yes Yes Yes
Controls No No Yes No No Yes
No. of Markets 604 604 604 516 516 516
No. of Merging Carriers 1,208 1,208 1,208 1,032 1,032 1,032
No. of Non-merging Carriers 3,843 3,843 3,843 3,262 3,262 3,262
Observations 24,244 24,244 24,244 20,776 20,776 20,776
Adjusted R
2
0.116 0.223 0.226 0.144 0.293 0.295
a
The dependent variable is the natural log of passenger weighted average ticket fare for the ticketing carrier on the
market.
b
Overlap Definition A is that an overlap markets is defined as a directional airport pair where both of the merged
airlines operated at least one quarter prior to the merger, and at least one of them operated through pre and post
merger periods. Overlap Definition B is that an overlap market is defined as a directional airport pair where both of
the merged airlines operated during the sample pre-merger period, and at least one of them operated through pre and
post merger periods.
c
Standard errors are in parentheses and clustered at the market-airline level.
d
** : 1% significance, * : 5% significance,y: 10% significance.
38
Table 1.11: The Quality Effects Results (Balanced Panel)
Overlap Definition A Overlap Definition B
OLS FE FE OLS FE FE
(1) (2) (3) (4) (5) (6)
WN/FL -0.160 -0.383*
(0.112) (0.179)
Post 0.0191 0.109
(0.148) (0.219)
WN/FL Post -0.0972 -0.0972 -0.0925 -0.231 -0.231** -0.325*
(0.159) (0.0642) (0.0568) (0.229) (0.0838) (0.127)
Unemployment: Org -0.0385 -0.0906
(0.0423) (0.0688)
Unemployment: Dest -0.0377 -0.0895
(0.0424) (0.0689)
Population: Org -0.504 -0.427
(0.714) (0.820)
Population: Dest -0.566 -0.501
(0.711) (0.815)
Constant 5.544** 5.357** 9.696** 5.862** 5.836** 10.16**
(0.106) (0.0228) (1.552) (0.00318) (0.00271) (1.943)
Market-Carrier FE No Yes Yes Yes Yes Yes
Quarter FE No Yes Yes Yes Yes Yes
Controls No No Yes No No Yes
No. of Markets 62 62 62 42 42 42
No. of Merging Carriers 96 96 96 78 78 78
No. of Non-merging Carriers 14 14 14 6 6 6
Observations 660 660 660 504 504 504
Adjusted R
2
0.008 0.014 0.026 0.029 0.032 0.057
a
The dependent variable is the natural log of quarterly flight frequency of the airline on the market.
b
Overlap Definition A is that an overlap markets is defined as a directional airport pair where both of the merged
airlines operated at least one quarter prior to the merger, and at least one of them operated through pre and post
merger periods. Overlap Definition B is that an overlap market is defined as a directional airport pair where both
of the merged airlines operated during the sample pre-merger period, and at least one of them operated through
pre and post merger periods.
c
Standard errors are in parentheses and clustered at the market-airline level.
d
** : 1% significance, * : 5% significance,y: 10% significance.
39
Table 1.12: The Quality Effects Results (Unbalanced Panel)
Overlap Definition A Overlap Definition B
OLS FE FE OLS FE FE
(1) (2) (3) (4) (5) (6)
WN/FL 0.0991 0.0461
(0.196) (0.302)
Post -0.638* -1.289**
(0.284) (0.425)
WN/FL Post 0.678* -0.151* -0.147* 1.164** -0.323** -0.422**
(0.295) (0.0742) (0.0667) (0.432) (0.0960) (0.142)
Unemployment: Org -0.0296 -0.0945
(0.0443) (0.0746)
Unemployment: Dest -0.0295 -0.0934
(0.0444) (0.0746)
Population: Org -0.527 -0.683
(0.908) (1.075)
Population: Dest -0.586 -0.754
(0.904) (1.071)
Constant 4.994** 5.061** 9.449** 5.304** 5.193** 11.62**
(0.189) (0.0271) (2.186) (0.299) (0.0339) (3.226)
Market-Carrier FE No Yes Yes Yes Yes Yes
Quarter FE No Yes Yes Yes Yes Yes
Controls No No Yes No No Yes
No. of Markets 62 62 62 42 42 42
No. of Merging Carriers 124 124 124 84 84 84
No. of Non-merging Carriers 38 38 38 21 21 21
Observations 794 794 794 557 557 557
Adjusted R
2
0.032 0.032 0.034 0.100 0.056 0.071
a
The dependent variable is the natural log of quarterly flight frequency of the airline on the market.
b
Overlap Definition A is that an overlap markets is defined as a directional airport pair where both of the
merged airlines operated at least one quarter prior to the merger, and at least one of them operated through
pre and post merger periods. Overlap Definition B is that an overlap market is defined as a directional airport
pair where both of the merged airlines operated during the sample pre-merger period, and at least one of them
operated through pre and post merger periods.
c
Standard errors are in parentheses and clustered at the market-airline level.
d
** : 1% significance, * : 5% significance,y: 10% significance.
40
Table 1.13: The Capacity Effects Results
Overlap Definition A Overlap Definition B
(1) (2) (3) (4)
Balanced Panel
WN/FL Post -0.144* -0.138* -0.257** -0.347**
(0.0659) (0.0575) (0.0901) (0.130)
No. of Markets 62 42
No. of Merging Carriers 96 78
No. of Non-merging Carriers 14 6
Observations 660 504
Adjusted R
2
0.020 0.030 0.039 0.061
Unbalanced Panel
WN/FL Post -0.194* -0.194* -0.346** -0.442**
(0.0753) (0.0675) (0.101) (0.143)
No. of Markets 62 42
No. of Merging Carriers 124 84
No. of Non-merging Carriers 38 21
Observations 794 556
Adjusted R
2
0.033 0.035 0.062 0.075
Market-Carrier FE Yes Yes Yes Yes
Quarter FE Yes Yes Yes Yes
Controls No Yes No Yes
a
Only the coefficients on WN/FL Post are reported.
b
The dependent variable is the natural log of quarterly number of seats of the airline on
the market.
c
Overlap Definition A is that an overlap markets is defined as a directional airport pair
where both of the merged airlines operated at least one quarter prior to the merger, and
at least one of them operated through pre and post merger periods. Overlap Definition B
is that an overlap market is defined as a directional airport pair where both of the merged
airlines operated during the sample pre-merger period, and at least one of them operated
through pre and post merger periods.
d
Standard errors are in parentheses and clustered at the market-airline level.
e
** : 1% significance, * : 5% significance,y: 10% significance.
41
Table 1.14: The Price Effects for Nonstop Flights
Overlap Definition A Overlap Definition B
(1) (2) (3) (4)
lnFF lnFare lnFF lnFare
Balanced Panel
WN/FL Post -0.0925 0.0783** -0.325* 0.0678
y
(0.0568) (0.0217) (0.127) (0.0348)
No. of Markets 62 62 42 42
No. of Merging Carriers 96 94 78 76
No. of Non-merging Carriers 14 14 6 6
Observations 660 648 504 492
Adjusted R
2
0.026 0.702 0.057 0.736
Unbalanced Panel
WN/FL Post -0.147* 0.0768** -0.422** 0.0781**
(0.0667) (0.0215) (0.142) (0.0335)
No. of Markets 62 62 42 42
No. of Merging Carriers 124 123 84 84
No. of Non-merging Carriers 38 33 21 18
Observations 794 777 557 546
Adjusted R
2
0.034 0.662 0.071 0.710
a
Only the coefficients on WN/FL Post are reported.
b
Market-Carrier FE, Quarter FE, and Controls are included in all specifications.
c
The dependent variable in column (1) and (3) is the natural log of quarterly flight
frequencies of the airline on the market, and the dependent variables in column (2) and
(4) is the natural log of passenger weighted average ticket fare of the airline on the
market.
d
Overlap Definition A is that an overlap markets is defined as a directional airport pair
where both of the merged airlines operated at least one quarter prior to the merger, and
at least one of them operated through pre and post merger periods. Overlap Definition B
is that an overlap market is defined as a directional airport pair where both of the merged
airlines operated during the sample pre-merger period, and at least one of them operated
through pre and post merger periods.
e
Standard errors are in parentheses and clustered at the market-airline level.
f
** : 1% significance, * : 5% significance,y: 10% significance.
42
Table 1.15: Parallel Trends Assumption Tests: Price Effects
Overlap Definition A Overlap Definition B
(1) (2) (3) (4)
FE Event Study FE Event Study
WN/FL Post 0.0547** 0.0565**
(0.00540) (0.00575)
Pre -3 0.00840 0.00768
(0.0085) (0.0092)
Pre -2 0.00639 0.00291
(0.0062) (0.0067)
Post +0 0.0631** 0.0639**
(0.0083) (0.0087)
Post +1 0.0709** 0.0713**
(0.0072) (0.0077)
Post +2 0.0449** 0.0448**
(0.0068) (0.0072)
No. of Markets 604 516
No. of Merging Carriers 1,008 920
No. of Non-merging Carriers 2,265 1,907
Observations 19,638 16,962
Adjusted R
2
0.288 0.288 0.295 0.296
a
Market-Carrier fixed effects, Time fixed effects, and Controls are included in all specifications.
The controls are not reported.
b
Overlap Definition A is that an overlap market is defined as a directional airport pair where
both of the merged airlines operated at least one quarter prior to the merger, and at least one of
them operated through pre- and post-merger periods. Overlap Definition B is that an overlap
market is defined as a directional airport pair where both of the merged airlines operated during
the sample pre-merger period, and at least one of them operated through pre- and post-merger
periods.
c
Standard errors are in parentheses and clustered at the market-airline level.
d
** : 1% significance, * : 5% significance,y: 10% significance.
43
Table 1.16: Parallel Trends Assumption Tests: Quality Effects
Overlap Definition A Overlap Definition B
(1) (2) (3) (4)
FE Event Study FE Event Study
WN/FL Post -0.0925 -0.325*
(0.0568) (0.127)
Pre -3 -0.0221 -0.0601
(0.110) (0.160)
Pre -2 0.0342 0.0266
(0.0375) (0.0513)
Post +0 0.00873 -0.270
(0.118) (0.174)
Post +1 -0.134 -0.383*
(0.0991) (0.189)
Post +2 -0.139
y
-0.356*
(0.0800) (0.153)
No. of Markets 62 42
No. of Merging Carriers 96 78
No. of Non-merging Carriers 14 6
Observations 660 504
Adjusted R
2
0.026 0.023 0.057 0.050
a
Market-Carrier fixed effects, Time fixed effects, and Controls are included in all specifica-
tions. The controls are not reported.
b
Overlap Definition A is that an overlap markets is defined as a directional airport pair
where both of the merged airlines operated at least one quarter prior to the merger, and at
least one of them operated through pre and post merger periods. Overlap Definition B is that
an overlap market is defined as a directional airport pair where both of the merged airlines
operated during the sample pre-merger period, and at least one of them operated through pre
and post merger periods.
c
Standard errors are in parentheses and clustered at the market-airline level.
d
** : 1% significance, * : 5% significance,y: 10% significance.
44
Figure 1.4: Price Effects: Overlap Definition A
Figure 1.5: Price Effects: Overlap Definition B
45
Figure 1.6: Quality Effects: Overlap Definition A
Figure 1.7: Quality Effects: Overlap Definition B
46
Chapter 2
Effects of Codeshare Exit:
Evidence from the Merger of Alaska and Virgin America
1
2.1 Introduction
In 2016, Alaska Airlines (AS) acquired Virgin America (VX). This merger between a fringe
legacy carrier and a low-cost carrier (LCC) resulted in the birth of a strengthened firm capable
of competing with other major airlines in the U.S. (Figure 2.1). The merger between Alaska
and Virgin has been unique in that the Department of Justice (DOJ) required the modification of
Alaska’s codesahre alliance with American Airlines (AA). Most remedies in recent airline mergers
have been enforced with the divestiture of gates or slots to the competitors, put in place as structural
remedies. However, the remedies involved in this merger were conduct remedies, or behavioral
remedies.
2
1
This is joint work, coauthored with Rihyun Park.
2
In merger cases for which the U.S. federal antitrust enforcement agencies—the Antitrust Division of the
United States Department of Justice (DOJ) and the Federal Trade Commission (FTC)—raise concerns about
competitive harm, appropriate remedies can be issued to protect competition in the market along with the merger
process. There are two forms of merger remedies: structural and behavioral. Structural remedies include the
sale or licensing of firms’ assets. Behavioral remedies include restrictions on the merged firm’s conduct (An-
titrust Division Policy Guide to Merger Remedies, U.S. Department of Justice, Antitrust Division, June 2011:
https://www.justice.gov/atr/page/file/1098656/download (last accessed in March, 2022)). Histor-
ically, the use of behavioral remedies has been discouraged due to the difficulty of implementation, increased
monitoring costs, and risk of circumvention by merged firms (Jessica K. Delbaum, John Skinner, ”Merger
remedies in the US: An overview of the leading cases,” 5 March 2020, e-Competitions US Merger remedies,
Art. N 922390). In the case of the merger between US Airways and American Airlines, for example, the
Department of Justice required merging parties to divest slots and gates at key constrained airports across
the country to low-cost carrier airlines in order to enhance system-wide competition in the airline industry,
47
The government authority was greatly concerned about a potential loss of competition in mar-
kets where Virgin America fiercely competed with American Airlines while Alaska Airlines was
dependent on American Airlines, given their codeshare agreements in many markets. The mod-
ification of the codeshare agreement between the merging parties and American was designed to
ensure that Alaska would have additional incentives to vigorously compete with American, just as
Virgin had done before 2016, and would directly impact the codeshare connection in the network
of Alaska and American Airlines. Since the required behaviors represented exogenous shocks to
Alaska-American (AS-AA) codeshare products, this study investigates whether and how codeshare
discontinuation impacted market prices and consumers.
In order to examine the changes associated with the merger and the impact of codeshare dis-
continuation, this study begins with a difference-in-differences (DID) fare comparison regarding
the merger. In this analysis, we compare the market level prices of different types of markets (types
being defined by the presences of the merging airlines on the markets). The results support neither
pro-competitive nor anti-competitive effects; the market level fares as well as the market ticketing
level fares exhibit no significant changes following the merger. In the next section, we use two-
way fixed DID regressions to estimate codeshare exit effects. In this analysis, we compare price
changes in affected markets with those of unaffected markets. The findings show that the code-
share cessation was associated with relative price decreases, especially on the tickets marketed
by the affected airlines (Alaska and American). To examine if these effects are associated with
decreased demand, we estimate the demand of air travel on the relevant markets. Using a set of
relevant dummy variables and their interactions, we conclude that the discontinuation of AS-AA
codeshare products did decrease demand for products on the codeshare exit markets, even though
the deceased demand was not specific to AS-AA products.
The remainder of this paper is organized as follows. Section 2.2 presents relevant prior liter-
ature on merger analysis and the competitive effects of airline codeshare alliances. Section 2.3
resulting in more choices and more competitive airfares for consumers (https://www.justice.gov/opa/pr/
justice-department-requires-us-airways-and-american-airlines-divest-facilities-seven-key).
48
provides a working definition of codeshare products and briefly discusses AS-AA codeshare cir-
cumstances. Section 2.4 describes the background of the Alaska/Virgin (AS/VX) merger and the
related remedies enforced by the authorities. Section 2.5 describes our data sources and the con-
struction of our sample. Section 2.6 illustrates how the price changes were compared in terms of
the merger, as well as the results. Section 2.7 describes the empirical methodology for a codeshare
discontinuation analysis and presents the results. Section 2.8 explores the structural estimation on
consumer demand to explain the connection between codeshare and consumer utility. Finally, the
last section of this paper presents a discussion and conclusion.
2.2 Literature Review
This section provides a brief overview of previous studies focusing on merger effects and code-
share agreements in the airline industry. Mergers and merger behavior within the U.S. airline in-
dustry have been extensively studied. There have been dynamic changes in airline competition
in recent years, resulting in many notably large mergers. This has enabled investigation into the
competitive effects of each merger case and related government enforcement. Older studies in
this field were generally concerned with the anti-competitive effects of mergers, with many argu-
ing that the loss of competition increased fare prices and brought harm to consumers (Borenstein
1990; Werden et al. 1991; Kim and Singal 1993; Peters 2006; Kwoka and Shumilkina 2010).
More recent studies, however, have opened up new possibilities for understanding the different
effects of mergers. Luo (2014) found that the examined legacy airline merger did not significantly
impact fares on the affected routes. She concluded that the price impact of low-cost carrier com-
petition was much larger than that of legacy carrier competition. Das (2019) studied the effects
of mergers on fares and product quality. His difference-in-differences analysis showed that the
merger between American Airlines (AA) and US Airways (US) actually decreased market fares in
large markets, although it did not affect the frequency of flights or the number of seats. Carlton
49
et al. (2019) studied the effects of three recent legacy mergers on fares and outputs.
3
The paper
confirmed that all three mergers were pro-competitive, with no adverse effect on nominal fares and
a significant increase in passenger traffic.
In addition to mergers, the codeshare agreements within the airline industry have also been
widely studied. Although far more studies have been conducted into international codeshare agree-
ments (e.g., Brueckner and Whalen, 2000; Park and Zhang, 2000; Brueckner, 2003; Whalen, 2007),
we focus here on studies of domestic codeshare agreements. Ito and Lee (2005) and Ito and Lee
(2007) documented definitions for different types of air travel products in terms ticketing and op-
erating carriers.
4
After addressing certain stylized facts about codesharing practices in the U.S.
airline industry, their study showed that virtual codesharing tickets are priced lower than those
ticketed and operated by a single carrier.
5
Gayle (2008) addressed the U.S. Department of Transportation’s (DOT) main concern that the
codeshare alliance among three major airlines—Delta, Continental, and Northwest—could facili-
tate collusive behaviors among the partners, especially on their overlapping routes. His research
estimated the actual effects of the codeshare alliance on price and traffic (number of passengers)
on the partners’ overlapping routes. He compared the changes in average price or total traffic of
city pairs for which the partners codeshare (alliance city pairs) with those of city pairs for which
at least one of the three provides service but for which there is no code sharing between any of the
three carriers (non-alliance city pairs). The aggregated analysis showed that, overall, the alliance
is associated with both price and traffic increases. He also investigated whether different types
of codeshare products had different effects. The findings showed that traditional code sharing is
3
The analyzed mergers are the merger between Delta (DL) and Northwest (NW) in 2008, between United (UA)
and Continental (CO) in 2010, and between American (AA) and US Airways (US) in 2013.
4
A ticketing carrier is the airline that sells the ticket under its airline code, whereas an operating carrier is the
airline whose aircraft is used to operate the flight.
5
The definition of a virtual codeshare in Ito and Lee (2005) and Ito and Lee (2007) is slightly different from
the one used in this paper. In those studies, the researchers divide virtual codeshare into two groups: a semi-virtual
codeshare and a fully virtual codeshare. A semi-virtual codeshare product is an itinerary operated by a single carrier
but partly ticketed by another airline. A fully virtual codeshare product is defined as an itinerary operated by a single
carrier and ticketed by another single carrier. A virtual codeshare product as defined in this study would be considered
a fully virtual codeshare in Ito and Lee (2005) and Ito and Lee (2007), and we do not consider semi-virtual codeshares
because they involve multiple ticketing carriers.
50
associated with price decreases and traffic increases, while virtual code sharing is associated with
both price and traffic increases. From these findings, he concluded that the examined alliance is
not associated with collusive behavior within the partners’ overlapping markets.
Recent research has also developed structural models of code sharing between U.S. domestic
airlines. Gayle (2013) focused on traditional codesharing, using a vertical relation approach to test
the existence of double marginalization in pricing of codeshare products.
6
Gayle (2013) used a
random coefficients logit demand model and structured the supply side of the model for a double
marginalization with two kinds of margins in the prices of codeshare products: an upstream margin
of the operating carrier and a downstream margin of the marketing carrier. His work found that
the upstream margin persists if the upstream operating carrier simultaneously offers competing
online products in the same market. This is because the operating carrier optimally chooses not
to eliminate its upstream margin of codeshare products to reduce the intensity of downstream
competition for its own product.
Two other papers have adopted structural models for studying the effects of virtual codeshar-
ing. Gayle and Brown (2014) investigated whether virtual codesharing is associated with collusive
pricing behavior. They studied the codeshare contracts between the Delta, Continental, and North-
west airlines and found that, in their overlapping market, the average price increased while the
average number of passengers decreased. A demand estimation, however, suggested that the code-
share alliance actually exhibited a demand increasing effect, given that the alliance creates new
opportunities for passengers to accumulate and redeem rewards points across airlines. Moreover,
their non-nested statistical test on price setting behavior assumptions suggested no statistical evi-
dence of collusive behavior between partners. Shen (2017) studied two codeshare alliances active
in 2003. Her work established a pricing equation in which the ticketing carrier keeps a share of
6
Codesharing can be seen as a form of vertical contracting between ticketing and operating carriers (Chen and
Gayle, 2007). The operating carrier is equivalent to an upstream supplier that provides an essential input (trip seg-
ments) to the downstream ticketing carrier, who then combines it with other inputs (complementary trip segments) in
order to provide the final products to consumers. The ticketing carrier sets the final price for the entire trip ticket and
compensates the operating carrier for its operating services on a segment. Therefore, the codesharing may have the
pro-competitive effect of eliminating double marginalization, since two carriers can jointly price the product, resulting
in the elimination of double markups.
51
the profits as a commission fee while the operating carrier acquires the remainder of the profits.
She found that codesharing reduces marginal costs for the airlines, and thus enables them to price
codeshare products at lower rates. More importantly, she found that, if codesharing creates new
products in the market, the demand increases, and consumers are given a larger surplus.
Our analysis differs with those of previous codeshare studies in two ways. First, while prior re-
search has focused on codeshare formations, this paper is unique in presenting a codeshare analysis
with a codeshare discontinuation case. Second, we can regard the codeshare exit change as exoge-
nous, with the justifiable rationale behind this assumption being that the decision was formally
ordered by the Department of Justice (DOJ).
2.3 Codeshare Alliance
In this section, we present working definitions for the different types of codeshare products and
describe the specific codeshare agreement between Alaska and American. We then document the
merger and remedy background in the following section.
2.3.1 Definition
A codeshare alliance allows one carrier (“a ticketing/marketing carrier”) to sell and market the
tickets of its partners’ flights using the code of the ticketing/marketing carrier. The partner carriers,
those that actually operate the flight and carry the passengers, are called “operating carriers”. For
example, suppose that American Airlines is selling tickets for a flight, AA999, that is actually
operated by its codeshare alliance partner, Alaska Airlines. In this case, even though a passenger
buys a ticket from American Airlines (AA999), that passenger might actually take a flight operated
by the codeshare partner, Alaska Airlines.
The literature on airline codeshare alliances has defined two types of codeshare itineraries: (1)
traditional codeshare; and (2) virtual codeshare. The different types of products are illustrated in
Figure 2.2. All three tickets represent an example of three types of products marketed by American
52
(AA) on the market between Los Angeles (LAX) and Boston (BOS). In this figure, the red airplanes
represent flights operated by American, and the blue airplanes represent flight operated by Alaska
(AS).
Traditional codeshare itineraries offer interline operating services from codeshare partners
while the flights are marketed solely by a ticketing carrier who is one of those two distinct op-
erating carriers.
7
This type of product is illustrated as the second ticket in Figure 2.2.
In contrast, virtual codeshare products offer an online operating service for a partner carrier,
meaning that passengers on these itineraries remain on a single operating carrier’s plane(s) for the
entire trip. The ticket for the entire flight is marketed by a ticketing carrier, which is distinct from
the operating carrier in these itineraries. Here, the ticketing carrier is always distinct from the
operating carrier. This type of codeshare itinerary is illustrated as the third ticket in Figure 2.2.
Thus, a key distinction between traditional and virtual codeshares is that the former requires the
passenger to travel on different operating carriers’ planes (interline air travel)—one of them is the
ticketing carrier—on a multi-segment route, while the latter does not involve interline air travel
even when passengers change planes on a multi-segment route.
The flight itineraries marketed and operated by one airline are known as “pure online” products.
In this case, the marketing carrier and the operating carrier are always identical. The first ticket in
Figure 2.2 illustrates an example of a pure online product.
2.3.2 Alaska (AS) and American (AA) Codeshare Agreement
In 2016, Alaska Airlines maintained codeshare partnerships with two other legacy airlines:
American Airlines (AA) and Delta Airlines (DL). The codeshare agreement between American
and Alaska was implemented in 1999. Since then, the two airlines repeatedly expanded their ar-
rangement, increasing the number of routes on which each partner could sell on behalf of the other.
As a result of the merger between American (AA) and US Airways (US) in 2013, Alaska could
7
An interline service describes when segments of the itineraries are operated by distinct airlines, requiring pas-
sengers to change airline brands when they change planes at some point in their itinerary. An online service describes
when passengers travel on a single airline brand through the entire trip.
53
sell more flights on American’s significantly expanded network in 2015. In April 2016, just prior
to the AS/VX merger, American and Alaska once again expanded their codeshare agreement. This
expansion enabled Alaska to market American flights on over 250 routes, and allowed American
to market Alaska flights on approximately 80 routes. This enabled Alaska and American to sell
each other’s flights on certain overlap routes where both partners already offered competing non-
stop services. This situation raised serious concerns regarding potential loss of competition with
the Department of Justice (DOJ). Existing literature has widely discussed how codeshare partners
“may choose to compete less vigorously than they otherwise would” (Ito and Lee, 2007).
Figures 2.3 and 2.4 show in how many routes AS-AA traditional or virtual codeshare products
were marketed, respectively. For these figures, AS/AA indicates codeshare products marketed by
Alaska (AS) and operated by American (AA), while AA/AS indicates codeshare products ticketed
by AA and operated by AS. The vertical dashed grey lines indicate 2016Q4, when the final judg-
ment was proposed, and the solid grey lines indicate 2017Q2, when the final judgment was revised
and finalized. The green, blue, and purple lines are drawn cumulatively, so the area under the
square-connected purple line (AS-AA cs) show the number of markets in which both AS/AA and
AA/AS codeshare products were sold. Similarly, the area between the square-connected purple
line and the triangle-connected blue line (AS/AA cs only) show the number of markets in which
only AS/AA codeshare products were marketed, while the area between the triangle-connected
blue line and the circle-connected green line (AA/AS cs only) show the number of markets in
which only AA/AS codeshare products were marketed. Figures 2.5 and 2.6 show how many pas-
sengers used AS-AA traditional or virtual codeshare products, respectively. The bars are drawn
cumulatively, and the trends are similar to those of the number of markets.
8
8
The vertical red lines indicate times when the final judgment was first proposed (dashed) and revised and finalized
(solid), similar to Figure 2.3 and 2.4.
54
2.4 Alaska Airlines’ acquisition of Virgin America
In this section, we briefly describe Alaska Airlines (AS)’ acquisition of Virgin American (VX)
and the remedies associated with that acquisition.
9
2.4.1 Merger Background
In April 2016, Alaska Airlines announced its plans to acquire Virgin America, with the ac-
quisition approved by Virgin America’s shareholders in July 2016. Although a lawsuit was filed
against Alaska Airlines by private plaintiffs to block the merger between these two carriers, the
Alaska Air Group was able to settle the lawsuit and the merger was approved by the Department
of Justice (DOJ) in December 2016. About a week after the approval, the merger was officially
closed. Even though no asset divestitures were required by the government authority, the merger
was approved conditional on modifications of the codeshare agreement between Alaska (AS) and
American Airlines (AA). These modifications are described in more detail in the following sub-
section. The merged Alaska was granted a single operating certificate from the Federal Aviation
Administration (FAA) in January 2018. The two airlines gradually integrated their various systems
until the final flight of Virgin America operated on April 24, 2018.
Although the merger was ultimately approved and consummated, there were concerns regard-
ing this merger. Many of the concerns focused on potential reductions in competition. Since the
two airlines’ routes were concentrated in the western U.S., an elimination of one airline might
significantly reduce competition in the area. The airlines even shared the same hub airports, Los
Angeles (LAX) and San Francisco (SFO). As perceived based on the lawsuit and the DOJ’s rem-
edy, the merger was expected to reduce competition and to result in general price increases. An-
other concern focused on the processes of integration. Operational inefficiencies represent one of
the problems associated with a merger, and the integration between Alaska and Virgin American
9
In this study, we do not distinguish an acquisition from a merger because differences between the two are not
relevant to the purpose of the study.
55
might not have been an exception. Despite these concerns, the two airlines made concerted efforts
toward integrating, and eventually stabilized as a combined airline.
2.4.2 Remedies
In order to resolve the likely competitive harm associated with the merger, the U.S. district court
for the District of Columbia constructed conduct remedies upon its approval of the merger between
Alaska and Virgin America. The authorities were primarily concerned with the loss of a competitor
against American Airlines, the world’s largest airline, as a consequence of this acquisition.
Virgin America had, until the merger, aggressively competed with American (AA) on their
overlap routes in ways that forced American to respond with lower fares and better services. In
fact, markets where Virgin America and American overlapped had expanded significantly in terms
of market total passengers compared to other American and Alaska-American overlap markets.
Moreover, in markets where both American and Virgin operated, they had similar passenger shares,
indicating that Virgin was indeed a strong competitor against American.
Alaska’s codeshare agreement with American, meanwhile, created an incentive for Alaska to
avoid aggressive head-to-head competition in order to preserve its relationship with American.
With constant expansion of the codeshare agreement between the two airlines, they operated code-
share products even on overlap routes where both already offered competing pure online products,
which raised concerns regarding a loss of competition.
10
Taking these issues into account, the DOJ required Alaska to significantly scale back its code-
share agreement with American, identifying four main markets where Alaska and American should
not operate codeshare products.
11
First, the DOJ prohibited Alaska and American from offering
their codeshare products on routes where Virgin America and American both offered competing
10
More details are provided in Appendix A.2.
11
The final judgement can be found at https://www.justice.gov/atr/case-document/file/
1039436/download. Other related documents can be found at https://www.justice.gov/opa/pr/
justice-department-requires-alaska-airlines-significantly-scale-back-codeshare-agreement
andhttps://www.justice.gov/atr/case/us-v-alaska-air-group-inc-and-virgin-america-inc; U.S.
V . ALASKA AIR GROUP, INC., AND VIRGIN AMERICA INC. (last accessed in March, 2022).
56
nonstop services during the pre-merger period.
12
Second, the DOJ prohibited Alaska and Amer-
ican from code sharing on Alaska-American (AS-AA) nonstop overlap routes.
13
Third, the DOJ
prevented either Alaska or American from marketing each other’s flights on routes that included
their respective key airports.
14
For example, Alaska is prohibited from selling American code-
shared flights on their routes from Seattle, one of Alaska’s key airports. Lastly, the DOJ prohibited
either carrier from codesharing on routes between Los Angeles (LAX) and either American or
Alaska key airports.
In the second quarter of 2016 in the DB1B data, 96 distinct routes can be identified as affected
by these codeshare modification remedies. The four categories of remedy routes mentioned above
were not mutually exclusive. To better understand the affected routes, we distinguish all the remedy
markets into two sets depending on the existence of codeshare products between Alaska (AS) and
American (AA). After exploring the data (Table 2.1), we found a stylized fact that, in the first
category (codeshare regulations on the VX-AA overlap routes), there had been little appearance
of codeshare products between Alaska and American. Inasmuch as the other remedy markets
were directly related to the network of Alaska and American airlines, we found that the effects of
codeshare elimination were relatively stronger in the other three categories of the remedy markets.
More details are provided in Appendix A.1.
12
The DOJ specified 42 VX-AA U.S. domestic overlap routes as of December 6, 2016 in its final judgement, and
the list is in appendix Table A1.
13
The DOJ described 62 AS-AA U.S. domestic overlap routes as of December 6, 2016 in its final judgement for
illustrative purposes, and the list is in appendix Table A2. For this definition of overlap routes only, the authority
identified the overlap routes not on the basis of airport, but rather on the basis of city. Thus, airports with the same
City Market ID, as identified by the U.S. Department of Transportation in the Airline Origin and Destination Survey
(DB1B), are regarded as serving the same city, with the following airports excepted: (1) Los Angeles International
Airport (LAX), and (2) Norman Y . Mineta San Jose International Airport (SJC). This list of Alaska/American overlap
routes may be subject to change as both airlines adjust their respective schedules.
14
The DOJ defined four Key Alaska Airports as follows: (1) Portland International Airport (PDX); (2) Seattle-
Tacoma International Airport (SEA); (3) San Francisco International Airport (SFO); and (4) Ted Stevens Anchorage
International Airport (ANC). They also identified twelve Key American Airports as follows: (1) Charlotte Douglas
International Airport (CLT); (2) Chicago Midway International Airport (MDW); (3) Chicago O’Hare International
Airport (ORD); (4) Dallas/Fort Worth International Airport (DFW); (5) Dallas Love Field (DAL); (6) Fort Lauderdale-
Hollywood International Airport (FLL); (7) John F. Kennedy International Airport (JFK); (8) Miami International
Airport (MIA); (9) New York LaGuardia Airport (LGA); (10) Philadelphia International Airport (PHL); (11) Phoenix
Sky Harbor International Airport (PHX); and (12) Washington Reagan National Airport (DCA).
57
2.5 Data
2.5.1 Data Sources
The main data used in this paper are the Airline Origin and Destination Survey (DB1B) dataset
from the Bureau of Transportation. This set represents a quarterly 10% sample of domestic air
travel tickets reported by reporting carriers.
15
The data contain detailed information on each
itinerary, including itinerary fares, number of passengers who purchased tickets at the given price,
origin/destination cities, and origin/connecting/destination airports. The data also include infor-
mation on ticketing (marketing), operating, and reporting carriers.
We supplement the DB1B data with other relevant data samples to construct control variables.
Population data are from the U.S. Census Bureau.
16
Although the population variable might not
have changed dramatically over time, it is included to capture market size as well as potential
economies of traffic density (Brueckner and Spiller, 1994). In order to capture changes in economic
conditions, unemployment rate data are included from the U.S. Bureau of Labor Statistics (BLS).
Data regarding one of the main cost shifters, jet fuel price, are from the U.S. Energy Information
Administration.
2.5.2 Sample Construction
A basic observation unit—before collapsing—is defined as a roundtrip itinerary on a market at
a particular time (quarter). A market is defined as a directional airport-pair of origin and destination
15
For analytical purposes, airlines can be categorized into three groups: (1) ticketing carriers, (2) operating carriers,
and (3) reporting carriers. Usually, a reporting carrier is an airline that operates the first segment of a trip. For example,
if an itinerary is marketed by American Airlines but operated by Alaska Airlines, it is likely that the reporting carrier
is Alaska Airlines.
16
Annual Estimates of the Residence Population: April 1, 2010 to July 1, 2018: U.S. Census Bureau, Population
Division. This report produces estimates of the population for the United States, individual states, metropolitan and
micropolitan statistical areas, counties, cities, and towns.
58
airports.
17
For example, a trip from Los Angeles (LAX) to Dallas (DAL) is treated as a different
market from a trip from DAL to LAX.
Several restrictions have been applied in order to construct the analysis sample. First, for every
analysis in this study, any itinerary that is either bulk fare or suspected to have incredible fare is not
included. Itineraries which include ground transportation or roundtrip itineraries with more than
two trip breaks are also dropped from the sample. To exclude anomalies, tickets with too low or
too high fares are not included.
18
Any tickets with more than two coupons are dropped. In other
words, the tickets included in the sample can have at most one stop per one-way trip. Itineraries
with multiple ticketing carriers or foreign carriers are not included. The sample includes only pure
online products and traditional codeshare products. This decision was made upon examination of
two types of codeshare products between Alaska and American. We confirmed that the traditional
codeshare was the prominent relationship between the two codeshare partners at the time of the
merger and remedy decision. As confirmed from data presented in Figures 2.4 and 2.6, almost no
virtural codeshare products existed between Alaska and American starting from 2015.
Although many of Alaska Airlines (AS)’ routes are from or to airports in the state of Alaska
(AK), airports from outside the contiguous U.S. are not included in this study, mainly because
routes in the contiguous U.S. are likely to have different features than those in non-contiguous
states.
19
Since all airports in the sample are located within the contiguous U.S., any particular
characteristics from non-contiguous states are irrelevant to this study.
Airlines included in this study are mainline carriers such as American Airlines (AA), Alaska
Airlines (AS), and Virgin America (VX).
20
Although some other airlines are present on the relevant
17
In empirical literature regarding the airline industry, two common ways to define a market are to use routes either
between city-pairs or airport-pairs. Each definition can be more suitable than the other depending on the specific
research questions and assumptions of any given study. In this paper, we use airport-pairs, because the majority of the
remedies introduced by the DOJ (i.e., reductions of the codeshare operations between AS and AA in certain markets)
were applied at an airport-pair level. One exception to this is the second category, which specifically targets the
city-pair market between AS and AA overlap. This category can be approximately covered by the airport-pairs.
18
Itineraries with fares lower than $50 or greater than $2,000 are dropped from the sample.
19
For instance, certain types of substitute transport available in the contiguous U.S. might not be viable options in
the case of non-contiguous states.
20
The included legacy airlines are American (AA), Alaska (AS), Delta (DL), Hawaiian (HA), United (UA), while
the included low-cost carriers are JetBlue (B6), Frontier (F9), Allegiant Air (G4), Spirit (NK), Sun Country (SY),
Virgin America (VX), and Southwest (WN).
59
markets, most of them experience little routine traffic. In addition, this restriction was predicted
to have an insignificant impact on our results because the included airlines represented more than
90% of the market shares in most of the sample markets. Moreover, our study recodes affiliated
regional carriers using the codes of the mainline carriers that they serve. In the absence of such
recoding, products with a mainline airline as a ticketing carrier and its affiliated regional carrier
as an operating carrier may be counted as codeshare products, since the operating and ticketing
carriers differ. The main recoding procedure conducted on the sample is clarified in Table 2.2. We
found five regional carriers that had been officially affiliated according to the annual report of the
Regional Airline Association (RAA).
21
2.5.2.1 CodeshareExit Effects Sample
Following application of the conventional restrictions, we collapse our data into market-ticketing
carrier-time level observations. For the fare variable, passenger weighted average prices are cal-
culated. In order to estimate codeshare (CS) discontinuation effects, we focus our analysis on the
relevant AS-AA markets, or markets in which at least one of the AS/AA and AA/AS traditional
codeshare products was operated during the pre-remedy period. For the AS-AA codeshare relevant
markets, two alternative definitions are used. Definition A defines the relevant markets as routes in
which AS-AA traditional codeshare products were operated during all sample pre-remedy periods.
Definition B defines the relevant markets as routes in which AS-AA traditional codeshare products
were operated during at least one pre-remedy period. Definition A is stricter, while Definition B
includes more markets. For both definitions, CS Exit markets are defined as routes for which there
were AS-AA codeshare products in the pre-remedy period but no AS-AA codeshare products in
the post-remedy period.
One caveat regarding these definitions is that these definitions are subject to the measurement
error problem. Because these definitions are based on the 10% sample data and are used to indi-
cate treatment status, results from these definitions might be subject to the infamous misclassified
21
The RAA Annual Reports can be found at https://www.raa.org/content-hub/raa-annual-reports
(last accessed in March, 2022).
60
regressor problem (Mahajan 2006; Lewbel 2007; Nguimkeu et al. 2021). For example, a market
is classified as a CS Exit market based on the presence of AS-AA codeshare products in the pre-
remedy period and the absence of such products in the post-remedy period. However, if there had
been AS-AA codeshare products in the post-remedy period that were not included in the data, then
the market would have to be classified as a CS Keep market. As described in many studies of mis-
classified binary variables, this potential measurement error problem is likely to cause attenuation
bias in estimates. Thus, the results should be interpreted while taking this into account.
Two quarters from the pre-remedy period and two quarters from the post-remedy period are
included in the sample. Quarters 2 and 3 from 2016 are chosen as the pre-remedy period, and
quarters 3 and 4 from 2017 are included as the post-remedy period. The pre-remedy quarters are
from before the DOJ’s final judgement was released, and the post-remedy quarters are from after
the revised version of the final judgement was released.
22
Summary statistics are presented in
Table 2.3.
2.5.2.2 Demand Estimation Sample
To estimate passengers’ air travel demand, we continued the analysis with product-level data. A
product is defined as a combination of a trip itinerary, a ticketing carrier, and an operating partner, if
the product is a codeshare product. Examples of such products include (1) LAX-JFK/AA/AA, (2)
LAX-SEA-JFK/AA/AS, and (3) JFK-ATL-LAX/AA/AA.
23
The first product, LAX-JFK/AA/AA,
is a nonstop product from LAX to JFK, ticketed and operated by American (AA), whereas LAX-
SEA-JFK/AA/AS is a onestop codeshare product from LAX to JFK, ticketed by American, al-
though a part of the trip is operated by Alaska (AS). Even though these two products are different
in number of stops (nonstop vs. onestop) and product types (pure online vs. traditional codeshare),
they are products in the same market, the market of LAX:JFK. The last example is an onestop pure
22
The final judgement specifies that “Beginning sixty calendar days after consummation of the Transaction, De-
fendents shall not directly or indirectly” offer codeshare products on the relevant markets. Since the merger was closed
on December 14, 2016, 2017Q3 was approximately 60 days after the transaction.
23
We instituted a simple rule to represent the product information. We put a flight itinerary, a ticketing carrier, and
an operating carrier in a specific order: Origin airport - (Connecting airport) - Destination airport / Ticketing carrier /
Operating carrier.
61
online product, ticketed and operated by American in the JFK:LAX market. As stated previously,
the LAX:JFK market is different from the market of JFK:LAX. Therefore, this definition of prod-
uct determines the product’s observed characteristics, such as the number of stops, miles flown,
direct distance between origin and destination, and product types.
We continue the demand analysis with the stricter given definition of codeshare discontinua-
tion markets (Definition A in the above subsection). Thus, we have 59 codeshare discontinuation
markets (CS Exit markets) and 49 codeshare continuation markets (CS Keep markets). We used
the same time period to maintain the consistency of the analysis. Summary statistics are presented
in Table 2.4.
2.6 Descriptive Merger Analysis
Prior to investigating the codeshare exit effects, which is the focus of this study, we conducted a
descriptive analysis on how ticket prices were changed before and after the Alaska-Virgin America
(AS-VX) merger. In this section, we conduct market level fare comparison as well as a comparison
of fare changes in the most relevant markets in terms of antitrust issues—routes the merging parties
operated before and after the merger.
2.6.1 Market Level Fare Changes
Similar to the work of Carlton et al. (2019), we focus on market level average prices to com-
pare changes in different types of markets. To compare market level changes, the markets are
categorized into five exclusive groups: Overlap, Either, Exit, Enter, and Neither markets. Over-
lap markets are routes in which both Alaska and Virgin America operated as direct competitors.
This type of market includes most relevant markets in terms of merger effects, since there will be
a direct reduction of airlines in the market, whereas efficiency gain can be realized substantially.
Either markets are routes in which only one of the merging airlines operated during the pre- and
post-merger periods. Exit markets are routes in which at least one of the merging airlines operated
62
prior to the merger but ceased to operate after the merger. Enter markets are routes in which neither
of the merging airlines operated prior to the merger, but in which they began to operate after the
merger. Lastly, Neither markets are routes in which neither of the merging airlines operated before
or after the merger. Because Neither markets are considered to be the least affected by the merger,
subsets of Neither markets are likely to be used as a control group when conducting market level
merger effects analysis. For these different types of markets, market level passenger weighted av-
erage fares are averaged for the pre- and post-merger periods, respectively. These differences are
also compared to the differences of Neither markets as if Neither markets are used as a control
group.
Table 2.5 shows the market level fare comparison results; these results are generally insignifi-
cant.
24
For each of the market types, Column (1) in the table shows the mean of the average market
fares in the pre-merger period; Column (2) shows the mean of the average market fares in the post-
merger period; Column (3) shows the differences between the pre- and post-merger periods; and
Column (4) shows the difference-in-differences that the difference of Neither markets subtracted
from each differences of markets. Even though the average prices dropped in the post-merger pe-
riod, with the exception of Exit markets, these differences are not significant, and the differences
are not significantly different compared to the difference of Neither markets. The results suggest
that the merger affected markets did not experience higher or lower fare changes compared to the
markets considered the least affected by the merger. More importantly, this merger appears not to
be associated with anti-competitive price effects.
25
24
The pre- and post-merger periods used to examine the merger effects might need to be different from those
used in the codeshare effects analysis. When conducting comparisons using alternative time periods, the results are
similar in the sense that none of the affected markets are shown to have experienced significantly different fare changes
compared to Neither markets.
25
The potential misclassified binary regressors problem may have caused biases in these results. In this case, the
insignificant results should be further examined using different methods, which may correct the problem to some
extent.
63
2.6.2 Fare Changes inOverlap andEither markets
Even though market level merger effects were not detected based on the results, it is possible
that the merging airlines’ prices were affected in specific markets. Since both the Overlap and
Either markets are routes in which the merging airlines presented during the pre- and post-merger
periods, we compare pricing changes from the merging airlines to those from the non-merging
airlines on the relevant routes.
The results are reported in Table 2.6. The first two columns of the table show the results
of the Overlap markets, and the last two columns show the results of the Eiether markets. All
specifications include populations and unemployment rates as controls, although the results for
these variables are not reported in the table. Columns (1) and (3) illustrate the OLS results of
estimating the following equation:
lnFare
imt
=b
0
+b
1
AS=V X
im
+b
2
Post
t
+b
3
AS=V X
im
Post
t
+Controls+e
imt
;
(2.1)
where i is an airline, m is a market, and t is a time. AS=V X is a dummy variable indicating
the merging airlines, Post is a dummy variable indicating the post-merger period, and e
imt
is an
idiosyncratic shock at market-ticketing carrier-time level. As shown in Table 2.6, no coefficients
on AS=V XPost are significant. In order words, compared to the fare changes of the non-merging
airlines, the price changes of the merging airlines were neither significantly higher nor significantly
lower on the routes in which the merging airlines competed with each other or in which one of the
merging airlines operated without the other.
Column (2) and (4) represents the results using the market-carrier and time fixed effects instead
of AS=V X and Post dummy variables. The price effects are found to be insignificant in these
specifications as well. Comparing this together with the results from the previous subsection, it is
evident that the merger between Alaska and Virgin America is not associated with significant price
effects.
64
2.7 Codeshare Exit Effects
2.7.1 Empirical Strategy
In order to estimate the AS-AA codeshare exit effects, we use two-way fixed effects difference-
in-differences regressions. This method has been widely used for examining treatment effects in
various studies, as unobserved time and unit characteristics can be controlled. The main issue with
applying this method is how to define a treatment group and a control group. In our analysis, the
treatment group is defined as markets in which AS-AA ceased their codeshare products after the
remedy, whereas the control group is defined as markets in which AS-AA continued to operate
their codeshare products.
One advantage of using the AS-VX merger to estimate codeshare exit effects is that we can
reasonably treat codeshare exit decisions as exogenous. As stated in Gayle and Brown (2014),
codeshare decisions are usually endogenous, since airlines choose on which routes they offer code-
share products with which partners. In such cases, endogeneity in codeshare decisions should be
addressed.
26
Conversely, AS-AA’s codeshare discontinuation decisions in 2016 were mainly due
to the remedy required by the governmental authorities. As a result, we can treat the codeshare exit
decisions as exogenous, and thus the results are less likely to suffer from selection bias.
A ticket price is specified as follows:
lnFare
imt
=b
0
+b
1
CS Exit
im
Post
t
+Controls+g
im
+l
t
+e
imt
;
(2.2)
where i is an airline, m is a market (an airport-pair), and t is a time (quarter). The outcome variable
is lnFare, which is the natural log of the passenger weighted average fare of the airline i on the
market m at time t. CS Exit is a dummy variable that takes a value of 1 if the observation is for
the CS Exit markets, as defined in Section 2.5.2. Post is a dummy variable indicating the post-
merger remedy period. Controls are control variables included to capture the potential effects
26
For example, Gayle and Brown (2014) use the predicted probability variable from the logit model to address the
potential endogeneity problem.
65
of economic conditions. We include the population and unemployment rates of the origin and
destination cities as the controls. For the unemployment rates, the rates are used at the level of
Metropolitan/Micropolitan statistical areas (MSA); if the MSA level rates are not available, state
level rates are used instead. g
im
is market-ticketing carrier fixed effect, andl
t
is time fixed effects.
Lastly,e
imt
is idiosyncratic shock at market-ticketing carrier-time level.
The coefficient of the main interest is the coefficient on CS Exit Post, b
1
. b
1
approximates
the percentage difference between the changes in fares of observations on markets in which Alaska
and American ceased their codeshare operations, compared to the changes in fare of observations
on the markets in which Alaska or American continued to operate their codeshare products. This
represents the average codeshare exit effect on the airline on the CS Exit markets.
2.7.2 Results
The results of this analysis are presented in Tables 2.7 and 2.8. Table 2.7 shows the results
from the Definition A sample. To see whether Alaska and American behaved differently from the
other examined airlines, the last three columns include only Alaska and American observations,
whereas the first three columns include all airlines in the sample. Because airlines enter and exit
markets, and because the DB1B data is a 10% sample, not all airlines are shown in the data for
the entire sample periods. To ensure that the results are robust for airlines present during the entire
sample periods, the lower part of the table shows the results from the balanced sample, which is
balanced in terms of market-carrier level. Since these airlines in the balanced sample were present
in all sample periods, they can be understood as more regular and significant airlines in the relevant
markets.
Columns (1) and (4) in Table 2.7 are the results from the OLS estimation. For these specifi-
cations, dummy variables CS Exit and Post are included, along with their interaction term, rather
than the market-ticketing carrier and time fixed effects for comparison purpose. Columns (2), (3),
(5), and (6) include the unit and time fixed effects, and the controls are included in Columns (3)
and (6). The negative coefficients on CS Exit in Columns (1) and (4) show that, on average, ticket
66
prices on the CS Exit markets were lower than the ticket prices on the CS Keep markets. When
the treatment of the post dummies are controlled, no codeshare exit effects are detected in the case
of all airlines included. When we focus on Alaska and American, however, 10% significant code-
share exit effects are detected. While the codeshare exit effects are not detected in Column (1),
the fixed effects results show the pro-competitive effects of the codeshare exit. The results suggest
that the price changes on the market in which Alaska and American ceased to operate their code-
share products were lower than the price changes from the airlines on the markets in which Alaska
and American continued to operate their codeshare products. These differences are approximately
4% when we include all airlines; the differences increase to around 9% when we focus on Alaska
and American observations. In other words, the price changes of AS-AA products on the CS Exit
markets were about 9% lower than the price changes of AS-AA products on the CS Keep market.
From these findings, we can conclude that the AS-AA codeshare discontinuation is associated with
average price decreases among the airlines on the CS Exit markets.
Table 2.8 shows the results produced when we include the markets in which AS-AA codeshare
products were included in at least one pre-remedy period on the relevant market, i.e., codeshare
market Definition B. Similar to the results shown in Table 2.7, the OLS results are included for
comparison, with the first three columns corresponding to the sample in which all airlines are
included, while the last three columns are the results from the sample focused only on Alaska and
American observations. Even though many of the coefficients of interest lost their significance
compared to the results in Table 2.7, the coefficients are all negative, suggesting that no anti-
competitive effects resulted from the codeshare discontinuation. Comparing the results from the
alternative definitions, we can conclude that, at a minimum, the codeshare exit produced no anti-
competitive effects, and is in fact associated with relative price decreases, especially for the AS
and AA products on the CS Exit markets.
While these results support the pro-competitive effects of codeshare discontinuation, a few crit-
ical issues must be addressed. The first such issue is potential measurement error in the treatment
indicator. As stated in Section 2.5.2.1, a misclassified binary regressor is likely to cause estimators
67
to be biased. In the current context, misclassification is mainly due to the fact that the treatment
indicator is classified based on the presence of the relevant products in the data, where the data
are a 10% sample. As a result, some of the markets classified as CS Exit markets might actually
be markets in which AS-AA continued their codeshare products, whereas the real codeshare exit
markets cannot be misclassified as the CS Keep markets. If some of the CS Exit markets were
misclassified, the estimators are likely to suffer from attenuation bias, which means that the actual
effects might have been stronger.
The other potential concern regarding these results is that the codeshare exit effects might be
contaminated by the merger effects. Because the AS-AA codeshare discontinuation was closely
related to the AS-VX merger, the merger effects must be controlled for when examining the code-
share exit effects. Even though it is difficult to clearly disentangle the two effects, we did examine
whether Virgin (VX) operated in the codeshare relevant markets. If Virgin had offered its prod-
ucts in these markets, the merger might have had significant effects in the markets. Figure 2.7
shows how many markets were served by Virgin among the AS-AA codeshare relevant markets
from 2014Q1 to 2016Q3. As shown in the figure, the majority of the relevant markets were not
served by Virgin America. Although there might have been some impact from the change of the
entire network due to the merger, the relatively small presence of Virgin America in the relevant
market suggests that the codeshare exit effects are less likely to be sensitive to the merger effects,
if any exist. In fact, the results in Section 2.6 also support the conclusion that the AS-VX merger
generally did not have strong price effects.
2.7.3 Parallel Trends Assumption
Causal interpretations of two-way fixed effect DID estimates require parallel trends and con-
stant treatment effects assumptions. One common method for arguing a parallel trends assumption
is to conduct pre-trends test, as in the event study literature (Autor 2003; Sun and Abraham 2021).
Even though the absence of pre-trends does not fully exclude the possibility of post-trends that may
exist without the intervention, we estimate the effects under the event study approach to determine
68
whether the treatment group and the control group had significantly different trends prior to the
remedy. Specifically, we estimate the following equation:
lnFare
imt
=a
0
+b
tå
t
CS Exit D
t
+Controls+g
im
+l
t
+e
imt
=a
0
+b
2
CS Exit D
2
+b
0
CS Exit D
0
+b
1
CS Exit D
1
+Controls+g
im
+l
t
+e
imt
;
(2.3)
where D
t
is a dummy variable for each time. The last pre-remedy period is excluded as a reference
group.
27
In other words, b
2
, b
0
, and b
+1
represent relative differences between the airlines on
the CS Exit markets and the airlines on the CS Keep markets, compared to their difference at the
last pre-remedy period.
The results of this analysis are reported in Table 2.9.
28
The upper part of the table reports the
results from the Definition A sample, while the lower part of the table reports the results from the
Definition B sample. Columns (1) and (3) are from Columns (3) and (6) of Tables 2.7 and 2.8.
The insignificant coefficients on Pre
2
show that there were no significant pre-merger trends in
the pre-remedy period.
2.8 Demand
In this section, we present an air travel demand model in order to investigate the effect of
eliminating AS-AA codeshare products. Similar to Verboven (1996), we use a version of the two-
level nested logit model. The model is also similar to the demand model employed in Gayle and
Brown (2014), which examined the demand-increasing effects of virtual codeshare.
27
Sun and Abraham (2021) document that most recent studies conducting event study specification exclude relative
periods close to the initial treatment.
28
Though we report the unbalanced panel case only, the results of the balanced panel are similar.
69
2.8.1 Model
In the two-level nested logit model, products are first divided into G+ 1 groups, g= 0;1;:::;G,
with each group g further divided into H
g
subgroups, h= 1;:::;H
g
. In our model, the first nest is an
identity of the ticketing carrier, while the second nest is whether a product is nonstop or onestop.
The first nest G also include an outside option, g= 0. The outside option includes options not to
travel or to travel using transportation other than airlines. In other words, a passenger first chooses
whether to air travel with a specific airline or not to air travel, then if the passenger decides to
travel with the specific airline, the passenger chooses whether to use a direct/nonstop flight or a
connecting/onestop flight. This two-level nested logit model allows preferences to be correlated
among products in a group as well as a subgroup. The nesting structure is illustrated in Figure 2.8.
Following Verboven (1996), a passenger i’s utility of choosing a product j on an airport-pair m
at time t is:
u
i jmt
=d
jmt
+n
i jmt
; (2.4)
whered
jmt
is a mean-utility that is equal for all passengers andn
i jmt
is an individual specific part.
n
i jmt
is specified as n
i jmt
=e
imt
+e
igmt
+(1s
g
)e
ihgmt
+(1s
h
)e
i jmt
, where s
h
represents the
correlation of the passenger’s utility across products in the same subgroup, while s
g
captures the
correlation of the utility across products in the same group. Because products in the same sub-
group should be considered closer substitutes compared with ones in other groups, a restriction of
0s
g
s
h
< 1 should hold for this model to be valid. e
igm
,e
ihgmt
, ande
i jmt
have unique distribu-
tion, such that e
igmt
, (1s
g
)e
ihgmt
+(1s
h
)e
i jmt
and e
igmt
+(1s
g
)e
ihgmt
+(1s
h
)e
i jmt
are
extreme value random variables.
d
jmt
is the mean level of utility of choosing product j on the airport-pair m at time t. It is a
function of the characteristics of product j, such as price, and observed non-price characteristics.
Specifically,d
jmt
is specified as follows:
d
jmt
= X
jmt
ba p
jmt
+q
1
CS Exit
m
+q
2
Post
t
+q
3
CS Exit
m
Post
t
+f
a
+l
t
+g
m
+x
jmt
; (2.5)
70
where X
jmt
is a vector of non-price observed product characteristics. We include dummy vari-
ables for nonstop flights, inconvenience of the product, and an indicator of traditional codeshare
as observed product characteristics. A nonstop dummy takes a value of 1 if the flight is a direct
flight. The coefficient on this variable is expected to be positive because passengers tend to prefer
direct flights over flights with intermediate stops. As suggested in Gayle (2013), onestop flights
might have different effects on utility according to their itinerary distance. In order to account
for this possibility, an inconvenience variable is calculated as the itinerary distance divided by the
nonstop miles of the route. The traditional codeshare dummy takes a value of 1 if the product is
a traditional codeshare of any airline (not just the AA/AS or AS/AA traditional one). p
jmt
is the
passenger weighted average price of the product. CS Exit takes a value of 1 if the market was
served by AS-AA traditional codeshare in the pre-merger period but not in the post period. It takes
a value of 0 if, in the market, AS-AA codeshare products were operated both in the pre-merger
and post-remedy periods. Post is a dummy variable which takes a value of 1 for the post period
and 0 for the pre period. f
a
,g
m
, andl
t
are ticketing carrier fixed effects, market fixed effects, and
time fixed effects, respectively. Even though the DB1B data contain many detailed information
of tickets, it does not include information about some of ticket-specific restrictions information,
which would have affected passengers’ choices. For example, whether the ticket is refundable or
not is not included in the data. In order to account for these kinds of unobserved product quality
measures,x
jmt
is included in the specification.
b is a vector of parameters representing marginal utilities associated with the observed product
characteristics. a captures the marginal utility of the price. q
1
represents whether there are any
differences in mean utility between the products in the CS Exit markets and the products in the
markets in which AS-AA codeshare products were operated in both the pre and post periods (the
CS Keep markets). q
2
captures whether there are any changes in mean utility over the pre and post
periods for the products on the CS Keep markets. q
3
captures whether the change in mean utility
over time differs across the CS Exit markets and CS Keep markets; this represents the demand
effects of the AS-AA codeshare discontinuation.
71
2.8.2 Estimation
The two-level nested logit model explained in the previous subsection can be estimated by the
following equation:
lnS
j
lnS
0
= X
jmt
ba p
jmt
+q
1
CS Exit
m
+q
2
Post
t
+q
3
CS Exit
m
Post
t
+s
h
lnS
jjhg
+s
g
lnS
hjg
+f
a
+l
t
+g
m
+x
jmt
; (2.6)
where S
j
is the unconditional probability of product j being selected and S
0
is the probability of an
outside option being chosen. S
jjhg
is the probability of product j being chosen given the selection of
its subgroup. Similarly, S
hjg
is the probability of the subgroup h being selected given the selection
of its group g. These values are used as natural log forms. Each probability is calculated as
the observed market shares. Following Berry and Jia (2010), the market size is calculated as
the geometric mean of the origin and destination cities’ populations, and each observed share is
calculated as the total number of passengers divided by the market size.
Since p
jmt
, lnS
jjhg
, and lnS
hjg
are likely to be correlated with unobserved product character-
istics x
jmt
, instrumental variables should be used to estimate the parameters. Proper instrumental
variables include cost shifters, which directly affect supply (and thus markup) but not demand. The
characteristics of other products can be also used as instruments. Instrumental variables should
provide variations to identify relative shares of the product type (i.e., nonstop or onestop) as well
as the within group share of each product. The instrumental variables we included are (1) flown
miles, (2) jetfuel price as interacted with the legacy airline dummy,
29
(3) the number of other prod-
ucts offered by the airline on the market, (4) market-airline level average of inconvenience, and (5)
market-airline level average number of stops on the market.
30
Using these instrumental variables,
the model is estimated with 2SLS.
29
Changes in jetfuel price might affect lagacy airlines and low-cost airline differently, because these two types of
airlines have different cost structures.
30
We consider IVs in Gayle (2013), Gayle and Brown (2014), and Shen (2017).
72
2.8.3 Results
The results of this analysis are provided in Columns (1)-(4) of Table 2.10. The first two columns
present the estimation without using the instrumental variables, while Columns (3) and (4) present
the estimation using the instrumental variables. Overall, the results for OLS and 2SLS are similar
even though the exogeneity of the endogenous variables are strongly rejected. For all specifi-
cations, ticketing carrier fixed effects, time fixed effects, and market fixed effects are included;
however, the results for these fixed effects are not reported.
The coefficients on the observed product characteristics are as expected. Higher fares are asso-
ciated with lower utility. Direct flights are associated with higher utility compared to connecting
flights. Products with higher inconvenience measures are associated with lower utility. From these
results, we can confirm that passengers prefer shorter trips. Traditional codeshare products are
valued less by passengers compared to pure online products. As virtual codeshare products might
be inferior substitutes for pure online products (Ito and Lee, 2007), traditional codeshare products
are likely to be considered as inferior substitutes.
In terms of the nesting parameters, passengers typically recognize products with the same
ticketing carriers as close substitutes with only one nest, as shown in Column (3). The products
with the same number of stops given the ticketing carriers are considered as close substitutes
with two nests, as shown in Column (4). Though the coefficient on lnS
hjg
is not significant, the
restriction of 0s
g
s
h
< 1 is partially satisfied. As the coefficients on lnS
jjhg
are statistically
significant, we can argue that there are correlations among products in a subgroup. In other words,
passengers perceive products in the same group as closer substitutes compared to ones in other
groups. Given thats
g
is not significant, if we consider it to be 0, the model collapses to a one-level
nested logit with the number of stops (i.e., direct or indirect flight) as nests.
The significantly negative coefficients on CS Exit Post support the conclusion that the AS-
AA codeshare discontinuation negatively impacted passengers’ utility. In other words, while the
codeshare remedy decreased the average fares on the CS Exit markets, it also might have decreased
73
utility for those goods. As a result, it is not straightforward to derive welfare impact of the code-
share cessation.
A nested logit model can be sensitive to the nesting structure. Because lnS
hjg
is insignificant
in Column (4), we conduct a demand estimation with an alternative nesting structure. In this
structure, the first nesting group is composed of whether to take a nonstop flight or an onestop
flight (or choose an outside option), and the subgroup is composed of each ticketing airline. This
nesting structure is illustrated in Figure 2.9. The results are reported in Columns (5)-(8), Table
2.10. The results are generally similar to those of the first nesting structure, with the exception
of the significance of s
g
. Note that in this nesting structure, s
g
captures preference correlation
among products with the same number of stops. The results for the main indicator variables are
consistent with the alternative nesting structure, which corroborate the demand-decreasing effects
of the codeshare discontinuation.
2.9 Conclusion
This study examines the effects of Alaska-American (AS-AA) codeshare discontinuation on
prices and demand, as well as the relative price changes associated with the merger between Alaska
(AS) and Virgin America (VX). The results from the market level fare change comparisons suggest
that the AS-VX merger was not associated with any significant impact on prices. This might be
due to their relatively small number of overlapping routes compared to mergers between larger
airlines. It might be also be due to the remedies imposed by the governmental authorities to reduce
potential anti-competitive effects from the merger. However, additional research is required to
precisely determine the underlying mechanism.
In terms of the codeshare exit effects, our findings show that the discontinuation of AS-AA
codeshare products is associated with price decreases, since relative price decreases are detected
on the routes for which AS-AA discontinued their codeshare products. While these results may
indicate that the codeshare discontinuation was overall beneficial to consumers in terms of lower
74
prices, the relative lower prices might be associated with a decrease in demand. In order to address
this issue, we estimate an air travel demand model, concluding that the codeshare discontinua-
tion did negatively impact consumers’ utility. As a result, welfare implications cannot be directly
derived.
Applying the results of this study to other mergers or industries might not be straightforward,
because each case has important contexts for the purpose of the analysis. Nevertheless, this study
does have important policy implications, as the remedies were determined to have helped to de-
crease prices. Given that partnerships between major airlines are crucial strategic choices for the
airlines as well as sources of anti-competitive concerns from the DOJ, the findings of this study
can provide useful insight for relevant issues.
31
31
Partnerships among major airlines are active concerns of the DOJ. For example, the
DOJ recently sued to block an alliance between American Airlines and JetBlue in Septem-
ber 2021. Relevant information can be found at https://www.justice.gov/opa/pr/
justice-department-sues-block-unprecedented-domestic-alliance-between-american-airlines-and
(last accessed in Apr. 2022).
75
Tables and Figures
Figure 2.1: US Airlines’ Market Share in 2016
Figure 2.2: Three Types of Airlines Products
These three tickets represent an example of the three types of products on the market between LAX and BOS. The
first nonstop ticket is marketed and operated by American Airlines, and is an example of a ”pure online” product.
The second ticket is operated partially by the marketing carrier (AA) and partially by the carrier (AS) apart from the
marketing carrier (AA), and is defined as a ”traditional codeshare” product. The last ticket represents a product wholly
operated by the carrier (AS) apart from the marketing carrier (AA), and is defined as a ”virtual codeshare” product.
76
Figure 2.3: AS-AA Codeshare Markets (Traditional)
Figure 2.4: AS-AA Codeshare Markets (Virtual)
77
Figure 2.5: AS-AA Codeshare Total Passnegers (Traditional)
Figure 2.6: AS-AA Codeshare Total Passnegers (Virtual)
78
Table 2.1: Remedy Relevant Markets with Codeshare Products
Time
CAT 1 CAT 2 CAT 3 CAT 4 Total
VX/AA Overlap AS/AA Overlap AS KEY AA KEY LAX:KEY Relevant Mkt
ASAA AAAS ASAA AAAS ASAA AAAS ASAA AAAS ASAA AAAS
2015Q1
58 36 1520 5155 30 6615
0 0 5 11 74 26 0 0 74 28
2015Q2
58 38 1593 5270 30 6803
0 0 3 7 71 22 0 1 71 24
2015Q3
58 52 1600 5274 30 6816
0 0 8 11 71 41 0 0 71 44
2015Q4
58 52 1557 5319 30 6816
1 1 7 6 68 35 1 2 68 39
2016Q1
58 48 1511 5207 30 6660
0 0 5 7 49 29 1 1 49 32
2016Q2
58 50 1586 5308 30 6836
1 4 7 18 58 38 1 4 58 45
2016Q3
58 49 1597 5308 30 6845
0 4 4 14 56 46 0 5 56 49
2016Q4
58 64 1560 5299 30 6807
0 4 5 12 59 31 0 4 59 33
2017Q1
58 62 1533 5229 30 6713
0 3 4 12 71 25 0 3 71 27
2017Q2
58 69 1570 5316 30 6840
0 0 6 2 66 30 0 0 68 32
2017Q3
58 67 1584 5279 30 6813
0 0 6 2 65 26 0 0 66 28
2017Q4
58 69 1578 5316 30 6849
0 0 2 0 59 18 0 0 59 18
a
Each column stands for each categories of the codeshare remedies defined on original Final Judgment documented
by the Department of Justice. The first row of each quarter contains the number of remedy markets that document
clarified with definitions and the second row for the number of markets where the codeshare products between
two airlines were observed in data. ASAA stands for the codeshare product marketed by Alaska and operated by
American, AAAS stands for the vice versa.
79
Table 2.2: Regional Airlines affiliated to Major Airlines
Regional Airlines Major Airlines Note
16, OH AA 16! US ( 2015Q2), OH! AA (2015Q3 )
b
17, PT AA 17! US (2015Q1), PT! AA (2015Q2 )
c
MQ AA 1998
9E DL 2013Q2
QX AS 1986
a
Source: 2016 Annual Report of Regional Airlines Association, based on 2016 schedules.
b
PSA Airlines changed its IATA code from 16 to OH at 20150701.
c
Piedmont Airlines changed its IATA code from 17 to PT at 20150401.
Table 2.3: Codeshare Exit Analysis Sample Summary Statistics
Variable Mean Std. Dev. Min Max
Definition A (observations: 1,960)
CS Exit 0.551 0.498 0 1
Post 0.484 0.5 0 1
lnFare 6.23 0.374 4.382 7.489
Fare 540.34 181.70 80 1789
Population: origin (mil.) 3.963 3.047 0.097 19.335
Population: destination (mil.) 3.155 3.669 0.148 19.335
Unemployment rate: origin 4.273 0.644 2.4 7.533
Unemployment rate: destination 4.342 0.807 2 9.167
Definition B (observations: 4,640)
CS Exit 0.657 0.475 0 1
Post 0.49 0.5 0 1
lnFare 6.218 .346 4.043 7.535
Fare 530.48 172.68 57 1872
Population: origin (mil.) 3.717 3.497 0.049 19.335
Population: destination (mil.) 3.315 3.604 0.023 19.335
Unemployment rate: origin 4.248 0.708 2 8.267
Unemployment rate: destination 4.31 0.796 2 9.167
a
Definition A defines codeshare markets as routes where AS-AA codeshare products are
operated during the all sample pre periods and at least one quarter during the post-remedy
period. Definition B defines codeshare markets as routes where AS-AA codeshare products
are operated at least one quarter during the pre and post periods, respectively.
80
Table 2.4: Demand Estimation Sample Summary Statistics
Variable Obs Mean Std. Dev. Min Max
Y (= lnS
j
lnS
0
) 5,143 -13.34 1.45 -16.02 -7.33
Fare (in $100) 5,143 5.5 2.01 0.8 19.98
lnS
jg
5,143 -1.7 1.51 -8.43 0
lnS
jjhg
5,143 -1.4 1.18 -6.49 0
lnS
hjg
5,143 -0.3 1.06 -8.19 0
Nonstop 5,143 0.07 0.26 0 1
Inconv 5,143 1.13 0.15 1 2.21
Tra CS 5,143 0.41 0.49 0 1
CS Exit 5,143 0.57 0.5 0 1
Post 5,143 0.48 0.5 0 1
CS Exit Post 5,143 0.27 0.44 0 1
a
Definition A is used for the sample.
Table 2.5: Pre & Post Differences: Market Level Fares
Market Def. No. of Markets Pre Post Differences Diff-in-Diff
[(2) - (1)] [(3) - Neither]
(1) (2) (3) (4)
Overlap 69 441.01 435.89 -5.12 13.39
(11.44) (11.35) (16.12) (16.27)
Either 304 401.57 394.49 -7.08 11.43
(4.95) (5.16) (7.15) (7.49)
Exit 15 473.84 476.79 2.95 21.46
(21.53) (28.45 ) (35.68) (35.75)
Enter 40 515.14 479.85 -35.29 -16.78
(13.66) (14.88) (20.2) (20.32)
Neither 2,583 419.68 401.16 -18.51**
(1.53) (1.64) (2.24)
a
Standard errors are in parentheses.
b
Only markets with at least 900 passengers per quarters and with minimum average number
of airlines that is greater than 2 are included.
c
Column (4) difference-in-differences use Neither market as a control group.
d
** : 1% significance, * : 5% significance,y: 10% significance.
81
Table 2.6: Merger Effects Results
Overlap Markets Either Markets
(1) (2) (3) (4)
OLS FE OLS FE
AS/VX 0.0108 -0.0148
(0.0256) (0.0159)
Post -0.165*** -0.0233*
(0.0245) (0.0128)
AS/VX Post -0.00874 -0.0440 -0.00315 0.0155
(0.0403) (0.0337) (0.0224) (0.0137)
No. of Markets 69 304
No. of AS/VX 138 359
No. of the Others 367 1,265
Observations 1,719 5,803
Adjusted R
2
0.112 0.019 0.006 0.004
Market-Carrier FE No Yes No Yes
Time FE No Yes No Yes
Controls Yes Yes Yes Yes
a
The dependent variable is the natural log of passenger weighted average
ticket fare for the ticketing carrier on the market.
b
All specifications include populations and unemployment rates of the ori-
gin and destination cities, but the results are not reported in the table.
c
*** : 1% significance, ** : 5% significance, * : 10% significance.
82
Table 2.7: The Codeshare Exit Effects Results (Definition A)
All Airlines AS - AA
(1) (2) (3) (4) (5) (6)
OLS FE FE OLS FE FE
Unbalanced Panel
CS Exit -0.182*** -0.159***
(0.0216) (0.0329)
Post 0.0351 0.0635*
(0.0218) (0.0345)
CS Exit Post -0.0387 -0.0478** -0.0421* -0.0952* -0.0850** -0.0897**
(0.0322) (0.0200) (0.0238) (0.0490) (0.0361) (0.0434)
No. of Markets 108 108 108 108 108 108
No. of Carriers 537 537 537 189 189 189
Observations 1,960 1,960 1,960 719 719 719
Adjusted R
2
0.071 0.004 0.005 0.091 0.015 0.019
Balanced Panel
CS Exit -0.173*** -0.166***
(0.0220) (0.0328)
Post 0.0316 0.0598*
(0.0212) (0.0340)
CS Exit Post -0.0481 -0.0481** -0.0458** -0.0881* -0.0881** -0.0917**
(0.0323) (0.0190) (0.0226) (0.0489) (0.0349) (0.0409)
No. of Markets 107 107 107 104 104 104
No. of Carriers 428 428 428 162 162 162
Observations 1,712 1,712 1,712 648 648 648
Adjusted R
2
0.076 0.006 0.007 0.103 0.016 0.019
Market-Carrier FE No Yes Yes No Yes Yes
Time FE No Yes Yes No Yes Yes
Controls No No Yes No No Yes
a
The dependent variable is the natural log of passenger weighted average ticket fare of the ticketing carrier on
the market.
b
Definition A defines codeshare markets as routes where AS-AA codeshare products are operated during the
all sample pre periods and at least one quarter during the post-remedy period. Definition B defines codeshare
markets as routes where AS-AA codeshare products are operated at least one quarter during the pre and post
periods, respectively.
c
Standard error are in parentheses and clustered at the market-airline level.
d
*** : 1% significance, ** : 5% significance, * : 10% significance.
83
Table 2.8: The Codeshare Exit Effects Results (Definition B)
All Airlines AS - AA
(1) (2) (3) (4) (5) (6)
OLS FE FE OLS FE FE
Unbalanced Panel
CS Exit -0.110*** -0.118***
(0.0141) (0.0216)
Post 0.0241 0.0467*
(0.0164) (0.0255)
CS Exit Post -0.0222 -0.0211 -0.0099 -0.0560* -0.0423* -0.0358
(0.0207) (0.0136) (0.0150) (0.0316) (0.0243) (0.0269)
No. of Markets 279 279 279 279 279 279
No. of Carriers 1,303 1,303 1,303 464 464 464
Observations 4,640 4,640 4,640 1,676 1,676 1,676
Adjusted R
2
0.027 0.004 0.007 0.051 0.011 0.012
Balanced Panel
CS Exit -0.101*** -0.117***
(0.0137) (0.0213)
Post 0.0209 0.0482*
(0.0158) (0.0259)
CS Exit Post -0.0253 -0.0253* -0.0140 -0.0466 -0.0466* -0.0406
(0.0203) (0.0132) (0.0147) (0.0318) (0.0239) (0.0258)
No. of Markets 276 276 276 256 256 256
No. of Carriers 992 992 992 365 365 365
Observations 3,968 3,968 3,968 1,460 1,460 1,460
Adjusted R
2
0.028 0.005 0.008 0.051 0.010 0.010
Market-Carrier FE No Yes Yes No Yes Yes
Time FE No Yes Yes No Yes Yes
Controls No No Yes No No Yes
a
The dependent variable is the natural log of passenger weighted average ticket fare of the ticketing
carrier on the market.
b
Definition A defines codeshare markets as routes where AS-AA codeshare products are operated dur-
ing the all sample pre periods and at least one quarter during the post-remedy period. Definition B
defines codeshare markets as routes where AS-AA codeshare products are operated at least one quarter
during the pre and post periods, respectively.
c
Standard error are in parentheses and clustered at the market-airline level.
d
*** : 1% significance, ** : 5% significance, * : 10% significance.
84
Figure 2.7: VX Presence on AS-AA Codeshare Routes
85
Table 2.9: Parallel Trends Assumption Tests
All Airlines AS-AA
(1) (2) (3) (4)
FE Event Study FE Event Study
Definition A
CS Exit Post -0.0421* -0.0897**
(0.0238) (0.0434)
Pre -2 -0.0287 -0.0370
(0.0237) (0.0409)
Post +0 -0.0426 -0.0730
(0.0283) (0.0514)
Post +1 -0.0699** -0.147**
(0.0332) (0.0651)
No. of Markets 108 108
No. of Carriers 537 189
Observations 1,960 719
Adjusted R
2
0.005 0.006 0.019 0.021
Definition B
CS Exit Post -0.0099 -0.0358
(0.0150) (0.0269)
Pre -2 -0.0166 -0.0190
(0.0156) (0.0261)
Post +0 0.00408 -0.0282
(0.0180) (0.0331)
Post +1 -0.0414** -0.0634
(0.0205) (0.0390)
No. of Markets 279 279
No. of Carriers 1,303 464
Observations 4,640 1,676
Adjusted R
2
0.007 0.009 0.012 0.013
a
Market-Carrier fixed effects, Time fixed effects, and Controls are included in
all specifications.
b
Definition A defines codeshare markets as routes where AS-AA codeshare
products are operated at least one quarter during the pre and post periods, re-
spectively. Definition B defines codeshare markets as routes where AS-AA
codeshare products are operated during the all sample periods (two quarters for
each pre and post).
c
Standard error are in parentheses and clustered at the market-airline level.
d
*** : 1% significance, ** : 5% significance, * : 10% significance.
86
Air Travel
Outside
Option
AA
nonstop onestop
AS
nonstop onestop
Airline
m
nonstop onestop
. . .
Figure 2.8: The Nesting Structure 1
Air Travel
Outside
Option
Nonstop
A
1
A
2
A
n
Onestop
A
1
A
2
A
o
. . . . . .
Figure 2.9: The Nesting Structure 2
87
Table 2.10: Demand Estimation Results
Nesting Structure 1 Nesting Structure 2
OLS 2SLS OLS 2SLS
(1) (2) (3) (4) (5) (6) (7) (8)
Variable 1-level 2-level 1-level 2-level 1-level 2-level 1-level 2-level
Fare (in $100) -0.0243*** -0.0297*** -0.157*** -0.301*** -0.00703** -0.00885** -0.129*** -0.256***
(0.00763) (0.00751) (0.0600) (0.0677) (0.00339) (0.00366) (0.0318) (0.0608)
lnS
jg
0.509*** 0.216*** 0.855*** 0.825***
(0.0240) (0.0242) (0.0247) (0.0388)
lnS
jjhg
(s
h
) 0.684*** 0.452*** 0.878*** 0.652***
(0.0149) (0.0443) (0.0204) (0.0754)
lnS
hjg
(s
g
) 0.330*** 0.0626 0.821*** 0.340**
(0.0287) (0.0465) (0.0302) (0.142)
Nonstop 2.149*** 2.270*** 2.532*** 2.537*** 2.362*** 2.360*** 2.326*** 2.460***
(0.148) (0.164) (0.112) (0.119) (0.316) (0.312) (0.101) (0.109)
Inconv -1.685*** -1.407*** -2.261*** -1.968*** -0.471*** -0.460*** -0.618*** -1.354***
(0.157) (0.145) (0.101) (0.117) (0.105) (0.104) (0.119) (0.270)
Tra CS -0.593*** -0.449*** -0.733*** -0.484*** -0.116*** -0.106*** -0.0912** -0.274***
(0.0452) (0.0371) (0.0466) (0.0617) (0.0225) (0.0206) (0.0439) (0.0844)
CS Exit 0.225* 0.512*** 0.0604 0.838** 2.395*** 2.371*** 2.625*** 1.718***
(0.123) (0.107) (0.349) (0.391) (0.120) (0.118) (0.261) (0.438)
Post 0.0665 0.0608 0.0902** 0.107** 0.0160 0.0171 0.0384 0.0814**
(0.0406) (0.0449) (0.0457) (0.0477) (0.0552) (0.0555) (0.0280) (0.0399)
CS Exit Post -0.0777 -0.119** -0.102** -0.164*** -0.201*** -0.202*** -0.212*** -0.196***
(0.0511) (0.0548) (0.0513) (0.0547) (0.0696) (0.0694) (0.0316) (0.0442)
Constant -10.08*** -10.48*** -8.768*** -8.771*** -12.10*** -12.11*** -11.50*** -9.902***
(0.285) (0.264) (0.337) (0.373) (0.178) (0.174) (0.288) (0.582)
No. of Markets 108 108 108 108 108 108 108 108
Observations 5,143 5,143 5,143 5,143 5,143 5,143 5,143 5,143
Adjusted R
2
0.696 0.724 0.606 0.575 0.871 0.872 0.851 0.727
a
Standard error are in parentheses.
b
*** : 1% significance, ** : 5% significance, * : 10% significance.
88
Chapter 3
Determinants of Low-Cost Carriers’ Operational Performance
3.1 Introduction
The United States (U.S.) airline industry has changed in various aspects since the Airline
Deregulation Act in 1978.
1
One of the most critical changes in this industry is the growth of
low-cost carriers (LLCs).
2
Figure 3.1 shows the changes in market share of both legacy and low-
cost carriers since 1990.
3
As clearly shown in the figure, LCCs’ market share has been growing
consistently since 1990, and it almost doubled in the last 15 years. To examine the rapid expan-
sion of LCCs more closely, Table 3.1 shows the domestic non-stop segment market shares of some
major full-service airlines and the LCCs. Most of the full-service airlines’ market shares have not
changed substantially, and some of them have actually decreased, whereas the market shares of
the LCCs have increased consistently. Among the LCCs, Southwest Airlines is the world’s largest
LCC as well as one of the largest airlines in the U.S. domestic market. The LCCs’ growth appears
to be due to both the consistent success of the largest one—Southwest Airlines (WN)—and the
rapid expansion of relatively smaller carriers such as Spirit Airlines (NK).
1
The Airline Deregulation Act is a U.S. federal law that removed specific controls primarily related to ticket fares
and airlines’ entry or exit on routes.
2
Passenger airlines can be largely categorized into two groups according to their cost structures; legacy carriers
(full-service carriers) and LCCs. Legacy airlines are traditional carriers that were mostly established before deregula-
tion, whereas low-cost airlines are relatively recent ones with lower fares and fewer services.
3
Market share is calculated by using revenue passenger miles (RPMs) and available seat miles (ASMs), respec-
tively. RPMs are a measure of air passenger traffic, and ASMs can be considered as a measure of an airline’s capacity.
89
Despite it being apparent that the LCCs’ market share has increased to approximately 30%, and
LCCs might have particular features different from full-service carriers, research on the behavior
and phenomena regarding these airlines under their perspectives is rather limited.
4
This lack of
research might be due to the fact that most of the LCCs are still relatively small compared to
the major full-service airlines, and empirically, fewer data are available because most of these
carriers have expanded relatively recently. Although there are many topics that can be investigated
that relate to LCCs, I focus on LCCs’ operational performance, particularly on-time performance,
which is considered to be an important quality measure in the airline industry. Because full-
service airlines and LCCs are expected to provide different service quality under varying fares,
and the LCCs compete with each other as well as with the full-service airlines, it is important to
understand the features of LCCs’ service quality. Moreover, research focusing on LCCs’ service
quality is worthwhile because the effects of changes in market structures from deregulation on
service quality are less obvious and may vary across different market characteristics, whereas the
effects on flight fares are relatively straightforward. In other words, given the critical roles of LCCs
and their potential distinct features, a better understanding of LCCs’ service quality would lead to
an improved understanding of the U.S. airline industry.
On-time performance (OTP) is one of the four major categories of airline performance in the
annual Airline Quality Rating (AQR), which is a statistical study of major airline performance in
the U.S.
5
Hence, OTP is crucial in consumers’ satisfaction and also important in terms of airline
functions because flights are a mode of transport. One of the most common causes that airline
companies blame for delays is unexpected weather conditions. Even though OTP is affected by
weather conditions and other factors, airlines can still make efforts to improve this form of per-
formance. In other words, they can invest more to enhance their resilience to unexpected events.
If airlines were aware that additional investment to improve OTP could increase their profits, they
would most likely choose to do so.
4
Most of the studies on LCCs have focused on the Southwest Effect because there has been a substantial impact in
terms of increases in travel traffic and decreases in ticket fares due to the entry of Southwest (e.g., Morrison (2001)).
5
The four categories are on-time performance (OT), denied boardings (DBs), mishandled baggage (MB), and
customer complaints (CCs).
90
Because the LCCs are growing, it has become possible to use their OTP data.
6
I not only utilize
these OTP data but also data for daily weather conditions and aircraft information to investigate
the determinants of LCCs’ operational performance. Specifically, I examine whether the LCCs
consider possible externalities that can cause delays and internalize in the case of their hub or
focus city airports, as full-service airlines do.
The LCCs initiated (and some of them still do) their operations focusing on short- and medium-
haul routes under a point-to-point system rather than long-haul ones, which means that their most
targeted consumers are local market passengers. However, the rapid expansion of these airlines has
enabled them to operate under a hub-and-spoke system similar to that provided by the full-service
airlines; hence, the LCCs can serve more connecting passengers through focus cities.
7
In this case,
LCCs might behave differently at their core vs. non-core airports. To examine this effect and some
other determinants of LCCs’ OTP, I specify a delay equation and estimate the effects of relevant
factors.
The remainder of this paper is organized as follows. In Section 3.2, I briefly present some
of the previous studies of service quality (OTP) in the airline industry. Section 3.3 displays data
sources as well as the sample and the variable constructions. Section 3.4 presents the empirical
specifications, and Section 3.5 discusses the estimation results. In the final section, I conclude.
3.2 Literature Review
In this section, I discuss some of the previous empirical works on airlines’ service quality,
focusing on OTP. Because airlines’ service quality, particularly OTP, is a crucial aspect of airline
carriers, there have been many studies regarding this topic. Research on OTP has increased in
6
Only airlines that account for at least 1% of domestic scheduled passenger revenues are required to report these
data.
7
A focus city tends to indicate an airport from which the airline operates several point-to-point routes. However,
it also has a function of a hub airport in the case of some low-cost airlines. For example, AirTran (FL) used Atlanta
International Airport (ATL) as a hub airport. Hub airports and focus cities might be distinguished according to whether
or not it serves as a transfer point, but the distinction became ambiguous.
91
terms of scale and scope because more detailed data have become available from the Bureau of
Transportation Statistics (BTS).
Previous empirical works on airlines’ service quality mainly focused on the effects of various
factors on major large airlines’ service quality. In other words, the relevant studies investigated
the determinants of airlines’ OTP. Mazzeo (2003) investigates the effects of competition—market
concentration—on OTP using several alternative measures of competition and concludes that flight
delays are more common and longer in duration on routes with less competition. Mayer and
Sinai (2003) examine OTP, focusing on two determinants of performance; network benefits from
hubbing and congestion externalities. The findings suggest that the network benefits are a more
critical reason for carriers accepting delays. Forbes and Lederman (2010) conduct empirical works
on the effects of vertical integration on airlines’ OTP as well as flight cancellations by estimating
whether the use of an owned (rather than independent) regional at a particular airport affects main
airlines’ performance. The findings show that the integrated (with regional airlines) airlines had
systematically better operational performance than the non-integrated mainline airlines. Ater and
Orlov (2015) investigate the relationship between Internet access and OTP to determine whether
lower search costs would make quality worse along with lower prices, and their empirical results
support this supposition.
All of the above studies include large mainstream airlines. Hence, they do not include most
of the LCCs in their sample.
8
This lack of inclusion is most likely related to the BTS’s on-time
performance reporting requirement criteria. Because only large airlines are required to provide
these data, OTP data for the relatively smaller LCCs are not available.
Some of the relevant studies focused on the changes and reactions of major full-service airlines
rather than concentrating on LCC performance itself. For example, Prince and Simon (2015)
examine whether and how incumbent airlines change their service quality in response to entry and
8
Mazzeo (2003) includes Alaska, American West, American, Continental, Delta, Northwest Southwest, TWA,
United, and US Airway. Forbes and Lederman (2010) include American, Continental, Delta, Northwest, TWA, United,
and US Airways. Ater and Orlov (2015) include Alaska, American, Continental, Delta, Northwest, Southwest, United,
and US airways. Because Southwest Airlines is one of the largest carriers, it is typically included in the empirical
studies among LCCs.
92
entry threat, mainly by Southwest Airlines. The main findings show that OTP worsens when entry
carriers are low-cost airlines, potentially due to cost-cutting or post-entry differentiation. Bubalo
and Gaggero (2015) examine whether a higher presence of LCCs can improve the OTP of flights
at airports using data for the European airline industry. The results in this paper support that the
presence of LCCs reduces delays; this main finding contrasts with the results of Prince and Simon
(2015), who use U.S. airline industry data.
3.3 Data and Summary Statistics
In this section, I present the data sources as well as sample construction. The variables used in
estimation are also explained in detail.
3.3.1 Data Sources
The main OTP data used in this paper are taken from the Bureau of Transportation Statistics
(BTS). Any certified U.S. airline carriers
9
that comprise at least one percent of domestic scheduled
passengers revenues are urged to report relevant information regarding OTP, including the sched-
uled and actual departure/arrival times of their flights. The data contain not only the duration of
delays but also the origin/destination airport pair and aircraft information; hence, it is useful to
analyze airlines’ delays in detail. The raw data contain OTP data for most of the major full-service
carriers as well as expanding LCCs and regional airlines.
The main OTP data are combined with a couple of other types of data. Daily weather data were
collected via a commercial online weather service website. The weather records were obtained
from the National Weather Service (NWS) database. Because weather stations are located in most
large airports, the data contain precise weather condition information at each airport for each day.
These data contain minimum, maximum, and average temperatures for each day as well as any
9
Officially, a certified air carrier is one holding a Certificate of Public Convenience and Necessity issued by the
DOT to conduct scheduled services interstate.
93
weather events such as rain, thunderstorm, and fog. I use such event information because the
takeoff and landing of a flight would be more affected by these factors rather than temperatures.
Aircraft-related data are gathered from the Federal Aviation Administration (FAA) U.S. Air-
craft Registry database. These data contain not only the aircraft model but also when it was man-
ufactured and registered. Using tail number information in the OTP data, I match each flight with
the registry data to obtain aircraft-specific information. To include big airports, enplanement data
are also taken from the FAA.
3.3.2 Sample Construction
Whereas the raw data contain information on many routes, I apply some restrictions on the data
to construct the sample to study LCCs’ operational performance.
I use the October 2015 data because most of the low-cost airlines did not report flight-level OTP
data when they were relatively small in terms of revenue. Table 3.2 shows whether or not each low-
cost carrier reported OTP data in each year. This table provides information about which LCCs
had at least 1% of revenue share in a particular year. More reporting LCCs over time indicates their
growth and success. Because I use 2015 year data, the LCCs included in the sample are Frontier
Airlines (F9), JetBlue Airways (B6), Southwest Airlines (WN), Spirit Air Lines(NK), and Virgin
America (VX).
A non-stop flight between two airports is used as a unit of observation. A flight is directional,
which means that the data distinguish origin and destination airports. For instance, a flight from
Los Angeles (LAX) to Portland (PDX) airports is different from a flight from Portland to Los An-
geles airports. Flights in the data are categorized into four groups; delayed, not delayed, canceled,
and diverted. I exclude canceled or diverted flights from the sample because these flights were not
the main interest of this paper.
The top busiest 30 airports, each of which is served by at least one of the five LCCs besides
full-service airlines, are included in the sample. The airport size is based on year 2015 enplanement
data from the FAA. Table 3.3 provides a list of the airports included, as well as which low-cost
94
airlines operated at each airport. Because all of the airports are in contiguous states, any particular
characteristics from non-contiguous states are irrelevant in this paper. This airport selection leaves
me with 504 directional routes served by at least one of the LCCs.
The final sample for estimation includes 57,605 flights departed from and arrived at one of the
30 airports. In the sample, there are 504 directional routes, each of which is served by at least one
low-cost airline, on 31 days of October 2015.
3.3.3 Variables
This subsection presents the variables used in the estimation. Although the explanatory vari-
ables can be categorized into variables of specific interest and control variables, I do not distinguish
them because they are all tested to be the determinants of low-cost airlines’ OTPs.
3.3.3.1 On-time Performance Variables
To study the determinants of LCCs’ operational performance, I use a couple of alternative
dependent variables—either continuous or binary measures of performance.
One of the most natural measures of OTP is a flight delay. Although the terms on-time and
delay can be somewhat arbitrary, there are several ways to measure OTP. One is a binary vari-
able—whether a flight is delayed by 15 minutes or more. This variable is included in the raw
data because a delay time duration of less than 15 minutes would be acceptable both in terms of
consumer satisfaction and operational efficiency (or airport schedule configurations). Table 3.4
and Table 3.5 show the summary statistics of these variables according to the day of the week and
scheduled departure/arrival times.
Although the above binary measure is one way to analyze airlines’ performance, the duration
of a delay also matters because a consumer is likely to perceive a 30-minute delay as differing from
a 16-minute one. To take this factor into account, I use a continuous measure of delays, dep delay
and arr delay, which are the time differences (in minutes) between scheduled departure/arrival
time and actual departure/arrival time. If a flight departs from or arrives at an airport earlier than
95
the scheduled time, the variables take negative numbers. Assuming consumers prefer an earlier
departure or arrival, flights with negative delays are left in the sample. Out of the 57,605 flights,
22,548 flights departed late (even one minute), and 19,547 flights arrived late. The statistics show
that the average early departure time (excluding exact time departure, i.e., dep delay = 0) is ap-
proximately 4 minutes, and the average early arrival time is about 12 minutes, whereas the average
delay times are roughly 24 minutes for departure and 27 minutes for arrival.
10
3.3.3.2 Explanatory Variables
Whereas many major carriers operate based on a hub-and-spoke system, some of the LCCs
function based on a point-to-point system. For example, one of the largest airlines—Southwest
Airlines—does not follow a hub-and-spoke system, whereas Frontier Airlines and Virgin America
operate based on a hub-and-spoke approach. In the case of the LCCs that do not use a hub-and-
spoke system, it is not obvious if the carriers’ operational performance, particularly OTP, depends
on whether a flight is from or heading to its core airports. Even though they do not have hub
airports, their point-to-point system also has some core cities, which are termed focus cities. A
focus city is one where an airline operates several routes focused on local market passengers rather
than connecting ones. Although these cities are selected to be large markets for local market
passengers, they can also serve as a hub airport. Because a focus city can be thought of as a type
of hub airport, any features specific to a hub airport might be applied to a focus city as well. For
instance, Southwest Airlines flights from or to focus cities might have different characteristics than
flights from or to non-focus cities. Using this feature, I use dummy variables that indicate whether
a flight is from or heading to the carriers’ hub or focus cities. In other words, I treat focus cities as
the same as hub airports in the analysis.
Defining a hub airport and a focus city is not trivial and is even more difficult in the case of
LCCs for two reasons. First, there is no systematic way to define hub and focus cities across
different carriers. Second, there is a substantial difference in sizes (market shares, the number of
10
The averages are calculated over the relevant observations. For example, the average departure delay time is
calculated over the flights that departed later than the scheduled departure time.
96
serving routes, etc.) among the LCCs (i.e., Southwest is dominant in terms of size). In this manner,
serving 15 routes from an airport would make the airport a hub for Virgin America, but this would
not be the case for Southwest. To take each airline’s size into account, I treat an airport as a hub
or a focus city for the airline if it serves relatively many routes compared to the other airports that
the carrier also serves. Whether a flight departs from or arrives at a hub or a focus city might affect
operational performance differently. Thus, I distinguish between these two and include dummy
variables separately. dep hub indicates whether a flight departs from a hub or a focus city, and
arr hub indicates whether a flight arrives at a hub or a focus city of the airline. If an airline takes
the number of routes served from an airport into account and internalizes possible congestion from
delays, the sign of the coefficients on these variables would be negative.
Travel distance is also included in the analysis. In this paper, a distance might have to do with
an aircraft size (and thus the number of transported passengers) and/or more time in the air. I
use both continuous distance and categorized groups in alternative specifications. For the dummy
variables, observations are categorized into three groups. The first group is a set of flights for
which the travel distance is less than 750 miles. The second one includes flights for which the
distance is from 750 miles to 2,000 miles, and the last group contains flight distances of more than
2,000 miles. The distance ranges are set to reflect short-, medium-, and long-haul travel. LCCs
initially began their services based on short- and medium-haul travel but have provided more long-
haul travel recently. A longer distance means more airborne time. This increase might affect the
OTP of departure and arrival differently because a longer distance could imply more time to adjust
for arrival punctuality. The mean distances for each carrier is 999.64 miles for Frontier, 1200.82
miles for JetBlue, 1,041.75 miles for Spirit, 883.98 miles for Southwest, and 1,438.764 for Virgin
America. Table 3.6 displays the proportions of each haul group for each carrier. Examining the
proportions of each haul group for each carrier, it appears that Frontier, Spirit, and Southwest did
not operate many long-haul flights compared to the other two.
11
11
The proportions do not necessarily represent the airlines’ entire network structure because the sample only in-
cludes the top 30 busiest airports.
97
For technical reasons, the age of an aircraft can affect operational performance. This variable
is particularly relevant in LCCs’ operational performance context because using a relatively old
aircraft is one way for low-cost airlines to reduce costs. To take the technical aspect into account, I
include the ages of each aircraft by combining the data sets from the Bureau of Statistics (BTS) and
U.S. Aviation Registry data from the FAA. Tail number information in the OTP data allows me to
obtain aircraft manufacture year information. The average age of the five LCCs is approximately 9
years.
12
Southwest has the oldest aircraft. Southwest uses 30-year-old planes, the oldest ones, on
various routes, whereas the oldest aircraft used by the others is 16 years (used by JetBlue). Among
the five airlines, the recently expanding airlines—Spirit and Virgin America—had younger aircraft
on average compared to the others. It is anticipated that more recently manufactured aircraft tend
to perform better than older ones because it would be easier to manipulate and adjust the speed of
a flight. Recently manufactured aircraft might also take less time to be inspected before departure,
and is less likely to cause a departure delay.
To examine the effects of a hub or a focus city on LCCs’ operation performance, several other
factors are controlled. One of the most common causes that airline companies blame for delays is
unexpected weather conditions. Hence, these conditions are included as dummy variables in the
analysis. Among weather conditions, events such as rain, thunderstorms, and fog are included as
explanatory variables because they might affect the takeoff and landing of a flight significantly. Be-
cause the weather in both origin and destination airports might have different effects on departure
and arrival, I distinguish weather in origin airport from destination airport. org thunder, org rain,
org fog, dep thunder, dep rain, and des fog take a value of 1 if the corresponding weather event
occurred at the origin or the destination airport. Extremely cold weather may affect carriers’ op-
erational performance, but I do not include temperature information because I use the data from
October. Indeed, none of the airport temperatures are below 30 degrees Fahrenheit.
Besides weather conditions, airport congestion might be a key OTP determinant. Regarding
airport congestion, there would be a within-week trend in arrivals and departures. For example,
12
aircraft age is calculated as 2015 - manufactured year.
98
one would expect an airport to be more crowded on Fridays than Tuesdays or Wednesdays. Because
dep delay 15 and arr delay 15 indicate whether or not a flight is late by 15 minutes or more, the
average values of those variables can provide a closer look into within-week fluctuations. Table
3.4 shows the summary statistics of those variables according to each day of the week. Friday has
the highest average values both for arrival and departure delays of 15 minutes or more, whereas
there are fewer delays in terms of a percentage on Tuesdays and Wednesdays. Thus, day of week
dummies are included to control the effects.
In a similar way, scheduled departure/arrival time might also affect airlines’ OTP. The effect
of scheduled time on delays can be considered from two perspectives. The first one is that there
are different numbers of flights in each time slot. More flights are scheduled during peak hours
than in the early morning, and this difference might, therefore, affect airline delays. The other
one is that delays might be worsened because earlier delays accumulate over time. This factor
would make later flights tend to be delayed more. The summary statistics of dep delay 15 and
arr delay 15 according to scheduled time are given in Table 3.5. In the table, each time represents
a time window from the specified time to the next specified time. For example, the row 5am
includes flights scheduled to depart or arrive between 5am-5:59am. More flights are scheduled in
the afternoon and evening than early in the morning and late at night, and delays of 15 minutes
or more tend to increase (in percentage terms) from morning to night. To take this difference
into account, I use two alternative variables; one is discrete, and the other is continuous. For the
discrete variable, I divide a day into seven three-hour time slots starting from 1:00 and add the
dummy variables in the estimation. The base group is from 1:00–3:59 (note that a 24-hour clock
is used in this paper). The continuous variables, dep time day and arr time day convert scheduled
departure/arrival times into 0 to 1 scales. 00:00 corresponds to 0.
Definitions and summary statistics for the variables are provided in Table 3.7.
99
3.4 Empirical Specification
I model LCCs’ operational performance as a function of several independent variables I pre-
sented in the previous section. Although I examine alternative specifications, the basic delay equa-
tions are specified as follows;
dep delay
i jod
=a+borg weather
ijod
+gdep hub
i jod
+ddistance
i jod
+qage
i jo
+hday of week
ijo
+z dep time
ijo
+e
j
+e
o
+ u
i jod
(3.1)
arr delay
i jod
=a+bdes weather
ijod
+garr hub
i jod
+ddistance
i jod
+qage
i jod
+hday of week
ijod
+z dep time
ijod
+e
j
+e
d
+ u
i jod
(3.2)
where an observation is characterized by an individual flight i, an airline j, an origin airport o, and
a destination airport d.
org weather and des weather are vectors of weather conditions containing thunderstorm, rain,
and fog dummies. day of week represents variables that control within-week trends. When a vec-
tor of day-of-week dummies is used, the signs and magnitudes of the coefficients on the dummies
should be interpreted compared to Sunday because Sunday is excluded from both departure and
arrival estimation to be a reference group. dep time and arr time capture the effects of scheduled
time on operational performance. When dummy variables are included, the groups of 4:00–6:59
for departures and 7:00-9:59 for arrivals are selected as bases due to the fact that they are the time
slots with the lowest delay rates in terms of the dep delay 15 and arr delay 15 statistics.
Because I use flight-level observations, I include carrier- (e
j
) and airport-specific fixed effects
(e
o
and/or e
d
). By adding the fixed effects, one can examine the effects of other variables on
airlines’ delays while controlling for carrier- and airport-specific unobserved characteristics, which
affect operational performance. u
i jod
represents an individual flight idiosyncratic shock, and it
100
captures all the other unobserved factors that are not captured by the above variables as well as the
controls.
3.5 Results
This section discusses the results, which reveal the determinants of LCCs’ on-time performance
(OTP).
3.5.1 Departure On-time Performance
3.5.1.1 Departure Delay Duration
Table 3.8 shows the results of departure delay regression with or without carrier- and airport-
specific unobserved factors. The first two columns show the results without any fixed effects,
Columns (2) provide the results with carrier and origin airport fixed effects, and the last two
columns show the results including carrier, origin, and destination airport fixed effects. Regardless
of whether or not the fixed effects are included, most of the variables have expected coefficient
signs, and they are significant at the 1% level. Weather conditions at the origin airports have a
relatively large impact on departure delay time. Among them, thunderstorms appear to be the most
critical determinant of departure delays. The coefficients on both day-of-week and departure time
variables seem to be intuitive. Compared to Sundays, Tuesdays are associated with the lowest
impact on departure delays, whereas Fridays are the only day of the week that increase departure
delays more compared to Sundays. The results for the scheduled time variables are also as antici-
pated. All other time slots than the 4:00–7:00 group appear to increase departure delays compared
to the reference group.
Regarding travel distance, longer travel distances increase departure delays. This effect might
be related to aircraft size. Airlines tend to use bigger aircraft to serve longer distances. Such larger
aircraft and/or more passengers would imply that more time is required to depart at an origin airport
because boarding takes more time. Higher aircraft age increases departure delays. As explained in
101
the previous section, old aircraft require more effort to operate, and such aged aircraft need more
time to be inspected before departure; hence, the results seem to be reasonable.
Lastly, whereas the significance levels are different in the alternative specifications, departing
from a hub or a focus city airport is associated with a decrease in departure delays. Even though
not significant in the case of no fixed effects being included, the negative signs on the org hub
coefficients show that departure delays are less severe when airlines departed from their hub or
focus cities. This finding can be interpreted as the carriers internalizing likely externalities caused
by delays when they depart from their core airports. A delay from one flight can increase delays of
other flights operated at the airport, and airlines that operate many flights in the airport are likely
to be affected by the accumulated delays more. Because airlines run more flights from their hub or
focus city airports, the airlines at the hub or focus city airports attempt to avoid delays more than
they would at non-core airports.
Table 3.9 shows the regression results using alternative control variables. Both the weather and
day-of-week variables are included in all of the specifications but are not reported in the table. The
signs and magnitudes of the coefficients on these variables are almost the same as in the previous
table, Table 3.8. The carrier, origin airport, and destination airport fixed effects are also included in
all of the specifications. Columns (1) show the results with the continuous measure of scheduled
departure time, whereas Columns (2) illustrate the results with the categorized distance measure.
Columns (3) show the results when the categorized variables for distance and schedule time are
included.
Columns (1) show that the continuous distance variable loses its significance when the con-
tinuous measure of schedule time is included. With respect to the discrete distance variables, the
short-haul group is excluded as the reference group. Compared to the short-haul group, medium-
haul travel is associated with an increase in departure delays, whereas long-haul travel is not sig-
nificantly different from the short-haul group. This observation implies that travel distance is
non-linearly associated with departure delays.
102
dep time day is associated with increases in departure time both in Columns (1) and (2). This
finding is reasonable because departure delays can accumulate over time from early morning to
late at night. The signs of the coefficients on the org hub and aircraft age are consistent across the
alternative variables.
Because each observation is a directional flight, the characteristics of the destination airport
might affect the departure performance of a flight. For example, even if a flight is ready to depart
at the origin airport, it cannot leave because of the weather conditions at the destination airport. To
see the effects, I estimate the departure delays, including some characteristics of the destination
airports. The results are given in Table 3.10, Columns (1). Whereas the weather conditions of the
destination airport have significant effects on departure delays, whether a flight is heading to its
hub or focus city airport does not seem to be a significant determinant of departure performance,
even though the sign on the variable is negative.
3.5.1.2 Probability of Being Delayed by 15 Minutes or More
Table 3.11 shows the results of the estimation using an alternative delay variable, dep delay 15.
Logit is used to estimate the determinants’ effects, and all specifications include the carrier, origin,
and destination fixed effects. The main findings of the variables except org hub are consistent
with the previous results in terms of the signs and relative magnitudes of the coefficients. When it
comes to org hub, the variable of main interest, only the first specification shows significant delay
decrease effects of departing from a hub or a focus city airport. For the last two specifications
(Columns (2) and (3)), the probability of being delayed by 15 minutes or more is approximately
14.3% among the flights departed from non-core airports and about 13.7% among the flights de-
parted from hub or focus city airports, and they are not statistically different.
13
The marginal
effects and their 95% confidence intervals are provided in Figure 3.2. As shown in the figure, none
of the effects are significant at the 95% level, even though the effects are all negative.
13
The probabilities are calculated, holding the variables at their means.
103
In sum, almost all determinants turn out to be as expected across the alternative specifications.
More importantly, departing from a hub or focus city appears to reduce the departure delay duration
as well as the probability of being delayed by 15 minutes or more, even though less significant.
Distance seems to have non-monotonic, but not decreasing, effects on departure delays, and aircraft
age and unfavorable weather conditions increase departure delays.
3.5.2 Arrival On-time Performance
3.5.2.1 Arrival Delay Duration
Table 3.12 shows the results of arrival delay time regressions with or without carrier- and
airport-specific fixed effects. Similar to Table 3.8, the first specification does not include any
fixed effects (Columns (1)), whereas the other columns show the results when the fixed effects are
included.
As in the results of departure delays, the most critical weather event is a thunderstorm among
the weather conditions. The signs and magnitudes of the coefficients on both day-of-week and
scheduled-arrival-time variables are all significant and as expected. The peak days of the week
and times are associated with an increase in arrival delays.
Contrary to the variables for which the effects are similar to those in the departure delay anal-
ysis, the distance variables provide interesting results. In contrast to the results of the departure
delay analysis, a longer distance reduces arrival delay duration. This relationship can be also seen
in Table 3.13 with alternative distance variables. Compared to short-haul travel, medium-haul de-
creases arrival delays, and long-haul travel reduces them even further. It appears that there is a
monotonic relationship between travel distance and arrival delays, potentially because, in contrast
to the case of a departure delay, a longer distance might imply more airborne time to adjust arrival
time. Thus, it is possible that a longer distance implies a larger-size aircraft or more passengers in
the case of departure, but it implies more time to manipulate airborne time once the flight departs.
des hub reduces arrival delay time, and this relation is significant in all of the specifications.
When a flight is heading to a hub or a focus city airport, it is likely to have more connecting
104
passengers than the case of a non-hub airport. Because an arrival delay at a core airport would
worsen carriers’ performance by successive delays in arrivals as well as departures, carriers might
take these effects into account and endeavor to avoid arrival delays more at a core airport than they
would at a non-core airport.
Table 3.13 shows the regression results using the alternative distance and scheduled-arrival-
time variables. The continuous scheduled-time variable is significantly positive, indicating that the
arrival delays increase because scheduled arrival times are late in the day. The effects of whether
arriving at a hub or focus city airport are also consistent under the alternative control variables.
An arrival delay is likely to depend on the characteristics of the origin airport. In particular,
the departure delay of the flight is likely to be a strong factor that causes a delay in arrival. To
examine the effects, I include some of the origin airport characteristics as well as departure delay
duration in the analysis. The relevant results are shown in Table 3.10, Columns (2) and (3). When
the origin airports’ characteristics are included, except the departure delay duration (Columns (2)),
both departing from or arriving at a hub or a focus city airport are associated with a decrease
in arrival delays. The weather conditions at the origin airport are also effective determinants of
arrival delays and even had stronger effects than the weather conditions at the destination airport.
When departure delay duration is included (Columns (3)), interestingly, the sign of the coefficient
on org hub becomes positive, even though the significance level is lower. Compared with the
results in Columns (2), Columns (3) show that the departure delay duration is one of the strongest
determinants of arrival delays. The adjusted R
2
(0.867) of this specification is also significantly
higher than the others. Even with departure delay duration being controlled, des hub is found to
be a significant factor.
3.5.2.2 Probability of Being Late by 15 Minutes or More
Table 3.14 shows the results of logit estimation under three different settings, using arr delay 15
as a delay measure. Most of the explanatory variables show similar results with the findings in Ta-
ble 3.12 and Table 3.13. Unlike the estimation results of dep delay 15, where des hub loses its
105
significance in some of the specifications, des hub is consistently significant to reduce arrival de-
lays in all of the cases. The marginal effects of arriving at hub or focus city airports and their 95%
confidence levels are given in Figure 3.3. As shown in the figure, the probability of arriving late
by 15 minutes or more is roughly 1% lower when airlines are heading to a hub or focus city airport.
In sum, arriving at a hub or a focus city and a longer travel distance appear to reduce not only
the probability of being late by 15 minutes or more but also arrival delay duration, whereas aircraft
age has the opposite effect. Most of the other controls have expected effects on arrival delay.
3.6 Conclusion
This paper examines the operational performance of LCCs, focusing on on-time performance
(OTP). Although many previous studies have conducted research on full-service carriers’ perfor-
mance because they were the main players in the U.S. domestic airline industry, this paper focuses
on LCCs using recently available data on their OTP. The main focus of this paper is whether de-
parting from or arriving at a core airport affects LCCs’ OTP as is in the case of full-service carriers
who operate based on a hub-and-spoke system.
For both departure and arrival performance, departing from and/or arriving at a hub or a focus
city airport reduce delay duration. These results are consistent under the alternative specifications
with the alternative variables. Such results can be understood as carriers internalizing the potential
externalities caused by delays to some degree. Regarding the probability of being late by 15
minutes or more, arriving at a core airport significantly reduces the probability of such a delay,
whereas departing from a core airport is not significant. Older aircraft and unfavorable weather
conditions increase not only the probability of being delayed by 15 minutes or more but also the
delay duration as expected. Concerning travel distance, this factor has the opposite effects on
departure and arrival delays. Although it increases (non-monotonically) departure delays (even
though by only a small amount), it reduces arrival delay duration, possibly because more airborne
106
time is available to manipulate arrival OTP. In all cases, the schedule-related variables are found to
be significant determinants of OTP.
Although there are distinct features of low-cost carriers (LCCs) from full-service airlines, the
determinants of OTP are found to be similar in the sense that they incorporate congestion external-
ities. Because an increased amount of data for LCCs is now available, more studies on the growing
airlines can be beneficial for understanding the industry better.
107
Tables and Figures
Figure 3.1: U.S. Domestic Market Shares
108
Table 3.1: Domestic Non-segment Market Share, 2000-2015
Full-Service Airlines
Year Alaska American Delta Hawaiian United US Airway
2000 2.07 11.17 16.15 0.96 11.86 9.34
2001 2.23 10.82 15.35 0.96 11.26 9.29
2002 2.31 13.76 14.91 1.01 10.33 7.78
2003 2.32 12.12 13.05 0.95 9.50 6.27
2004 2.33 11.28 12.35 0.86 9.38 5.90
2005 2.28 11.56 11.58 0.85 8.30 5.58
2006 2.32 11.50 9.48 0.91 8.62 4.80
2007 2.31 11.13 8.92 1.01 8.22 5.40
2008 2.31 10.82 8.97 1.16 7.83 7.36
2009 2.30 10.55 8.87 1.29 7.29 7.14
2010 2.39 10.30 14.09 1.28 6.82 7.11
2011 2.53 10.07 14.24 1.24 6.13 7.15
2012 2.65 9.99 14.63 1.32 10.41 7.33
2013 2.81 9.94 15.01 1.35 9.97 7.70
2014 2.92 9.87 15.80 1.35 9.66 7.61
2015 3.10 13.30 16.33 1.35 9.87 3.55
Low-Cost Carriers
Year Allegiant Frontier JetBlue Southwest Spirit Sun Country
2000 0.02 0.49 0.18 13.22 0.45 0.33
2001 0.01 0.54 0.54 14.19 0.57 0.28
2002 0.02 0.67 1.01 14.07 0.68 0.07
2003 0.07 0.86 1.49 13.59 0.71 0.14
2004 0.11 0.97 1.79 13.66 0.72 0.17
2005 0.17 1.07 2.15 14.17 0.64 0.21
2006 0.32 1.26 2.69 15.38 0.63 0.21
2007 0.46 1.41 2.95 15.85 0.82 0.22
2008 0.64 1.52 3.08 16.58 0.84 0.20
2009 0.83 1.44 3.17 17.29 0.80 0.17
2010 0.91 1.39 3.29 17.69 0.88 0.16
2011 0.93 1.57 3.51 18.21 1.10 0.20
2012 1.06 1.53 3.80 18.25 1.39 0.21
2013 1.10 1.51 3.93 18.55 1.68 0.26
2014 1.21 1.68 3.93 19.73 1.90 0.27
2015 1.35 1.74 4.09 20.89 2.30 0.31
a
Numbers of passengers transported for domestic non-stop segments are used to compute the market shares.
b
Even though Northwest Airlines (FSC), Continental Airlines (FSC), AirTran Airways (LCC), and ATA (LCC)
also had noticeable market shares in several years between 2000 - 2015, only the airlines that operate until
2015 are included. Northwest and Continental merged with Delta (2010) and United (2012), respectively, and
AirTran was integrated into Southwest (2011).
109
Table 3.2: On-Time Performance Data Reporting LCCs
Year Allegiant Frontier JetBlue Southwest Spirit Sun Country Virgin America
2000 x x x o x x -
2001 x x x o x x -
2002 x x x o x x -
2003 x x o o x x -
2004 x x o o x x -
2005 x o o o x x -
2006 x o o o x x -
2007 x o o o x x x
2008 x o o o x x x
2009 x o o o x x x
2010 x o o o x x x
2011 x o o o x x x
2012 x o o o x x o
2013 x o o o x x o
2014 x o o o x x o
2015 x o o o o x o
a
Only low-cost carriers operating as of 2015 are included.
110
Table 3.3: List of Airports
Airport Code City Frontier JetBlue Southwest Spirit Virgin America
ATL Atlanta o - o o -
BOS Boston - o o o o
BWI Baltimore - o o o -
CLT Charlotte o o o - -
DAL Dallas - - o - o
DCA Washington D.C o o o - o
DEN Denver o o o o -
DFW Dallas-Fort o o - o -
DTW Detroit o o o o -
EWR Newark - o o - o
FLL Fort Lauderale - o o o o
IAD Dulles o o o - o
IAH Houston o - - o -
JFK New York - o - - o
LAS Las Vegas o o o o o
LAX Los Angeles o o o o o
LGA Flusing o o o o o
MCO Orlando o o o o o
MDW Chicago - - o - -
MIA Miami o - - - -
MSP Minneapolis o - o o -
ORD Chicago o o - o o
PDX Portland o o o o o
PHL Philadelphia o o o o -
PHX phoenix o o o o -
SAN San Diego o o o o o
SEA Seattle o o o - o
SFO San Francisco o o o - o
SLC Salt Lake City o o o - -
TPA Tampa o o o o -
111
Table 3.4: Statistics of dep delay 15 and arr delay 15: Day of the Week
Day of week Observations Mean S.D Min Max
dep delay 15
Sunday 7,414 0.217 0.412 0 1
Monday 7,703 0.142 0.349 0 1
Tuesday 7,578 0.098 0.298 0 1
Wednesday 7,568 0.128 0.334 0 1
Thursday 9,543 0.163 0.370 0 1
Friday 9,601 0.223 0.416 0 1
Saturday 8,198 0.160 0.366 0 1
Total 57,605 0.164 0.370 0 1
arr delay 15
Sunday 7,414 0.192 0.394 0 1
Monday 7,703 0.130 0.336 0 1
Tuesday 7,578 0.088 0.284 0 1
Wednesday 7,568 0.123 0.328 0 1
Thursday 9,543 0.170 0.376 0 1
Friday 9,601 0.235 0.424 0 1
Saturday 8,198 0.144 0.351 0 1
Total 57,605 0.158 0.364 0 1
112
Table 3.5: Statistics of dep delay 15 and arr delay 15: Scheduled Time
dep delay 15 arr delay 15
Time Obs Mean S.D. Obs Mean S.D.
0 am 97 0.144 0.353 571 0.186 0.389
1 am 75 0.200 0.403 83 0.205 0.406
2 am 0 0
3 am 0 5 0.200 0.447
4 am 0 173 0.127 0.334
5 am 840 0.048 0.213 642 0.159 0.366
6 am 4,096 0.052 0.222 651 0.163 0.369
7 am 4,345 0.072 0.259 1,988 0.093 0.291
8 am 3,923 0.089 0.284 2,627 0.074 0.262
9 am 3,934 0.100 0.300 3,612 0.086 0.280
10 am 3,542 0.144 0.351 3,583 0.110 0.313
11 am 3,318 0.149 0.356 2,883 0.094 0.291
12 pm 2,983 0.147 0.354 3,444 0.116 0.321
1 pm 3,337 0.186 0.389 3,269 0.126 0.332
2 pm 3,346 0.212 0.409 3,452 0.141 0.348
3 pm 3,203 0.210 0.407 2,700 0.145 0.352
4 pm 3,346 0.221 0.415 3,761 0.169 0.374
5 pm 3,587 0.220 0.414 3,314 0.190 0.392
6 pm 3,408 0.250 0.433 3,486 0.208 0.406
7 pm 3,363 0.251 0.434 3,447 0.192 0.394
8 pm 3,089 0.219 0.413 3,618 0.207 0.405
9 pm 2,307 0.192 0.394 3,156 0.246 0.430
10 pm 601 0.225 0.418 4,534 0.221 0.415
11 pm 865 0.192 0.3948 2,606 0.196 0.397
Total 57,605 0.164 0.370 57,605 0.158 0.364
113
Table 3.6: Distance Group Summary Statistic
Carrier Variable Obs # Mean S.D
Frontier dis short 4,907 0.315 0.465
dis medium 4,907 0.655 0.475
dis long 4,907 0.030 0.169
JetBlue dis short 10,078 0.305 0.461
dis medium 10,078 0.451 0.498
dis long 10,078 0.244 0.430
Spirit dis short 7,210 0.319 0.466
dis medium 7,210 0.622 0.485
dis long 7,210 0.060 0.237
Southwest dis short 30,352 0.475 0.499
dis medium 30,352 0.490 0.500
dis long 30,352 0.035 0.183
Virgin America dis short 5,058 0.370 0.483
dis medium 5,058 0.255 0.436
dis long 5,058 0.375 0.484
a
dis short = 1 if distance < 750 miles.
b
dis medium = 1 if 750 miles distance < 2,000 miles.
c
dis long = 1 if 2,000 miles distance.
d
Distance is distance between the origin airport and the destination
airport of a flight.
e
For all categories, Min value is 0 and Max value is 1.
114
Table 3.7: Variable Definitions and Summary Statistics
Variable Definition Mean S.D. Min Max
Dependent Variables
dep delay actual departure time - scheduled departure time 7.330 28.132 -25 577
dep delay 15 dummy = 1 if departure delay time 15 minutes 0.164 0.370 0 1
arr delay scheduled arrival time - actual arrival time 1.296 31.229 -81 577
arr delay 15 dummy = if arrival delay time 15 minutes 0.158 0.364 0 1
Independent Variables
dep hub dummy = 1 if a flight departs from a hub or a focus city 0.447 0.497 0 1
arr hub dummy = 1 if a flight arrives at a hub or a focus city 0.446 0.497 0 1
org thunder dummy = 1 if thunderstorm (origin airport) 0.056 0.229 0 1
org rain dummy = 1 if rain (origin airport) 0.262 0.440 0 1
org fog dummy = 1 if foggy (origin airport) 0.041 0.198 0 1
des thunder dummy = 1 if thunderstorm (destination airport) 0.055 0.228 0 1
des rain dummy = 1 if rain (destination airport) 0.260 0.438 0 1
des fog dummy = 1 if foggy (destination airport) 0.041 0.198 0 1
distance distance between airports (1,000 miles) 1.018 0.635 0.177 2.704
dis short dummy = 1 if distance < 750 miles 0.403 0.490 0 1
dis medium dummy = 1 if 750 distance < 2,000 miles 0.493 0.500 0 1
dis long dummy = 1 if distance 2,000 miles 0.104 0.305 0 1
age age of aircraft used in a route (year) 8.766 5.983 0 30
sun dummy = 1 if Sunday 0.129 0.335 0 1
mon dummy = 1 if Monday 0.134 0.340 0 1
tue dummy = 1 if Tuesday 0.132 0.338 0 1
wed dummy = 1 if Wednesday 0.131 0.338 0 1
thu dummy = 1 if Thursday 0.166 0.372 0 1
fri dummy = 1 if Friday 0.167 0.373 0 1
sat dummy = 1 if Saturday 0.142 0.349 0 1
dep time day scheduled departure time scaled in[0;1] 0.565 0.207 0.003 1.000
arr time day scheduled arrival time scaled in[0;1] 0.642 0.216 0.002 1.000
a
The distance group and day of week variables are included as factor variables in some specifications, but the summary statistics are reported
for separate dummy variables only.
115
Table 3.8: Determinants of Departure Delays: Main Results
(1) (2) (3)
Variable Coeff. S.E. Coeff. S.E. Coeff. S.E.
org hub -0.311 (0.227) -0.722* (0.405) -0.953** (0.415)
dist 0.946*** (0.195) 0.758*** (0.199) 0.546*** (0.205)
aircraft age 0.128*** (0.0209) 0.130*** (0.0212) 0.127*** (0.0218)
org thunder 8.261*** (0.793) 8.337*** (0.786) 8.343*** (0.786)
org rain 4.340*** (0.294) 4.517*** (0.301) 4.516*** (0.301)
org fog 3.368*** (0.742) 3.689*** (0.783) 3.687*** (0.782)
mon -3.432*** (0.458) -3.420*** (0.458) -3.418*** (0.458)
tue -7.835*** (0.399) -7.849*** (0.399) -7.858*** (0.399)
wed -3.969*** (0.472) -3.975*** (0.471) -3.975*** (0.471)
thu -2.580*** (0.441) -2.578*** (0.441) -2.578*** (0.441)
fri 2.078*** (0.485) 2.078*** (0.483) 2.080*** (0.483)
sat -3.130*** (0.444) -3.117*** (0.443) -3.131*** (0.444)
dep time 1 6.454 (4.171) 7.124* (4.208) 7.039* (4.279)
dep time 7 2.067*** (0.342) 1.941*** (0.348) 2.146*** (0.354)
dep time 10 5.963*** (0.385) 5.942*** (0.388) 6.140*** (0.393)
dep time 13 9.296*** (0.391) 9.126*** (0.395) 9.286*** (0.399)
dep time 16 11.54*** (0.408) 11.64*** (0.414) 11.84*** (0.420)
dep time 19 10.11*** (0.423) 10.01*** (0.430) 10.28*** (0.438)
dep time 22 6.805*** (0.814) 6.537*** (0.849) 6.254*** (0.875)
Constant 0.881 (0.602) 3.144*** (0.888) 2.909*** (1.121)
Carrier FE No Yes Yes
Origin Airport FE No Yes Yes
Dest Airport FE No No Yes
adj.R
2
0.0461 0.0496 0.0514
Obs 57,605 57,605 57,605
a
The dependent variable is dep delay.
b
*** : 1% significance, ** : 5% significance, * : 10% significance.
116
Table 3.9: Determinants of Departure Delays: Alternative Specification
(1) (2) (3)
Variable Coeff. S.E. Coeff. S.E. Coeff. S.E.
org hub -1.040** (0.410) -0.967** (0.410) -0.902** (0.416)
aircraft age 0.120*** (0.0217) 0.121*** (0.0216) 0.126*** (0.0217)
dist 0.216 (0.201) - - - -
dist medium - - 1.818*** (0.284) 1.772*** (0.285)
dist long - - -0.663 (0.483) 0.161 (0.490)
dep time day 17.05*** (0.554) 17.28*** (0.556) - -
dep time 1 - - - - 6.470 (4.279)
dep time 7 - - - - 2.155*** (0.354)
dep time 10 - - - - 6.137*** (0.393)
dep time 13 - - - - 9.241*** (0.399)
dep time 16 - - - - 11.87*** (0.420)
dep time 19 - - - - 10.37*** (0.439)
dep time 22 - - - - 6.568*** (0.877)
Constant 0.247 (1.141) 0.164 (1.107) 3.232*** (1.088)
adj.R
2
0.0399 0.0405 0.0444
Observations 57,605 57,605 57,605
a
The dependent variable is dep delay.
b
The weather conditions and the day of the week dummies are included in all specifications. The
signs and magnitudes of coefficients on these variables are almost the same as in Table 3.8.
c
Carrier fixed effects and airport fixed effects are included.
d
*** : 1% significance, ** : 5% significance, * : 10% significance.
117
Table 3.10: Origin & Destination Characteristics
(1) (2) (3)
dep delay arr delay arr delay
Variable Coeff. S.E. Coeff. S.E. Coeff. S.E.
dep delay - - - - 1.004*** (0.002)
org hub -1.026** (0.455) -1.203** (0.485) 0.295* (0.177)
des hub -0.267 (0.432) -1.278*** (0.465) -1.084*** (0.169)
aircraft age 0.127*** (0.022) 0.222*** (0.024) 0.0813*** (0.009)
dist 0.531*** (0.205) -3.996*** (0.226) -3.705*** (0.092)
org thunder 8.408*** (0.785) 10.72*** (0.854) 2.288*** (0.305)
org rain 4.044*** (0.297) 6.000*** (0.322) 1.923*** (0.121)
org fog 3.865*** (0.783) 4.725*** (0.843) 0.882*** (0.273)
des thunder 4.507*** (0.674) 8.468*** (0.758) 3.968*** (0.283)
des rain 3.622*** (0.304) 6.044*** (0.331) 2.421*** (0.126)
des fog 2.230*** (0.694) 4.028*** (0.784) 1.757*** (0.296)
Constant 4.917 (4.225) 3.814 (4.669) -12.79*** (1.457)
adj.R
2
0.0572 0.0943 0.867
Observations 57,605 57,605 57,605
a
The day of week dummies and the scheduled time dummies are included in all specifications, but
the coefficients on these variables are not reported.
b
Carrier fixed effects and airport fixed effects are included.
c
*** : 1% significance, ** : 5% significance, * : 10% significance.
118
Table 3.11: Determinants of Departure Delay Probability
(1) (2) (3)
Variable Coeff. S.E. Coeff. S.E. Coeff. S.E.
org hub -0.0654* (0.0375) -0.0518 (0.0376) -0.0464 (0.0376)
aircraft age 0.0175*** (0.00231) 0.0181*** (0.00233) 0.0180*** (0.00232)
org thunder 0.457*** (0.0461) 0.465*** (0.0462) 0.466*** (0.0462)
org rain 0.430*** (0.0273) 0.434*** (0.0274) 0.434*** (0.0274)
org fog 0.246*** (0.0586) 0.247*** (0.0587) 0.247*** (0.0587)
dist -0.0107 (0.0204) 0.0264 (0.0209) - -
dist medium - - - - 0.164*** (0.0304)
dist long - - - - -0.0717 (0.0492)
mon -0.479*** (0.0447) -0.474*** (0.0448) -0.475*** (0.0448)
tue -0.974*** (0.0492) -0.972*** (0.0493) -0.972*** (0.0493)
wed -0.631*** (0.0457) -0.626*** (0.0458) -0.626*** (0.0458)
thu -0.351*** (0.0406) -0.343*** (0.0407) -0.343*** (0.0407)
fri 0.0429 (0.0384) 0.0528 (0.0385) 0.0535 (0.0385)
sat -0.345*** (0.0428) -0.364*** (0.0429) -0.364*** (0.0429)
dep time day 2.132*** (0.0557) - - - -
dep time 1 - - 1.280*** (0.331) 1.226*** (0.331)
dep time 7 - - 0.532*** (0.0742) 0.536*** (0.0743)
dep time 10 - - 1.158*** (0.0725) 1.161*** (0.0726)
dep time 13 - - 1.553*** (0.0712) 1.550*** (0.0712)
dep time 16 - - 1.746*** (0.0707) 1.753*** (0.0709)
dep time 19 - - 1.658*** (0.0717) 1.671*** (0.0719)
dep time 22 - - 1.398*** (0.0965) 1.436*** (0.0964)
Constant -2.521*** (0.111) -2.529*** (0.124) -2.517*** (0.123)
pseudo R
2
0.0622 0.0713 0.0720
Observations 57,605 57,605 57,605
a
The dependent variable is dep delay 15.
b
*** : 1% significance, ** : 5% significance, * : 10% significance.
c
Carrier fixed effects and airport fixed effects are included.
119
Table 3.12: Determinants of Arrival Delays: Main Results
(1) (2) (3)
Variables Coeff. S.E. Coeff. S.E. Coeff. S.E.
des hub -0.344 (0.260) -0.999** (0.423) -0.911** (0.432)
dist -3.383*** (0.216) -3.750*** (0.222) -3.926*** (0.226)
aircraft age 0.230*** (0.0229) 0.221*** (0.0234) 0.229*** (0.0240)
des thunder 8.454*** (0.761) 8.416*** (0.760) 8.418*** (0.759)
des rain 6.698*** (0.328) 6.864*** (0.338) 6.846*** (0.337)
des fog 3.214*** (0.752) 3.703*** (0.787) 3.705*** (0.785)
mon -4.026*** (0.498) -4.044*** (0.497) -4.019*** (0.494)
tue -8.899*** (0.442) -8.930*** (0.442) -8.906*** (0.439)
wed -3.918*** (0.518) -3.939*** (0.516) -3.920*** (0.515)
thu -1.183** (0.483) -1.196** (0.482) -1.170** (0.480)
fri 4.654*** (0.539) 4.632*** (0.538) 4.662*** (0.536)
sat -4.699*** (0.486) -4.726*** (0.486) -4.682*** (0.483)
arr time 1 9.249** (4.599) 10.56** (4.639) 10.55** (4.579)
arr time 4 5.655*** (0.867) 4.527*** (0.892) 2.793*** (0.910)
arr time 10 2.278*** (0.370) 2.300*** (0.375) 2.409*** (0.376)
arr time 13 4.760*** (0.398) 4.762*** (0.401) 4.540*** (0.399)
arr time 16 8.811*** (0.408) 8.796*** (0.413) 8.531*** (0.413)
arr time 19 11.28*** (0.441) 11.21*** (0.443) 11.10*** (0.443)
arr time 22 9.950*** (0.465) 9.893*** (0.470) 9.874*** (0.471)
Constant 0.277 (0.624) -3.008*** (0.878) -0.167 (1.206)
Carrier FE No Yes Yes
Origin Airport FE No Yes Yes
Dest Airport FE No No Yes
adj.R
2
0.0659 0.0697 0.0762
Obs 57,605 57,605 57,605
a
The dependent variable is arr delay.
b
*** : 1% significance, ** : 5% significance, * : 10% significance.
120
Table 3.13: Determinants of Arrival Delays: Alternative Specification
(1) (2) (3)
Variable Coeff. S.E. Coeff. S.E. Coeff. S.E.
des hub -0.877** (0.429) -0.782* (0.429) -0.778* (0.432)
aircraft age 0.228*** (0.0239) 0.245*** (0.0238) 0.250*** (0.0239)
dist -3.542*** (0.221) - - - -
dist medium - - -2.873*** (0.312) -3.101*** (0.314)
dist long - - -7.130*** (0.531) -7.634*** (0.541)
arr time day 16.40*** (0.602) 16.35*** (0.601) - -
arr time 1 - - - - 9.972** (4.580)
arr time 4 - - - - 1.901** (0.907)
arr time 10 - - - - 2.286*** (0.378)
arr time 13 - - - - 4.432*** (0.400)
arr time 16 - - - - 8.326*** (0.414)
arr time 19 - - - - 10.87*** (0.442)
arr time 22 - - - - 9.675*** (0.470)
Constant -4.581*** (1.242) -7.493*** (1.208) -3.289*** (1.175)
adj.R
2
0.0730 0.0725 0.0754
Observations 57,605 57,605 57,605
a
The dependent variable is arr delay.
b
The weather conditions and the day of the week dummies are included in all specifications. The signs
and magnitudes of coefficients on these variables are almost the same as in Table 3.8.
c
Carrier fixed effects and airport fixed effects are included.
d
*** : 1% significance, ** : 5% significance, * : 10% significance.
121
Table 3.14: Determinants of Arrival Delay Probability
(1) (2) (3)
Variable Coeff. S.E. Coeff. S.E. Coeff. S.E.
des hub -0.0872** (0.0377) -0.0885** (0.0380) -0.0820** (0.0380)
aircraft age 0.0204*** (0.00242) 0.0206*** (0.00243) 0.0213*** (0.00242)
des thunder 0.529*** (0.0467) 0.530*** (0.0468) 0.530*** (0.0467)
des rain 0.581*** (0.0275) 0.583*** (0.0276) 0.582*** (0.0276)
des fog 0.240*** (0.0596) 0.240*** (0.0600) 0.240*** (0.0600)
dist -0.129*** (0.0210) -0.163*** (0.0213) - -
dist medium - - - - -0.129*** (0.0310)
dist long - - - - -0.323*** (0.0493)
mon -0.404*** (0.0465) -0.401*** (0.0466) -0.402*** (0.0466)
tue -0.935*** (0.0518) -0.935*** (0.0519) -0.935*** (0.0519)
wed -0.513*** (0.0470) -0.512*** (0.0471) -0.512*** (0.0471)
thu -0.126*** (0.0415) -0.125*** (0.0415) -0.125*** (0.0416)
fri 0.282*** (0.0392) 0.288*** (0.0393) 0.287*** (0.0393)
sat -0.338*** (0.0448) -0.335*** (0.0450) -0.335*** (0.0450)
arr time day 1.684*** (0.0588) - - - -
arr time 1 - - 1.139*** (0.287) 1.107*** (0.289)
arr time 4 - - 0.506*** (0.0891) 0.473*** (0.0887)
arr time 1 - - 0.335*** (0.0531) 0.332*** (0.0531)
arr time 1 - - 0.623*** (0.0517) 0.619*** (0.0518)
arr time 1 - - 1.029*** (0.0487) 1.023*** (0.0487)
arr time 1 - - 1.178*** (0.0482) 1.170*** (0.0482)
arr time 1 - - 1.114*** (0.0512) 1.107*** (0.0511)
Constant -2.460*** (0.115) -2.116*** (0.116) -2.252*** (0.114)
pseudo R
2
0.0750 0.0794 0.0792
Observations 57,605 57,605 57,605
a
The dependent variable is arr delay 15.
b
*** : 1% significance, ** : 5% significance, * : 10% significance.
c
Carrier fixed effects and airport fixed effects are included.
122
Figure 3.2: Conditional Marginal Effects of org hub
Figure 3.3: Conditional Marginal Effects of des hub
123
Bibliography
Arkhangelsky, D., Athey, S., Hirshberg, D. A., Imbens, G. W., and Wager, S. (2021). Synthetic
difference-in-differences. American Economic Review, 111(12):4088–4118.
Ater, I. and Orlov, E. (2015). The effect of the internet on performance and quality: Evidence from
the airline industry. Review of Economics and Statistics, 97(1):180–194.
Athey, S., Bayati, M., Doudchenko, N., Imbens, G., and Khosravi, K. (2021). Matrix comple-
tion methods for causal panel data models. Journal of the American Statistical Association,
116(536):1716–1730.
Autor, D. H. (2003). Outsourcing at will: The contribution of unjust dismissal doctrine to the
growth of employment outsourcing. Journal of labor economics, 21(1):1–42.
Berry, S. and Jia, P. (2010). Tracing the woes: An empirical analysis of the airline industry.
American Economic Journal: Microeconomics, 2(3):1–43.
Bilotkach, V . (2011). Multimarket contact and intensity of competition: evidence from an airline
merger. Review of Industrial Organization, 38(1):95–115.
Borenstein, S. (1990). Airline mergers, airport dominance, and market power. The American
Economic Review, 80(2):400–404.
Borenstein, S. and Rose, N. L. (2014). How airline markets work. . . or do they? regulatory reform
in the airline industry. In Economic regulation and its reform: What have we learned?, pages
63–135. University of Chicago Press.
Brueckner, J. K. (2003). International airfares in the age of alliances: The effects of codesharing
and antitrust immunity. Review of Economics and Statistics, 85(1):105–118.
Brueckner, J. K. and Flores-Fillol, R. (2007). Airline schedule competition. Review of Industrial
Organization, 30(3):161–177.
Brueckner, J. K. and Luo, D. (2014). Measuring strategic firm interaction in product-quality
choices: The case of airline flight frequency. Economics of Transportation, 3(1):102–115.
Brueckner, J. K. and Spiller, P. T. (1994). Economies of traffic density in the deregulated airline
industry. The Journal of Law and Economics, 37(2):379–415.
Brueckner, J. K. and Whalen, W. T. (2000). The price effects of international airline alliances. The
Journal of Law and Economics, 43(2):503–546.
124
Bubalo, B. and Gaggero, A. A. (2015). Low-cost carrier competition and airline service quality in
europe. Transport Policy, 43:23–31.
Capodaglio, G. (2019). On the reliability of structural methods in merger review. evidence from
the us airline industry. Evidence from the US Airline Industry (December 2019).
Carlton, D., Israel, M., MacSwain, I., and Orlov, E. (2019). Are legacy airline mergers pro-or anti-
competitive? evidence from recent us airline mergers. International Journal of Industrial
Organization, 62:58–95.
Chen, Y . and Gayle, P. G. (2007). Vertical contracting between airlines: An equilibrium analysis
of codeshare alliances. international Journal of industrial organization, 25(5):1046–1060.
Chen, Y . and Gayle, P. G. (2019). Mergers and product quality: Evidence from the airline industry.
International Journal of Industrial Organization, 62:96–135.
Das, S. (2019). Effect of merger on market price and product quality: American and us airways.
Review of Industrial Organization, 55(3):339–374.
De Chaisemartin, C. and d’Haultfoeuille, X. (2020). Two-way fixed effects estimators with het-
erogeneous treatment effects. American Economic Review, 110(9):2964–96.
Forbes, S. J. and Lederman, M. (2010). Does vertical integration affect firm performance? evidence
from the airline industry. The RAND Journal of Economics, 41(4):765–790.
Gayle, P. G. (2008). An empirical analysis of the competitive effects of the
delta/continental/northwest code-share alliance. The Journal of Law and Economics,
51(4):743–766.
Gayle, P. G. (2013). On the efficiency of codeshare contracts between airlines: is double marginal-
ization eliminated? American Economic Journal: Microeconomics, 5(4):244–73.
Gayle, P. G. and Brown, D. (2014). Airline strategic alliances in overlapping markets: Should
policymakers be concerned? Economics of Transportation, 3(4):243–256.
Goodman-Bacon, A. (2021). Difference-in-differences with variation in treatment timing. Journal
of Econometrics, 225(2):254–277.
He, G. and Wang, S. (2017). Do college graduates serving as village officials help rural china?
American Economic Journal: Applied Economics, 9(4):186–215.
Imai, K. and Kim, I. S. (2021). On the use of two-way fixed effects regression models for causal
inference with panel data. Political Analysis, 29(3):405–415.
Ito, H. and Lee, D. (2005). Domestic codesharing practices in the us airline industry. Journal of
Air Transport Management, 11(2):89–97.
Ito, H. and Lee, D. (2007). Domestic code sharing, alliances, and airfares in the us airline industry.
The Journal of Law and Economics, 50(2):355–380.
125
Kim, E. H. and Singal, V . (1993). Mergers and market power: Evidence from the airline industry.
The American Economic Review, pages 549–569.
Kwoka, J., Hearle, K., and Alepin, P. (2016). From the fringe to the forefront: Low cost carriers
and airline price determination. Review of Industrial Organization, 48(3):247–268.
Kwoka, J. and Shumilkina, E. (2010). The price effect of eliminating potential competition: Evi-
dence from an airline merger. The journal of industrial economics, 58(4):767–793.
Lewbel, A. (2007). Estimation of average treatment effects with misclassification. Econometrica,
75(2):537–551.
Luo, D. (2014). The price effects of the delta/northwest airline merger. Review of Industrial
Organization, 44(1):27–48.
Mahajan, A. (2006). Identification and estimation of regression models with misclassification.
Econometrica, 74(3):631–665.
Mayer, C. and Sinai, T. (2003). Network effects, congestion externalities, and air traffic delays: Or
why not all delays are evil. American Economic Review, 93(4):1194–1215.
Mazzeo, M. J. (2003). Competition and service quality in the us airline industry. Review of
industrial Organization, 22(4):275–296.
Morrison, S. A. (2001). Actual, adjacent, and potential competition estimating the full effect of
southwest airlines. Journal of Transport Economics and Policy (JTEP), 35(2):239–256.
Nguimkeu, P., Rosenman, R., and Tennekoon, V . (2021). Regression with a misclassified binary
regressor: Correcting for the hidden bias.
Pai, V . (2010). On the factors that affect airline flight frequency and aircraft size. Journal of air
Transport management, 16(4):169–177.
Park, J.-H. and Zhang, A. (2000). An empirical analysis of global airline alliances: cases in north
atlantic markets. Review of industrial organization, 16(4):367–384.
Peters, C. (2006). Evaluating the performance of merger simulation: Evidence from the us airline
industry. The Journal of law and economics, 49(2):627–649.
Prince, J. T. and Simon, D. H. (2015). Do incumbents improve service quality in response to entry?
evidence from airlines’ on-time performance. Management Science, 61(2):372–390.
Prince, J. T. and Simon, D. H. (2017). The impact of mergers on quality provision: Evidence from
the airline industry. The Journal of Industrial Economics, 65(2):336–362.
Roth, J., Sant’Anna, P. H., Bilinski, A., and Poe, J. (2022). What’s trending in difference-
in-differences? a synthesis of the recent econometrics literature. arXiv preprint
arXiv:2201.01194.
126
Russell, L. (2021). Price effects of nonprofit college and university mergers. Review of Economics
and Statistics, 103(1):88–101.
Shen, C. (2017). The effects of major us domestic airline code sharing and profit sharing rule.
Journal of Economics & Management Strategy, 26(3):590–609.
Sun, L. and Abraham, S. (2021). Estimating dynamic treatment effects in event studies with
heterogeneous treatment effects. Journal of Econometrics, 225(2):175–199.
Verboven, F. (1996). International price discrimination in the european car market. The RAND
Journal of Economics, pages 240–268.
Weinberg, M. C. and Hosken, D. (2013). Evidence on the accuracy of merger simulations. Review
of Economics and Statistics, 95(5):1584–1600.
Werden, G. J., Joskow, A. S., and Johnson, R. L. (1991). The effects of mergers on price and
output: Two case studies from the airline industry. managerial and Decision economics,
12(5):341–352.
Whalen, W. T. (2007). A panel data analysis of code-sharing, antitrust immunity, and open skies
treaties in international aviation markets. Review of industrial organization, 30(1):39–61.
127
Appendix A
Appendix to Chapter 2
A.1 Codeshare Remedy Details
When the Department of Justice (DOJ) approved the merger between Alaska Airlines and
Virgin America, the DOJ proposed modification of AS-AA codeshare conduct through the final
judgment.
1 2
In the final judgment, the DOJ clarified the market criteria ordering that the merging
parties and American Airlines should not market codeshare products with one another.
The first market criterion specified markets where Virgin America and American Airlines were
competing head-to-head as of December 6, 2016. The second market criterion specified city mar-
kets where Alaska Airlines and American Airlines were competing. These two market criteria pro-
hibited AS-AA codeshare products on Virgin/American overlap routes and Alaska/American over-
lap routes. In the final judgment, the DOJ defined 21 origin–destination pairs as “Virgin/American
Domestic U.S. Overlap Routes”
3
and 31 origin-destination pairs as “Alaska/American Domestic
U.S. Overlap Routes”
4
as of December 2016. The third market criterion specified markets that
1
The final judgment can be found at the following link: https://www.justice.gov/atr/case-document/
file/915971/download. Related documents can be found at https://www.justice.gov/atr/case/
us-v-alaska-air-group-inc-and-virgin-america-inc; U.S. V . ALASKA AIR GROUP, INC., AND VIR-
GIN AMERICA INC. (last accessed in March, 2022).
2
A press release from the Department of Justice regarding the final judg-
ment can be found at the following link: https://www.justice.gov/opa/pr/
justice-department-requires-alaska-airlines-significantly-scale-back-codeshare-agreement
3
See Table A1.
4
See Table A2.
128
originated or terminated at each key airport.
5
The fourth criterion specified codeshare flights oper-
ating between Los Angeles (LAX) and any key airports.
For these overlap routes and key airports, we counted the number of origin-destination airport
pair markets that fulfilled any of the above criteria for every quarter samples and plotted line graphs
of these “document remedy markets”, with figures shown on the right-hand axis (Figure A1-A6).
We also checked whether the merging parties and American Airlines had actually marketed
codeshare flights in these document remedy markets. Although our data set represents only 10%
of all tickets, checking the actual codeshare flights constitutes a useful way to evaluate the DOJ’s
remedy measures. We identified the flights within the document remedy markets that were also
codeshare flights between American Airlines and the merging parties. Then, we counted the num-
ber of markets with these flights and plotted bar graphs of “remedy markets with actual codeshare
flights”, with figures shown on the left-hand axis. We distinguished markets in three ways: mar-
kets with only AS/AA codeshares, markets with only AA/AS codeshares, and markets with both
AS/AA & AA/AS codeshares on—AS-AA markets.
6
We used dashed lines to indicate 2016Q4
(when the final judgment was proposed) and solid lines for 2017Q2 (when the final judgment was
revised and finalized).
We assessed the document remedy markets using the DOJ-specified criteria. For the first cri-
terion, we found that a small number of codeshare products were marketed prior to the merger
in the AS/VX overlap markets but eliminated immediately after the merger (Figure A1). For the
second criterion, we found that the number of AS/AA overlap markets significantly increased after
announcement of the merger, which may have increased the DOJ’s concerns regarding the compet-
itive harm of codeshare flights. With the final judgement, however, the number of AS/AA overlap
markets decreased and both airlines stopped using their codeshare flights in AS/AA overlap mar-
kets (Figure A2). In the markets with key AS airports, Alaska was steadily dependent on AS/AA
5
There are four key Alaska airports (PDX, SEA, SFO, and ANC) and 12 key American airports (CLT, MDW,
ORD, DFW, DAL, FLL, JFK, MIA, LGA, PHLM PHX, and DCA).
6
AS/AA represents codeshare products where Alaska is the ticketing carrier and American is the operating carrier
and AA/AS represents codeshare products where American is the ticketing carrier and Alaska is the operating carrier.
129
codeshare flights and never truly internalized the DOJ remedy measures (Figure A3).
7
Meanwhile,
in the markets with AA key airports, American mitigated its reliance on AA/AS codeshare flights
(Figure A4). In the markets connecting airlines’ key airports to LAX, Alaska and American mit-
igated their reliance on each other’s codeshare products (Figure A5). Overall, only some of the
document remedy markets featured codeshare products of interest, with American cutting its de-
pendency on AA/AS codeshare flights. Additionally, following the merger, American and Alaska
stopped selling their cross-codeshare flights in the same remedy markets—no AS-AA tra cs on
(Figure A6).
7
However, among 1,500 document remedy markets, AS/AA codeshare flights operated in less than 100 markets.
130
Table A1: Virgin/American Domestic U.S. Overlap Routes
Non-Directional Origin - Destination Pairs
Origin Destination
Boston Logan International Airport BOS LAX Los Angeles International Airport
Chicago O’Hare International Airport ORD LAX Los Angeles International Airport
Dallas Love Field Airport DAL LAX Los Angeles International Airport
Dallas/Fort Worth International Airport DFW LAX Los Angeles International Airport
Fort Lauderdale – Hollywood International Airport FLL LAX Los Angeles International Airport
Los Angeles International Airport LAX MIA Miami International Airport
Honolulu International Airport HNL LAX Los Angeles International Airport
McCarran International Airport LAS LAX Los Angeles International Airport
Los Angeles International Airport LAX IAD Washington Dulles International Airport
Los Angeles International Airport LAX DCA Ronald Regan Washington National Airport
Los Angeles International Airport LAX JFK John F. Kennedy International Airport
Los Angeles International Airport LAX EWR Newark Liberty International Airport
Los Angeles International Airport LAX MCO Orlando International Airport
Los Angeles International Airport LAX SEA Seattle – Tacoma International Airport
Dallas Love Field Airport DAL SFO San Francisco International Airport
Dallas/Fort Worth International Airport DFW SFO San Francisco International Airport
Fort Lauderdale – Hollywood International Airport FLL SFO San Francisco International Airport
Miami International Airport MIA SFO San Francisco International Airport
John F. Kennedy International Airport JFK SFO San Francisco International Airport
Los Angeles International Airport LAX SFO San Francisco International Airport
Chicago O’Hare International Airport ORD SFO San Francisco International Airport
Dallas Love Field Airport DAL DCA Ronald Regan Washington National Airport
Dallas/Fort Worth International Airport DFW DCA Ronald Regan Washington National Airport
Dallas Love Field Airport DAL LGA LaGuardia Airport
Dallas/Fort Worth International Airport DFW LGA LaGuardia Airport
Dallas Love Field Airport DAL LAS McCarran International Airport
Dallas/Fort Worth International Airport DFW LAS McCarran International Airport
Fort Lauderdale – Hollywood International Airport FLL JFK John F. Kennedy International Airport
Miami International Airport MIA JFK John F. Kennedy International Airport
Los Angeles International Airport LAX OGG Kahului Airport
McCarran International Airport LAS JFK John F. Kennedy International Airport
131
Table A2: Alaska/American Domestic U.S. Overlap Routes
Non-Directional Origin - Destination Pairs
Origin Destination
Ted Stevens Anchorage International Airport ANC LAX Los Angeles International Airport
Ted Stevens Anchorage International Airport ANC PHX Phoenix Sky Harbor International Airport
Chicago O’Hare International Airport ORD PDX Portland International Airport
Chicago O’Hare International Airport ORD SEA Seattle – Tacoma International Airport
Dallas/Fort Worth International Airport DFW PDX Portland International Airport
Dallas/Fort Worth International Airport DFW SEA Seattle – Tacoma International Airport
Los Angeles International Airport LAX PDX Portland International Airport
Los Angeles International Airport LAX SLC Salt Lake City International Airport
Los Angeles International Airport LAX SEA Seattle – Tacoma International Airport
John F. Kennedy International Airport JFK SEA Seattle – Tacoma International Airport
Philadelphia International Airport PHL SEA Seattle – Tacoma International Airport
Phoenix Sky Harbor International Airport PHX SEA Seattle – Tacoma International Airport
Phoenix Sky Harbor International Airport PHX PDX Portland International Airport
Ronald Regan Washington National Airport DCA LAX Los Angeles International Airport
Baltimore – Washington International Airport BWI LAX Los Angeles International Airport
Newark Liberty International Airport EWR SEA Seattle – Tacoma International Airport
John F. Kennedy International Airport JFK SAN San Diego International Airport
Newark Liberty International Airport EWR SAN San Diego International Airport
Miami International Airport MIA SEA Seattle – Tacoma International Airport
Fort Lauderdale–Hollywood International Airport FLL SEA Seattle – Tacoma International Airport
Washington Dulles International Airport IAD LAX Los Angeles International Airport
132
Figure A1: No. of Document Remedy 1 Markets and Those with Actual Codeshares
Figure A2: No. of Document Remedy 2 Markets and Those with Actual Codeshares
133
Figure A3: No. of Document Remedy 3 Markets With AS and Those with Actual Codeshares
Figure A4: No. of Document Remedy 3 Markets With AA and Those with Actual Codeshares
134
Figure A5: No. of Document Remedy 4 Markets and Those with Actual Codeshares
Figure A6: No. of All Document Remedy Markets and Those with Actual Codeshares
135
A.2 AS-AA Codeshare Details
Markets with codeshare products can be categorized into four groups based on the presence of
alternative pure online products offered by the codeshare partners. These four types are illustrated
in Figure A7. For all figures, (a) - (d), solid lines indicate flights operated by American (AA),
whereas dashed lines show flights operated by Alaska (AS). Figure (b) shows a traditional product
where the segment from the origin to a connecting airport is operated by American and the segment
from the connecting airport to the destination is operated by Alaska. If we assume that this product
is ticketed by Alaska (without loss of generality), this product represents an AS/AA traditional
codeshare product. In this case, no other pure online products are offered by either Alaska or
American in this market. In contrast, Figure (a) shows a market where competing pure online
products are offered by both Alaska and American along with the codeshare products, assuming
all nonstop flights are ticketed by the operating airlines (i.e., pure online products). Similarly,
Figure (c) shows a market where both AS/AA codeshare and pure online AA products are offered,
while Figure (d) shows a market where both AS/AA codeshare and pure online AS products are
offered. If we assume that Alaska is the ticketing carrier and American is the operating partner
of the codeshare products, Figure (c) represents a case where only the operating carrier offers a
competing pure online product, and Figure (d) represents a case where only the ticketing carrier
offers its own pure online product.
Because pricing of traditional codeshare products may depend on other products offered by the
codeshare partners in the market, we divided the relevant AS-AA codeshare markets based on the
above categories.
8
To distinguish the ticketing carrier and the operating airline partner, we divided
the relevant AS-AA codeshare markets into routes where AS/AA codeshare products were offered
and routes where AA/AS codeshare products were offered and determined how many routes were
classified as one of the four market types, 2014Q1 to 2016Q4.
8
Gayle (2013) shows that the margins of both the operating partner and the ticketing carrier persist when the
operating airline of a codeshare product offers competing pure online products on the relevant markets. In other
words, in the case of Figure (c), margins are not fully eliminated.
136
Figure A8 shows the numbers of each type of market for AS/AA codeshare markets. Note that
the operating airline was American (AA) in these markets. Among the total number of routes where
AS/AA codeshare products were operated, American also offered competing pure online products
for more than half of the routes. There were also some markets where both Alaska and American
offered pure online products. Among the four types, Neither markets are markets where neither
Alaska nor American offered pure online products (cf. Figure A7 (b)). These markets support the
network expansion effects of codeshare products, because without the codeshare products, these
markets would not be served by the relevant airlines.
For AA/AS codeshare routes, the proportions of each market type were similar to those of the
AS/AA markets, except the Neither type (Figure A9). Interestingly, there were no Neither type
markets among the AA/AS markets. In other words, unlike the AS/AA codeshare markets, none of
the AA/AS markets provided network expansion effects. This classification is likely to be helpful
for understanding the pricing of the relevant codeshare products, even though it is not used directly
to interpret the results of this study.
137
Figure A7: AS-AA Codshare Market Types
138
Figure A8: AS/AA Codeshare Markets
Figure A9: AA/AS Codeshare Markets
139

Abstract (if available)

Abstract This dissertation comprises three essays on the competition and antitrust issues in the airline industry, focusing on the competitive effects of airline consolidation and the behavior of low-cost carriers (LCCs).
The first chapter investigates the effects of the merger between two low-cost carriers, Southwest Airlines and AirTran Airways, and focuses on overlapping routes. To control factors that could affect the outcomes, two-way fixed effects difference-in-differences regressions are used on outcomes at the market-carrier-level. The results suggest that changes in price are about 5-6% higher for merging airlines than non-merging airlines, and there is no evidence of a relative increase in the quality of the merging airlines on the overlapping routes. The main takeaway from this paper is that, unlike the small price effects of mergers between legacy carriers, this merger between low-cost carriers appears to have significant anti-competitive effects on overlapping routes.
The second chapter analyzes the competitive effects of the merger between Alaska Airlines and Virgin America, and the impact of the Alaska-American codeshare discontinuation that resulted mainly from remedies by the Department of Justice (DOJ). The results from the descriptive merger analysis show that the merger was neither anti-competitive nor pro-competitive. Regarding codeshare exit effects, the findings from a two-way fixed difference-in-differences analysis and demand estimation indicate that the cessation of the codeshare agreement between Alaska and American worked to reduce prices on the relevant markets; meanwhile, the demand in these markets decreased.
In the last chapter, I examine the determinants of low-cost airlines’ service quality, focusing on on-time performance (OTP), using recently available low-cost carriers’ (LCCs) OTP data as well as other data for control variables. Although LCCs initiated their operations focusing on short- and medium-haul routes under a point-to-point system, the rapid expansion of these carriers has allowed them to operate under a hub-and-spoke system and to serve more connecting passengers through focus cities. The focus of this paper is whether departing from or arriving at a core airport affects an LCC's OTP, as is the case for full-service carriers. The findings show that departing from or arriving at a hub or a focus city airport tends to reduce departure and arrival delays of LCCs after controlling for several factors such as weather conditions, ages of aircraft, within-week trends, and scheduled times. This relationship can be understood as an internalization of delay externality because a carrier can suffer more from a delay on routes from or to a hub or a focus city airport.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Essays on competition between multiproduct firms

PDF

Essays on competition and strategy within platform industries

PDF

Vertical integration and two-sided market pricing: evidence from the video game industry

PDF

Empirical essays on industrial organization

PDF

Three essays on agent’s strategic behavior on online trading market

PDF

Cost structure and regulation in the telecommunications industry

PDF

Essays on competition for customer memberships

PDF

Three essays on industrial organization

PDF

Essays in the economics of education and conflict

PDF

Essays on family and labor economics

PDF

Essays on pricing and contracting

PDF

Essays on the microeconomic effects of taxation policies

PDF

The essays on the optimal information revelation, and multi-stop shopping

PDF

Essays on the platform design and information structure in the digital economy

PDF

Behavioral approaches to industrial organization

PDF

Essays on price determinants in the Los Angeles housing market

PDF

Essays on quality screening in two-sided markets

PDF

Essays in political economics

PDF

Essays on bundling and discounts

PDF

Essays in information economics and marketing

Asset Metadata

Creator Park, Hae Yeun (author)

Core Title Essays on competition and antitrust issues in the airline industry

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Economics

Degree Conferral Date 2022-08

Publication Date 07/24/2022

Defense Date 05/03/2022

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag airline industry,codeshare,merger,OAI-PMH Harvest

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Ridder, Geert (committee chair), Hovenkamp, Erik (committee member), Tan, Guofu (committee member)

Creator Email haeyeunp@usc.edu,haeyeunp12@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC111375389

Unique identifier UC111375389

Legacy Identifier etd-ParkHaeYeu-10949

Document Type Dissertation

Format application/pdf (imt)

Rights Park, Hae Yeun

Type texts

Source 20220728-usctheses-batch-962 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu