Changes in Airline Service Differ
Significantly for Smaller Communities,
but Limited Data on Ancillary Fees
Hinders Further Analysis
Report No. EC2020036
May 27, 2020
Changes in Airline Service Differ Significantly for Smaller
Communities, but Limited Data on Ancillary Fees
Hinders
Further Analysis
Self
-initiated
Office of the Secretary
| EC2020036 | May 27, 2020
What We Looked At
In 2013 and 2014, reports from the Government Accountability Office (GAO) and the Massachusetts
Institute of Technology (MIT) documented a disproportionate decline in commercial air service to
smaller communities. Since that time, there have been concerns that small- and medium-sized
communities continue to have limited access to the National Airspace System. The lack of a recent
analysis, as well as major changes in the industry, prompted our office to update the GAO and MIT
reports. Accordingly, our objective for this self-initiated audit was to detail recent trends in the
aviation industry, particularly as they relate to small- and medium-sized communities.
What We Found
Compared to larger metropolitan areas, smaller communities have experienced disparate effects from
several recent aviation industry trends. For example, departures declined in larger communities by
roughly 12 percent and in smaller communities by about 34 percent. Connectivity—the ability to
connect to and move throughout the national air system—declined by 16 percent in smaller
communities, double the rate in larger communities; however, data limitations hindered our analysis
of delays and cancellations. Similarly, competitive conditions improved in larger communities, but
grew worse in smaller communities, where the cost to fly was also greater. Finally, we found that some
airlines have dramatically increased their revenues from booking charges and other ancillary fees.
However, the Department of Transportation (DOT) does not collect adequate data on ancillary fees,
which reduces its ability to fully assess competition in the industry. Also, ancillary fees are not subject
to the excise tax that funds the Airport and Airway Trust Fund (AATF). We conservatively estimate that
certain carriers’ use of booking fees as a revenue source reduced AATF revenues by $60.6 million in
2019 alone.
Our Recommendations
We made three recommendations to address DOT’s data shortcomings and improve departmental
clarity on the impact of ancillary fees on AATF receipts. The Department concurred with one of our
three recommendations.
All OIG audit reports are available on our website at www.oig.dot.gov.
For inquiries about this report, please contact our Office of Government and Public Affairs at (202) 366-8751.
EC2020036
Contents
Memorandum 1
Result
s in Brief 3
Backgr
ound 4
Depart
ures Decreased Substantially System-Wide but Smaller
Communities Experienced the Greatest Percent Losses 8
Pass
enger Numbers Have Grown Through Increases in Seats and Load
Factors, Despite Departure Declines 13
Smalle
r Communities Lost the Most Connectivity to the National Airspace
System and Data Availability Limits Analysis of Delays and
Cancellations 18
Compet
itive Conditions Improved in Larger Communities but Worsened
in Smaller Communities 23
Flying From Smaller Communities Became Relatively More Expensive, but
Lack of Data on Growing Fees Hinders Analysis 34
Conc
lusion 45
Recom
mendations 45
Agency Comments and OIG Response 46
Acti
ons Required 48
Exhi
bit A. Scope and Methodology 49
Exhi
bit B. Organizations Visited or Contacted 65
Exhib
it C. List of Acronyms 66
Exhib
it D. Major Contributors to This Report 67
Exhi
bit E. Categorization of Select Airlines 68
Exhibit F. List of Multi-Airport PSAs 70
Appe
ndix. Agency Comments 72
EC2020036 1
U.S. DEPARTMENT OF TRANSPORTATION
OFFICE OF INSPECTOR GENERAL
Memorandum
Date: May 27, 2020
Subject: ACTION: Changes in Airline Service Differ Significantly for Smaller Communities,
but Limited Data on Ancillary Fees Hinders Further Analysis |
Report No. EC2020036
From:
Charles A. Ward
Assistant Inspector General for Audit Operations and Special Reviews
To: Assistant Secretary for Aviation and International Affairs
Director of the Bureau of Transportation Statistics
A community’s ability to develop economically is impacted by its connections
with other communities and ability to transport people quickly and regularly. In
2013 and 2014, the Government Accountability Office (GAO) and the
Massachusetts Institute of Technology’s (MIT) International Center for Air
Transportation (ICAT) released a series of reports documenting a
disproportionate decline in commercial air service to smaller communities—
relative to large communities—between 2007 and 2013.
1
When accounting for
service changes affecting smaller communities, GAO and MIT researchers cited
higher fuel costs, reduced demand, demographic changes, industry
consolidation, and capacity discipline.
2
The GAO a
nd MIT reports predated a decline in jet fuel prices
3
in late 2014 and
may not have included the full impact of recent airline mergers. For example, the
final judgment in the merger between US Airways Group, Inc. and AMR
Corporation was issued by the Department of Justice (DOJ) in April 2014, the
firms did not integrate their reservation systems until October 2015. However, the
GAO and MIT analyses only used data through 2013.
1
For example, GAO, Status of Air Service to Small Communities and the Federal Programs Involved (GAO-14-454T),
April 2014 and MIT ICAT, Trends and Market Forces Shaping Small Community Air Service in the United States (ICAT-
2013-02), May 2013.
2
The losses airlines incurred in the late 2000s—in part due to the economic recession and historically high jet fuel
prices—contributed to changes in airlines’ business models. In an effort to cut costs, airlines transitioned to a
capacity-discipline strategy. This strategy reduced seating capacity by offering fewer flights, while reducing the share
of unfilled seats on flights.
3
The per-gallon price fell from $2.73 in September 2014 to $1.50 in January 2015.
EC2020036 2
Despite the airline industry’s profitability since these reports were issued, there
were concerns that many communities’ ability to access the National Airspace
System has not subsequently improved. In particular, these concerns have
focused on airline service to small- and medium-sized communities. For example,
the potential economic impact of this decline in air service received congressional
attention, and the Federal Aviation Administration (FAA) Reauthorization of 2016
authorized the Secretary of the Department of Transportation (DOT) to establish
a Working Group on Improving Air Service to Small Communities.
The lack of recent analysis, airlines’ financial recovery during the past few years,
and the completion of major airline mergers have prompted our office to update
the earlier GAO and MIT analyses to better inform the ongoing policy debate
regarding service to smaller communities. Accordingly, our objective for this self-
initiated audit was to detail recent aviation industry trends, particularly as they
relate to service to small- and medium-sized communities.
4
Specifically, we detail
trends in airline service levels; numbers of passengers flown; airline service
quality, including connectivity; airline competition; and prices paid by airline
passengers for airfare and ancillary services—particularly as they relate to small-
and medium-sized communities.
To meet the objective, we analyzed U.S. Census Bureau (Census) and DOT
datasets that highlighted changes in activity, competition, prices, and service
quality from 2005 through 2017.
5
Because we found that some fees—which are
not included in the base ticket price—have grown considerably, we compiled
information on certain fees through November 2019 in order to account for this
trend. We reviewed airline industry research conducted by Government agencies,
academic economists, and transportation researchers, with a focus on articles
that analyzed competitive practices and service to smaller communities. To better
understand the industry’s considerations in serving smaller communities, we
interviewed representatives from Airlines for America, the Regional Airline
Association, and the Air Line Pilots Association. We also contacted GAO to
discuss their previous research on ancillary fees. We conducted this performance
audit in accordance with generally accepted Government auditing standards.
We appreciate the courtesies and cooperation of DOT representatives during this
audit. If you have any questions concerning this report, please call me at
(202) 366-1249 or Betty Krier, Chief Economist, at (202) 366-1422.
cc: The Secretary
DOT Audit Liaison, M-1
FAA Audit Liaison, AAE-100
4
We included a second audit objective when we announced this review. In a subsequent review, we will provide a
descriptive analysis of factors associated with changes in airline service to small- and medium-sized communities.
5
We started with 2005 so that our baseline would be unaffected by the recession that began in 2008. Our analyses of
ticket prices and competition used data beginning in 2006.
EC2020036 3
Results in Brief
In comparison to larger metropolitan areas, smaller
communities have experienced disparate effects of several
recent aviation industry trends.
• Departures decreased substantially system-wide
6
but smaller communities
experienced the greatest percentage losses. While departures declined in
larger communities by roughly 12 percent, departures dropped about
34 percent in smaller communities. Further, small communities without
Essential Air Service (EAS)
7
saw an even larger decline.
• Passenger numbers have increased through growth in seats per flight and
load factors. The number of seats per flight and passenger load factors
had the largest percentage growth in smaller communities, by more than
35 percent and 12 percentage points, respectively. Still, the total number
of seats fell significantly in smaller communities.
• Smaller communities lost the most connectivity to the National Airspace
System, and data limitations hinder analysis of delays and cancellations.
Connectivity—a measure of a passenger’s ability to easily connect to and
move throughout the national air system—declined among smaller
communities by 16 percent, twice as much as the 8 percent decline in
connectivity among larger communities. Differences in cancellations and
delays by community size appear modest, but coverage of smaller
community service quality was limited until 2018.
• Competitive conditions improved in larger communities, but worsened in
smaller communities. While competition increased on routes originating
from larger communities due primarily to non-legacy carriers entering
these routes, it declined for smaller communities. Further, the price
premium associated with flying from a smaller community—compared
with taking similar flights from a large community—has risen in recent
years.
• Ancillary fee revenue has grown significantly, which may degrade the
quality of DOT’s airline revenue and ticket price data and decrease Airport
6
In this report, the term “system-wide” refers to passenger flights between airports in the contiguous United States.
7
EAS is a DOT program that was put into place following the Airline Deregulation Act of 1978 to guarantee that small
communities that were served prior to deregulation maintain a minimal level of scheduled air service.
EC2020036 4
and Airway Trust Fund (AATF)
8
receipts. Also, information we gathered
while conducting our audit shows that some airlines’ revenues from
ancillary fees, such as booking fees, have grown dramatically. However,
while DOT collects data on airline revenues and ticket prices, it does not
collect adequate data on ancillary fees. Without this information, DOT and
the traveling public may not know the impact of these fees on the costs
to passengers and airline revenues for air service from smaller–or larger–
communities. This may reduce DOT’s ability to assess competitive
conditions in the industry. Using available information, we determined
that certain carriers’ use of booking fees as a revenue source can be
conservatively estimated to reduce AATF revenues by $60.6 million in
2019 alone.
We are making recommendations to address DOT’s data shortcomings and
improve departmental clarity on the impact of ancillary fees on AATF receipts.
Background
DOT’s Office of Aviation Analysis initiates and supports the development of
DOT’s public policies regarding economic oversight of the airline industry. The
Office of Aviation Analysis analyzes and supports DOT’s decision makers on
major airline issues, including mergers and acquisitions, joint venture agreements
and immunized international alliances between U.S. and foreign carriers, and
airline distribution practices. Additionally, the Office of Aviation Analysis
administers the EAS program and its Competition and Policy Analysis division
monitors changes in the industry, analyzes industry trends – including
assessments of airline fares, and evaluates policy options on a wide range of
issues. DOT’s Bureau of Transportation Statistics (BTS) publishes data and
statistics on commercial aviation, which includes data on airfares, air carrier
traffic, and airlines’ financial data. This data is used by analysts within and outside
DOT, and in our report we rely heavily on data published by BTS.
In our analysis, we defined communities using Census criteria and DOT
information on EAS subsidies. We also categorized airlines into two primary
groups—mainline and regional—as well as divided mainline and regional carriers
into subgroups. All of our analysis focused exclusively on airline service between
communities in the contiguous United States.
9
The following describes our
8
The AATF was created under the Airport and Airway Revenue Act of 1970 to provide a dedicated source of funding
for the U.S. aviation system, independent of the General Fund.
9
We also restricted our analysis to airports which had at least 2,500 enplanements on scheduled passenger flights in
at least one year between 2005 and 2017. Throughout this report, airport refers to an airport which met this
enplanement threshold.
EC2020036 5
criteria for identifying communities and community size categories, EAS
communities, and airline categories.
Defining Communities and Community Size
Groups
We determined community boundaries using Census statistical area definitions.
Communities were defined as either a county or set of counties. A set of counties
was considered a single community when Census determined that they were
significantly economically and socially integrated, see exhibit A for further details
on our community definitions. This resulted in a number of communities that
contain multiple airports, see exhibit F.
We categorized communities into five size groups—large (L), medium-large (ML),
medium (M), medium-small (MS), small (S)—such that the combined population
of all communities within a size group was approximately 20 percent of the
population of the contiguous United States in 2010. Throughout the report, when
we use the term “smaller” to describe communities, we are referring to both small
and medium-small communities. Also, we use the term “larger” to refer to both
medium-large and large communities. Table 1 below provides information on our
community size categories and their population statistics.
EC2020036 6
Table 1. Community Size Groups
Category
a
Number of
Communities
Total
Population
Percent of
Total U.S.
Population
b
Median Size
Community and
Its Population
c
Population Range
Large 5
68.0
million
22.4
Chicago-
Naperville, IL-IN-
WI;
9.84 million
8.15 million to 23.1 million
Medium-Large 10
57.5
million
19.0
Houston-The
Woodlands, TX;
6.11 million
3.68 million to 7.89 million
Medium 26 59.5
million
19.6 Salt Lake City-
Provo-Orem, UT;
2.27 million
1.46 million to 3.52 million
Medium-Small 83 60.5
million
20.0 Jackson-
Vicksburg-
Brookhaven, MS;
660,368
374,536 to 1.41 million
Small with
commercial
service & no
EAS service
134
20.4
million
6.7
Mesa County,
CO;
146,723
5,172 to 373,802
Small with EAS-
subsidized
commercial
service
102
7.8
million
2.6
Franklin County,
NY;
51,599
7,369 to 279,771
Small, without
commercial
service
1,215 29.6
million
9.7 Winn Parish, LA;
15,313
614 to 331,298
a
Community statistics including EAS community information are for 2010.
b
Total U.S. population is for the contiguous U.S.
c
The median is the middle value and is less sensitive to outliers than an average. There are as many
communities larger than the median size community as there are communities smaller than it. The
median population column reports the higher of the central two communities when there is an
even number of communities.
Source: OIG analysis of Census and DOT data
Essential Air Service
The Airline Deregulation Act of 1978 granted airlines the freedom to determine
which routes they serve. This included granting carriers the ability to terminate
airline service to any community without Government approval—raising concerns
EC2020036 7
that communities with relatively low traffic could lose service entirely. To address
these concerns, the EAS program was established to ensure that small
communities retain a link to the National Air Space System. Service at EAS
communities is typically maintained by giving an air carrier a direct subsidy to
provide flights between the EAS community and a medium or large hub airport,
where passengers can connect to the national network.
Throughout this report, we define an EAS community as a small community with
service subsidized under the EAS program in at least one quarter of a given year.
That is, if a small community received an EAS subsidy in at least one quarter of a
particular year, then we consider it to be an EAS community in that year. As of
May 2018, there were 108 active EAS contracts.
10
In 2018, the yearly cost of the
subsidies was $285.8 million, for an average of $2.65 million per EAS contract.
Categorization of Air Carriers
To understand air carriers’ roles in serving smaller communities, we categorized
them into broad groups based on a few components of their operations. At the
highest level, provision of air services is divided between mainline carriers and
regional carriers. A mainline carrier—such as Delta Air Lines or Frontier Airlines—
is often the carrier that sells the ticket for an air travel itinerary, also known as the
marketing carrier.
11
Also, mainline carriers often operate the associated flights,
particularly on long-distance flights and flights using larger aircraft. In other
cases, the mainline carrier markets a flight, while a regional carrier—such as
SkyWest Airlines or Air Wisconsin Airlines—operates the flight under contract
with the mainline carrier.
We further categorized mainline carriers as either legacy or non-legacy carriers.
Legacy carriers operated routes prior to passage of the Airline Deregulation Act
of 1978. Six of these airlines remained in operation in 2005—American Airlines,
Continental Airlines, Delta Air Lines, Northwest Airlines, United Airlines, and US
Airways.
12
Airlines that began operation after deregulation are considered non-
legacy carriers.
10
This figure includes some contracts which were awarded to airports which were not in a small community. EAS
contracts which lie outside a small community are included in the cost figures presented in this paragraph, but
otherwise are not defined as EAS communities in the report.
11
Marketing carriers may sell tickets through direct channels such as the carrier’s webpage, or through third party
distributors, such as online travel agents.
12
Alaska Airlines and Southwest Airlines operated prior to deregulation, but neither had a significant network in the
contiguous United States at that time. Because our analysis focuses on service within the contiguous United States,
we do not define them as legacy carriers for the purposes of this report.
EC2020036 8
We then divided non-legacy carriers into two categories. Low-cost carriers
(LCCs)—such as Southwest and JetBlue—typically achieve lower costs than legacy
carriers. Ultra-low-cost carriers (ULCCs)—such as Allegiant Air and Spirit
Airlines—achieve even lower costs than LCCs. ULCCs are also distinct from LCCs
and legacy carriers in their reliance on ancillary fees for a significantly greater
share of their revenue than the other mainline carriers.
Lastly, we categorized regional carriers based on their ownership status, as some
regional carriers are held by a mainline carrier’s holding company while others
are independently-owned. The latter are referred to in this report as independent
regional carriers. Regional carriers that are held by a mainline carrier’s holding
company are referred to as “non-independent.” Exhibit E lists carriers by category.
Figure 1 below depicts the categorization of airlines used throughout this report.
Figure 1. Categorization of Air Carriers
Source: OIG generated
Departures Decreased Substantially System-Wide
but Smaller Communities Experienced the Greatest
Percent Losses
Departures have fallen in every community size group since 2005 but smaller
communities experienced the greatest percent declines. Reductions in flight
frequencies accounted for most of the system-wide decline but reductions in
nonstop destinations served also contributed substantially. Small non-EAS
communities experienced much larger reductions in both departures and
nonstop destinations than small EAS communities.
EC2020036 9
Smaller Communities on Average Had
the Greatest Percent Reductions in
Departures
From 2005 through 2017, the number of system-wide passenger flights fell by
19.2 percent. However, the percent change in departures varied considerably
across different community size groups. On average, larger communities lost
roughly one-tenth of their departures, medium-sized communities slightly less
than one quarter, and smaller communities approximately one-third. Notably,
excluding EAS communities, the decline in departures from small communities
was even larger, about 40 percent on average. Figure 2 below is a line chart that
depicts the percent changes in departures by community size.
Figure 2. Percent Change in Departures by Community Size
Note: Excludes EAS communities. Baseline is 2005.
Source: OIG analysis of DOT data
In addition to large average declines across different community size groups,
certain communities experienced significant reductions in departures during this
period. For example, five medium-sized communities—Cincinnati, OH, which saw
a 77.1 percent decline; Pittsburgh, PA; Greensboro, NC; Cleveland, OH; and
Milwaukee, WI—lost over half of their departures, often resulting from an airline
shifting the focus of its network away from an airport after a merger. Despite
such dramatic changes, the median
percent decline in departures—a measure
EC2020036 10
that is not sensitive to outliers—for each community size group was similar to its
total percent decline. Table 2 below shows the median percent change in
departures that occurred for the different community size groups.
Table 2. Median Percent Declines in Departures
Community Size
Group
Median Percent Decline
in Departures
Large 5.9
Medium-Large 18.0
Medium 21.8
Medium-Small 31.3
Small 32.3
Source: OIG analysis
Flight Frequency Reductions Accounted
for Most Departure Declines but
Destination Losses Were Also Sizeable
Reductions in flight frequencies to a destination accounted for 69.7 percent of
declines in departures while reductions in the number of nonstop destinations
13
served accounted for 30.3 percent. Although all community size groups
experienced a decline in flight frequencies, only medium and smaller
communities lost nonstop destinations. Nevertheless, while large communities
saw substantial growth in the number of nonstop destinations served, their total
departures still fell because of substantial losses in flight frequency.
Reductions in flight frequencies accounted for the majority of departure declines
in all except the medium and medium-small communities. Average flight
13
In this section, we focused exclusively on daily nonstop destinations, which we consider to be particularly important
for non-leisure travel. On average, daily destinations accounted for 95.5 percent of departures from small
communities and 99.6 percent of departures from large communities in 2017. We define “daily” destinations as those
that have over 250 flights per year and connect an origin and a destination community. This requires an average of at
least one flight per weekday, while allowing for a small number of cancellations. Note that this definition does not
differentiate between carriers. If Delta offered once-daily service on a route from January to June before exiting the
market, then United entered the market and offered once-daily service on this route from July to December, the route
is coded as a daily route. In addition, this definition includes seasonal routes as long as they accumulate over 250
flights in the year.
EC2020036 11
frequency
14
fell 19.0 percent in large communities, as shown by the decline from
15.3 to 12.4 in the figure below. It fell by 23.5 percent in small communities,
including EAS communities, and by 23.7 percent when EAS communities are
excluded. The comparable figure for medium-large communities showed a
15.7 percent drop. Figure 3 is a bar chart that depicts flight frequencies over time
by community size.
Figure 3. Average Daily Flights on Daily Routes
Note: Daily routes offer at least 250 flights in a given year.
Source: OIG analysis of DOT data
In contrast to reductions in flight frequencies as shown above, larger
communities experienced increases in total daily destinations served while
smaller and medium-sized communities saw significant declines between 2005
and 2017. Specifically, large communities saw a 12.5 percent average increase in
daily destinations and medium-large communities saw a 3.7 percent average
increase. Medium and medium-small communities had declines in daily
destinations of 17.1 and 21.0 percent on average respectively. Small communities’
daily destinations fell by only 5.7 percent when EAS communities were included,
which is not surprising given that 118 of the 236 small communities were EAS
communities at some point from 2005 through 2017. Figure 4 below is a line
14
Average flight frequency is the unweighted average of flight frequencies across all daily routes within each
community size group.
EC2020036 12
chart that depicts the percent change in daily nonstop destinations by
community size.
Figure 4. Percent Change in Daily Non-Stop Destinations by
Community Size
Note: Excludes EAS communities. Baseline is 2005.
Source: OIG analysis of DOT data
However, when the sample is restricted to non-EAS communities, small
communities lost service to 20.8 percent of their daily destinations on average.
Concurrently, the proportion of small community service accounted for by EAS
flights has grown from 12.7 to 20.1 percent. Figure 5 is a line chart that depicts
the percent changes in small community flights by whether or not they are
subsidized by the EAS program.
EC2020036 13
Figure 5. Percent Change in Small Community Departures: EAS-
Subsidized vs. Unsubsidized
Note: Baseline is 2005.
Source: OIG analysis of DOT data
In summary, large and medium-large communities saw an increase in the number
of nonstop daily destinations on average, but a decrease in flight frequency on
average to those destinations. Medium and medium-small communities saw a
large decline in nonstop destinations, but those destinations that remained saw a
modest decline in frequency. Among small communities, the median community
neither gained nor lost destinations when EAS communities are included, but the
median small non-EAS community experienced a significant decline in daily
destinations. Small communities also saw a considerable decline in flight
frequency.
Passenger Numbers Have Grown Through Increases
in Seats and Load Factors, Despite Departure
Declines
Despite departure declines, system-wide passenger numbers grew between 2005
and 2017. Only medium-small communities experienced a decline in passenger
EC2020036 14
numbers. In other community size groups, increases in seats per flight and load
factors were sufficiently large to offset departure declines. Both of these increases
were largest in percent terms in small communities. Still, the total number of
seats fell significantly only in smaller communities.
Passenger Numbers in 2017 Exceeded
2005 Levels for Most Community Size
Groups
From 2005 through 2017, the number of passengers flown by air carriers grew
from 636 million to 711 million—an 11.8 percent increase. Small communities
saw an increase of 9.6 percent in the number of passengers flown, while medium-
small communities were the only community size group with a decline in
passengers between 2005 and 2017. However, even in medium-small
communities, the number of passengers was down by only 2.5 percent overall
and has risen every year since 2013. Figure 6 is a line chart that depicts the
percent change in passengers by community size.
Figure 6. Percent Change in Passengers by Community Size
Note: Baseline is 2005.
Source: OIG analysis of DOT data
EC2020036 15
The data shows that airlines were able to increase the number of passengers
carried, despite the decline in departures, through increasing the number of
passengers per flight. The greatest growth in the average number of passengers
per flight—62.4 percent—occurred on flights originating in small communities,
which carried an average of 26.6 passengers in 2005 and 43.2 in 2017.
The Number of Seats Per Flight Grew
Substantially but Total Seats Still Fell
Significantly in Smaller Communities
System-wide, seating capacity was 0.8 percent lower in 2017 than it was in 2005.
This is a markedly smaller decline than the nearly 20 percent reduction in flights
during this same period. This was largely a result of airlines’ upgauging since
2005. Upgauging involves the airline changing the aircraft they use to models
with higher seating capacities. Notably, airlines also increased the quantity of
seats within aircraft models during this time. Figure 7 is a bar chart that shows the
growth in seats on several airplane models.
Figure 7. Average Seating Capacity of Select Aircraft
Note: MD-80 includes the MD-81, MD-82, MD-83, and MD-88.
Source: OIG analysis of DOT data
EC2020036 16
The average seating capacity per flight has increased in all community size
groups. In absolute terms, medium-sized communities saw the greatest increase
in average seating capacity, with the number of seats per flight rising from 96.6 in
2005 to 124.6 in 2017, a 29.0 percent gain. The other four community size groups
all saw similar absolute increases in average seating capacity, but this represented
the highest percent change—35.4 percent—for small communities. Notably,
average seating capacity changes in small communities were even greater when
only considering non-EAS communities—42.7 percent. In comparison, the
percent increases in average seating capacity for large and medium-large
communities were 14.8 percent and 16.2 percent, respectively. Figure 8 is a bar
chart depicting average seats per flight by community size.
Figure 8. Average Seats per Flight by Community Size
Note: Excludes EAS communities.
Source: OIG analysis of DOT data
However, the marked growth in seats per flight did not fully offset the impact of
departure declines on airline seating capacity in smaller communities. The total
number of seats fell 8.7 percent in small communities, and 14.2 percent in
medium-small communities. In contrast, total seating capacity in large
communities grew 4.8 percent, and fell by just 0.3 percent in medium-large
communities. The minor decline in system-wide seating capacity was the net
result of these different changes in larger and smaller communities. Figure 9 is a
line chart showing changes in total seats by community size.
EC2020036 17
Figure 9. Percent Change in Total Seats by Community Size
Note: Excludes EAS communities. Baseline is 2005.
Source: OIG analysis of DOT data
Load Factors Increased in All Community
Size Groups
Higher load factors—defined as the ratio of passengers to seats on a plane—
enabled the significant growth in passenger numbers despite the slight decline in
system-wide seating capacity. Apart from small communities, the different
community size groups’ load factors in 2005 were between 69.9 and 73.6 percent,
and grew 8.4 to 9.6 percentage points by 2017. In contrast, small communities
had significantly lower load factors in 2005 than other community size groups,
but experienced the highest load factor growth rate, particularly in non-EAS
communities—12.0 percentage points. Figure 10 is a bar chart that shows that
load factors increased significantly for all five community groups.
EC2020036 18
Figure 10. Load Factor Percent by Community Size
Note: Excludes EAS communities.
Source: OIG analysis of DOT data
Smaller Communities Lost the Most Connectivity to
the National Airspace System and Data Availability
Limits Analysis of Delays and Cancellations
Connectivity—a measure of a passenger’s ability to easily connect to and move
throughout the National Airspace System—declined across all community size
groups from 2005 to 2017. However, the average decline in connectivity among
smaller communities was twice as big as in larger communities. Data on delays
and cancellations were not reported for a large share of flights in smaller
communities during this time, but the available data showed modest differences
in delays and cancellations across community sizes.
EC2020036 19
Smaller Communities’ Connectivity Has
Declined Twice as Much as Larger
Communities’ Since 2005
Between 2005 and 2017, the average connectivity loss in smaller communities
was about twice as high as in larger communities, 16.3 and 7.8 percent
respectively. Further, every community size group lost connectivity between 2005
and 2017. However, the average decline for the medium-sized communities was
only 1.4 percent. Figure 11 is a line chart that shows the percent change in
average connectivity by community size.
Figure 11. Percent Change in Average Connectivity by Community
Size
Note: Baseline is 2005.
Source: OIG analysis of DOT data
Our connectivity calculations were based on the Airport Connectivity Quality
Index (ACQI), developed by researchers at MIT’s International Center for Air
Transportation. The ACQI accounts for: the number of nonstop and connecting
destinations, with connecting destinations receiving less weight; the frequency of
available scheduled flights to the nonstop destinations; and the quality of a
destination, as a proxy for economic, social, cultural, and political importance. The
ACQI captures the quality of an airport destination by assigning weights to
EC2020036 20
airports based on their FAA airport hub type designation.
15
This means that a
flight to a large city or a major connecting hub is weighted more heavily than a
flight to a smaller community with limited connecting options. Instead of
calculating the connectivity by airport, we calculated it by community, community
size group, and system-wide, see exhibit A for connectivity calculation details.
Our calculations show that the National Airspace System experienced a large
decline in connectivity from 2007 to 2010. Subsequently, the average connectivity
of smaller communities continued to decline, while the average connectivity from
other communities stabilized or grew. For all but small communities, the average
connectivity score bottomed out in 2014 and subsequently began to increase.
Further, from 2005 through 2017, smaller individual communities were far more
likely than larger individual communities to undergo a significant decline in
connectivity. During this time, nearly half of smaller communities saw their
connectivity decline by more than 20 percent, while fewer than one in seven
larger communities saw their connectivity decline by more than 20 percent.
Differences in Cancellations and On-Time
Performance Across Community Sizes are
Modest, but Data Limitations Hinder
Analysis
We found that passengers flying to smaller communities were just as likely to
have their flights canceled, but less likely to be delayed, as those flying to larger
communities. Also, the average delay of late arriving flights in smaller
communities increased over time to nearly the same level as in the larger
communities by 2017. However, lack of data coverage may limit the
representativeness of these conclusions for smaller communities—as the services
provided by carriers that fell below a revenue threshold were not reported (e.g.,
Allegiant Air and Air Wisconsin Airlines). Specifically, prior to 2018, FAA required
carriers with more than 1.0 percent of total domestic scheduled passenger carrier
revenues to report flight delays and cancellations. Consequently, over the period
of our analysis, the average proportion of flights with service quality data
16
was
only 60.0 percent for smaller communities in comparison to 81.6 percent for
larger communities. Importantly, starting in 2018 FAA reduced this reporting
threshold to 0.5 percent. This change brought the share of flights with service
15
FAA classifies airports’ hub type as Large, Medium, Small, or Nonhub, based on annual passenger enplanements.
16
FAA’s Airline Service Quality Performance database is the primary source for information on airline delays and
cancellations.
EC2020036 21
quality data for 2018 to 92 percent in the smaller communities and to nearly
100 percent in the larger communities.
The available data show that, from 2005–2017, flight cancellation percentages
generally declined and were lowest for passengers flying to medium-sized
communities. While the cancellation percentages for both small and large
communities declined by 2017, both experienced increased cancellation
percentages in 2011. Figure 12 is a bar chart showing cancellation percentages by
community size.
Figure 12. Cancellation Percentage by Community Size
Source: OIG analysis of DOT data
On-time performance—the percentage of flights no more than 15 minutes late—
to all communities generally improved by 2017, although large communities
experienced deteriorating performance between 2011 and 2017.
17
Figure 13 is a
bar chart showing on-time performance percentage by community size.
17
Some of the improvement in on-time performance may have resulted from increased schedule padding by airlines.
For more details, see Dennis Zhang, Yuval Salant, and Jan A. Van Mieghem, Where Did the Time Go? On the Increase
in Airline Schedule Padding Over 21 Years (August 24, 2018).
EC2020036 22
Figure 13. On-Time Performance Percentage by Community Size
Source: OIG analysis of DOT data
While passengers arriving late to smaller communities experienced shorter delays
than those arriving late to larger communities, that difference narrowed by 2017,
as the length of delays in smaller communities worsened. Lateness—the minutes
of delay for flights that arrived later than 15 minutes after their scheduled arrival
time—for passengers flying into smaller communities was 5 minutes lower than
in larger communities in 2005; by 2017, the difference decreased to 1 minute. The
downward trend was the result of increasing lateness at smaller communities,
from approximately 48 minutes in 2005 to 59 minutes in 2017. Figure 14 below is
a bar chart showing average minutes of delay for delayed flights by community
size.
EC2020036 23
Figure 14. Average Minutes of Delay for Delayed Flights by
Community Size
Source: OIG analysis of DOT data
Competitive Conditions Improved in Larger
Communities but Worsened in Smaller
Communities
Domestic airline services consolidated substantially from 2006 to 2017. This
occurred within both the mainline and regional segments. However, since 2006,
different community size groups have experienced diverging outcomes.
Competition increased on routes from larger communities, but declined on
routes from smaller communities. Accounting for part of this difference is that
non-legacy carrier service in larger communities expanded more than in smaller
communities. Lastly, non-legacy carriers differ substantially in their strategies for
serving smaller communities.
EC2020036 24
Both the Mainline and Regional Airline
Industry Segments Underwent
Substantial Consolidation
Mainline carriers and regional carriers divide the provision of commercial airline
services between them, and both industry segments have become substantially
more consolidated.
18
The share of passengers purchasing tickets from the four
largest mainline carriers has risen considerably—from 58 percent in 2006 to
79 percent in 2017. Similar changes occurred in the regional segment during this
time. The four largest regional airline holding companies combined to carry
55 percent of all passengers flying on a regional carrier in 2006. By 2017, this
figure rose to 76 percent. Even greater consolidation occurred among the subset
of regional holding companies that are independent. Specifically, the four largest
independent regional airline holding companies combined to carry 66 percent of
all passengers flying on an independent regional carrier in 2006, and this figure
rose to 94 percent by 2017.
While the passenger share of the four largest firms in each segment illustrates the
scale of consolidation among larger firms, it offers a limited image of how each
segment evolved. The industry has restructured considerably since 2000. Legacy
airlines struggled financially for much of 2000 through 2010, and underwent a
series of major mergers from 2005 through 2013. Of the six legacy airlines
operating in 2005, only three remained in 2017. Also during this period, LCCs
expanded and ULCCs grew dramatically. For example, LCC JetBlue’s passenger
share rose from 3.8 percent in 2006 to 5.4 percent in 2017. The combined
passenger share of ULCCs Allegiant Air, Frontier Airlines, and Spirit Airlines rose
from 3.0 percent in 2006 to 9.5 percent in 2017.
Additionally, non-legacy carriers’ passenger share rose from 35.4 percent in 2006
to 45.9 percent in 2017. The entry of non-legacy carriers into new routes has
been cited by regulatory agencies and researchers as a means to promote
competition—in light of legacy carriers’ consolidation. For example, the DOJ
ruled that US Airways Group and AMR Corporation could merge under the
subsequently formed American Airlines Group. However, they were also required
to offer 26 slots
19
to non-legacy carriers—16 at Reagan National Airport to
JetBlue Airways, Inc. and 10 slots at LaGuardia Airport to Southwest Airlines, Inc.
18
We conducted our analysis of competition of mainline carriers throughout this section using marketing carriers.
Because regional carriers primarily operate flights marketed by mainline carriers, we computed measures of regional
market structure using operating carriers.
19
A slot is an authorization to either take-off or land at a particular airport on a particular day during a specific time
period. In addition to divestitures at Reagan National Airport and LaGuardia Airport, the ruling also required
EC2020036 25
The regional segment also underwent significant restructuring over this
timeframe. We conducted our analysis of regional airlines using airline holding
companies rather than individual airlines in scenarios where multiple airlines were
held by the same company. Regional airline holding companies often own
multiple regional airlines. Some of the increase in regional concentration can be
traced to merging mainline carriers’ subsidiaries falling under the same holding
company after the mainline partners merged. For example, the regional
subsidiaries of US Airways Group and AMR Corporation were each placed under
the newly formed American Airlines Group, Inc. after the two companies merged.
However, changes in the structure of the regional airline industry did not result
entirely from mergers of mainline carrier holding companies. For example, the
independent regional holding company SkyWest, Inc. acquired two large
independent regional airlines—Atlantic Southeast Airlines, Inc. in 2005 and
ExpressJet Airlines, Inc. in 2010. In 2005, SkyWest, Inc. carried 18.2 percent of
passengers flying on independent regional airlines, while ExpressJet carried
13.7 percent of passengers, and Atlantic Southeast carried 8.4 percent of
passengers. By 2012, the SkyWest, Inc. holding company carried 46.8 percent of
passengers flying on independent regional airlines.
To better measure the changes in airline industry concentration, we use the
Hirschmann-Herfindahl Index (HHI). This is a standard measure of industry
concentration used by DOJ and the Federal Trade Commission (FTC). The HHI is
calculated as the sum of the squared value of the passenger share of each
airline.
20
The HHI ranges from zero to 10,000, and a greater HHI corresponds to a
more concentrated market. DOJ and FTC generally classify markets as
unconcentrated if the HHI falls below 1,500; moderately concentrated if the HHI
lies between 1,500 and 2,500; and highly concentrated if the HHI lies above 2,500.
The maximum value of 10,000 indicates a monopoly.
The changes in the HHI from 2006 through 2017 shown in the figure below
indicate that every industry segment underwent a sizeable degree of
consolidation. The increase was greater among the regional segment (859 points)
and the independent regional segment (1,714 points) than it was among the
mainline segment (550 points). Based on the DOJ and FTC classification, both the
mainline and regional markets were unconcentrated in 2006, but became
moderately concentrated by 2017. The independent regional market was likewise
divestiture of gates at Boston Logan International Airport, Chicago O’Hare International Airport, Dallas Love Field, Los
Angeles International Airport, and Miami International Airport. Research has shown that these divestitures improved
gate access of non-legacy carriers and resulted in lower airfares on routes with forced divestitures. For more details,
see: Zhou Zhang, Federico Ciliberto, and Jonathan Williams, “Effects of Mergers and Divestitures on Airline Fares,”
Transportation Research Record: Journal of the Transportation Research Board, vol. 2603, no. 1 (2017), pp. 98-104.
20
Letting
represent firm j’s share of all passenger enplanements in a given year, = 10,000
. The index
can alternatively be computed using revenue rather than numbers of passengers.
EC2020036 26
unconcentrated in 2006 and became highly concentrated by 2017. Figure 15 is a
bar chart with the HHI by market segment for 2006, 2011, and 2017.
Figure 15. HHI by Market Segment
Source: OIG analysis of DOT data
Regional carriers play a critical role in service to smaller communities, as
passengers in these communities are more likely to be served by the regional
carriers than passengers in larger communities. From 2005 through 2017,
regional carriers flew 75 percent of passengers in small communities and over
40 percent of passengers in medium-small communities, as compared to around
20 percent of passengers in larger communities.
Economists and other researchers have studied the relationship between mainline
competition and outcomes such as prices and service quality, but we are unaware
of any study that examines the possible impacts of regional consolidation. For
example, whether regional consolidation can impact ticket prices by impacting
contract negotiations with their mainline partners is unknown.
21
21
For example, Millou and Petrakis study mergers in the upstream sectors of vertically related industries, focusing on
the relationship between contract types and market structure. Economists have referred to the mainline and regional
airlines as vertically related, with the mainline carriers representing downstream firms and regional carriers
representing upstream firms. For more, see Chrysovalantou Millou and Emmanuel Petrakis, “Upstream horizontal
mergers, vertical contracts, and bargaining,” International Journal of Industrial Organization, vol. 25, no. 5 (2007),
pp. 963–987.
EC2020036 27
Route-Level Competition Increased in
Larger Communities and Declined in
Smaller Communities
Passengers flying from smaller communities had fewer carriers to choose from
when purchasing tickets in 2017 than in 2006. During this time, the average
number of effective competitors
22
—which are holding companies that sold at
least 5 percent of tickets between an origin and destination in the year—for a
passenger flying from a small community fell from 2.66 to 2.51. The average
number of effective competitors for a passenger flying from a medium-small
community fell from 3.73 to 3.33. In contrast, passengers flying from larger
communities had more carriers to choose between in 2017 than in 2006. During
this time, the average number of effective competitors for a passenger flying
from a medium-large community rose from 3.73 to 4.03, while the average
number of effective competitors for a passenger flying from a large community
grew from 4.26 to 4.56.
While useful, the measure of effective competitors does not account for the
relative size of the competitors. The HHI provides more information than the
average number of effective competitors, as it depends on both the number of
competitors and the difference in competitors’ size. Given two markets with the
same number of competitors, the HHI will be lower—indicating stronger
competition—in the market with more similarly-sized competitors. For example,
consider a route that initially has one competitor and consequently, 100
= 10,000.
If a second competitor begins serving this route and captures 10 percent of the
market, the HHI would fall to 90
+ 10
= 8,200. If a second competitor begins
serving this route and captures 40 percent of the market, the HHI would fall to
60
+ 40
= 5,200. Each scenario brings one additional competitor, but the lower
HHI indicates the latter has a greater effect on competition.
We find that the divergence in competitive conditions between smaller and larger
communities is also present when the HHI—rather than effective competitors—is
used to measure competition. The HHI in small and medium-small communities
rose between 2006 and 2017, indicating a decline in competition. In contrast, the
HHI in large and medium-large communities fell during the same period,
22
Effective competitors and the HHI are defined based on the marketing carrier, and weighted by the number of
passengers on the route. We included both direct and indirect itineraries between an origin and destination because
airlines which offer direct flights also compete with airlines offering indirect flights. In 2017, 74.5 percent of
passengers flew direct, 24.5 percent of passengers made one stop, 0.9 percent of passengers made two stops, and
0.1 percent of passengers made at least three stops.
EC2020036 28
indicating an increase in competition. Figure 16 is a bar chart showing route-level
HHI by community size for 2006, 2011, and 2017.
Figure 16. Average Route-Level HHI by Community Size
Note: Average HHI is calculated weighting by the number of passengers on each
route in a given year.
Source: OIG analysis of DOT data
Expansion of Non-Legacy Carriers Was
Substantial in Larger Communities, but
Modest in Smaller Communities
The divergence in competition on routes serving smaller communities, in
comparison to larger communities, can be partially explained by differences in
the expansion of non-legacy carriers across different community size groups. The
average number of legacy carriers competing on a route fell significantly across
all community size groups from 2006 to 2017. At the same time, non-legacy
carriers substantially expanded their presence in medium and larger
communities, but their expansion in smaller communities was comparatively
minor.
23
23
All measures throughout this section are defined based on the marketing carrier.
EC2020036 29
Compared to legacy carriers, non-legacy carriers draw a lesser share of their
passengers from smaller communities. Further, the share of non-legacy carriers’
passengers originating in smaller communities has declined. In 2006, legacy
carriers drew 18.0 percent of their passengers from smaller communities, while
non-legacy carriers drew 14.2 percent. By 2017, the proportion of legacy carriers’
passengers originating in smaller communities increased slightly to 18.3 percent,
while that of non-legacy carriers’ fell to 11.7 percent. Table 3 displays the share of
passengers drawn from each of the community size groups for legacy and non-
legacy carriers in 2006 and 2017.
Table 3. Percent of Passengers by Community Size for Legacy and
Non-Legacy Carriers
Carrier Type
Year
Small
Medium-
Small
Medium
Medium-
Large
Large
Legacy 2006 4.1% 13.9% 27.6% 27.4% 26.9%
2017 4.7% 13.6% 26.0% 28.1% 27.7%
Non-Legacy 2006 2.1% 12.1% 34.8% 23.7% 27.4%
2017 1.7% 10.0% 36.5% 24.6% 27.3%
Source: OIG analysis of DOT data
One factor that could explain the differential patterns of non-legacy carrier
service across different community sizes is their network structure. A simple
breakdown of airline network structures may categorize networks as either hub-
and-spoke or point-to-point. Hub-and-spoke networks are characterized by the
presence of a central hub and several spokes branching out from the hub.
Passengers in a hub-and-spoke network are transported between different points
on the network through the central hub. Point-to-point networks do not have a
central hub and passengers are transported directly between different points on
the network. Modern airline networks are most accurately characterized as a
hybrid between hub-and-spoke and point-to-point networks. Figure 17 is a
graphical representation of the two airline network structures from this simple
characterization.
EC2020036 30
Figure 17. Comparison of Airline Networks
Source: OIG-generated
Non-legacy carriers’ networks are more similar than those of legacy carriers’ to
the point-to-point network.
24
Compared to hub-and-spoke networks point-to-
point networks have features that can make it difficult for the carrier to serve
smaller communities. They typically require high-density markets, allowing
carriers to operate routes at a low average cost per passenger. In addition, they
are better suited to carriers which operate a more limited set of aircraft. This
means the carrier may not operate smaller aircraft, which are better suited for
serving smaller communities.
Non-legacy carriers had notably different patterns of network expansion in larger
communities than in smaller communities. These carriers substantially expanded
their networks in medium-sized and larger communities from 2006 to 2017. For
example, the average passenger could choose between 1.48 non-legacy
competitors in 2006 in large communities. This rose to 2.17 non-legacy
competitors by 2017. By comparison, non-legacy carriers only increased their
presence in smaller communities to a minor extent.
The number of legacy competitors declined across all community sizes. For
example, the average number of legacy competitors serving a route in a medium-
small community fell from 2.63 in 2006 to 2.08 in 2017, a decline of 0.55.
Medium-sized communities saw a similarly large decline in legacy competitors,
while larger communities saw a somewhat smaller—but still significant—decline.
The smallest decline in legacy competitors occurred in small communities. Figure
18 is a bar chart that shows the change in effective competitors across
24
In 2017, over 30 percent of passengers on each of the three legacy carriers made a connection on the same carrier.
Among non-legacy carriers, Southwest had the greatest share of passengers (20 percent) connect to another
Southwest flight. Alaska Airlines (12 percent), Frontier Airlines (6 percent), and Sun Country (5 percent) had modest
shares of connecting passengers. The three remaining carriers—Spirit, JetBlue, and Allegiant—had a share of
connecting passengers below 5 percent.
EC2020036 31
community sizes broken out by changes in legacy competitors and non-legacy
competitors.
Figure 18. Change in Effective Competitors, 2006–2017
Note: Number of effective competitors is calculated weighting by the number of
passengers on each route in each year.
Source: OIG analysis of DOT data
Overall, as shown in the figure, in larger communities the robust expansion of
non-legacy carriers more than counteracted the decline in legacy carriers. Despite
national consolidation, passengers departing from larger communities could
choose between more carriers in 2017 than in 2006. The number of legacy
competitors also fell in smaller communities, and the number of non-legacy
competitors rose. However, the magnitude of entry by non-legacy carriers was
not as large as the magnitude of exit by legacy carriers. As a result, passengers
flying from smaller communities had fewer carriers to choose between in 2017
than in 2006.
EC2020036 32
Non-Legacy Carriers Differ Substantially
in Their Strategies for Serving Smaller
Communities
Although non-legacy carriers as a whole showed limited expansion into smaller
communities from 2006 through 2017, these carriers differed notably in their
strategies for serving smaller communities. In particular, Alaska Airlines and
Allegiant Air offer significantly more service to small communities than other
non-legacy carriers.
All of the seven non-legacy carriers drew a smaller share of their passengers from
smaller communities in 2017 than in 2006.
25
Further, five of the seven non-legacy
carriers drew less than 1 percent of their passengers from small communities in
2017. The other two carriers—Alaska Airlines and Allegiant Air—differ from the
other five in ways that help explain their greater propensity to serve passengers
in small communities. Table 4 below shows the percent of passengers drawn
from each community size group in 2006 and 2017 for the seven largest active
non-legacy carriers.
25
We restrict this discussion to the non-legacy carriers which were active in both 2006 and 2017.
EC2020036 33
Table 4. Percent of Passengers by Community Size for Non-Legacy
Carriers
Carrier Type
Year
Small
Medium-
Small
Medium
Medium-
Large
Large
Alaska 2006 7.0% 8.7% 24.2% 32.8% 27.2%
2017 5.8% 7.6% 26.3% 32.9% 27.4%
Allegiant 2006 15.6% 27.6% 51.2% 1.7% 3.9%
2017 11.8% 29.5% 42.6% 9.2% 6.9%
Frontier 2006 0.6% 9.2% 59.3% 15.5% 15.5%
2017 0.5% 6.2% 60.7% 18.0% 14.6%
JetBlue 2006 1.0% 9.0% 17.3% 22.7% 50.0%
2017 0.6% 8.3% 21.3% 30.8% 39.1%
Southwest 2006 1.1% 12.0% 39.0% 19.8% 28.1%
2017 0.7% 10.4% 40.8% 21.4% 26.8%
Spirit 2006 0.0% 10.9% 15.6% 52.7% 20.8%
2017 0.3% 7.3% 26.9% 41.2% 24.3%
Sun Country 2006 0.8% 5.9% 19.5% 61.3% 12.6%
2017 0.1% 4.3% 18.0% 61.2% 16.4%
Source: OIG analysis of DOT data
Unlike the other six non-legacy carriers, Alaska Airlines sells tickets for flights that
are operated by its own regional subsidiary, Horizon Air, as well as by other
regional partners. In 2017, none of the seven non-legacy carriers operated aircraft
with fewer than 100 seats. However, Horizon Air’s fleet was composed entirely of
76-seat aircraft at that time. This enabled Alaska Airlines to serve small
communities—which may not have sufficient demand to fill larger aircraft—
through Horizon and its regional partners.
26
In 2017, there were six small
communities—four in Washington and two in Northern California—where Alaska
Airlines was the marketing carrier for at least 85 percent of passengers.
There were stark differences between the three ULCC’s service to passengers in
smaller communities. In 2017, Frontier Airlines and Spirit Airlines drew 6.7 and
7.6 percent of their passengers from smaller communities, respectively, while
Allegiant Air drew 41.3 percent. Although Allegiant’s passenger share on all
flights in 2017 was just 2.4 percent, it was 8.6 percent on flights from small
communities and 5.9 percent on flights from medium-small communities. During
26
Alaska Airlines has also marketed flights that were operated by independent regional carriers.
EC2020036 34
that year, Allegiant was present at a total of 81 smaller communities. In 34 of
these communities, its passenger share exceeded 20 percent. Further, in 12 of
these communities Allegiant had the highest passenger share of any airline, and
in 5 of these communities its passenger share was at least 90 percent.
Allegiant built its business around offering infrequent service from smaller
communities to leisure destinations. Passengers attempting to use Allegiant to
reach destinations not served directly by the airline may face difficulties for a few
reasons. First, Allegiant’s routes are often low frequency. For example, in 2017,
more than half of its routes flew three times or fewer per week. Thus, same-day
connections to other Allegiant flights may not have been available. Second,
Allegiant’s operations tend to be seasonal. For example, in 2017, Allegiant had
64.3 percent more departures in July than in September. Third, Allegiant has
based a significant share of their service to mid-sized and larger metropolitan
areas at secondary airports. For instance, Allegiant’s operations in the Orlando,
FL, area are based out of Orlando Sanford International Airport (SFB), while the
community’s primary airport is Orlando International Airport (MCO).
27
Passengers
seeking to connect from an Allegiant flight to almost any other carrier’s service
would need to exit SFB, drive over 30 miles to MCO, and pass through MCO
security screening.
Flying From Smaller Communities Became
Relatively More Expensive, but Lack of Data on
Growing Fees Hinders Analysis
Passengers flying from smaller communities’ pay a price premium, and this
premium has risen significantly in recent years. However, our analysis was limited
by a lack of information on ancillary fees. Certain fees have grown dramatically in
recent years, but are not reflected in DOT data on prices or ancillary fee revenue.
This lack of data could impact the Department’s understanding of both the costs
to consumers and airline industry competition. It could also impact
understanding of the effect on tax receipts supporting the Airport and Airway
Trust Fund (AATF) of airlines’ increased reliance on ancillary fees. In particular, we
conservatively estimate that airlines’ use of booking fees for purchasing tickets on
their websites may reduce AATF excise tax revenue by $60.6 million in 2019
alone.
27
In 2017, Allegiant accounted for 97.9 percent of departures from SFB but had no departures from MCO. Other
carriers had 130,461 departures at MCO compared to 189 departures at SFB.
EC2020036 35
Flying From Smaller Communities
Became Relatively More Expensive
From 2006 through 2017, passengers flying from smaller communities paid a
significant price premium, compared to passengers on similar flights in large
communities. Passengers flying roundtrip from small communities were
estimated to have paid a 21 percent premium in 2005, which rose to 27 percent
in 2017. The premium for medium-small communities rose from 8.5 percent to
15.6 percent over the same period. In contrast, passengers flying from medium
and medium-large communities consistently paid similar prices to passengers in
large communities.
From 2008 to 2010, the price premium associated with flying out of smaller
communities fell to a relative low point, and then fluctuated between 2011 and
2014. However, since 2014, the price premium paid by passengers from smaller
communities has increased steadily, surpassing 2005 levels. Figure 19 is a line
graph showing the percent price premiums by community size. The baseline for
our calculation of these price premiums is large community prices.
Figure 19. Price Premium Percent by Community Size
Note: Baseline is the large community price
Source: OIG analysis of DOT data
We focused on the price premium—the percentage difference between prices
paid in large versus other community size groups—because jet fuel prices varied
EC2020036 36
considerably from 2006 through 2017, and they are a significant component of
airline costs. We estimated these price premiums using quality-adjusted price
indices for the different community size groups. The quality factors accounted for
included: the number of seats per aircraft type; circuity or ratio of miles flown to
miles between the origin and destination, which accounts for the directness of
flights; the distance between communities; the number of trip segments; and the
carriers marketing the flights. See exhibit A for details on our price premium
estimation.
We calculated the prices using the DOT database reporting airfares—the Airline
Origin and Destination Survey (DB1B)—with Government and airport charges
removed.
28
We adjusted the reported fares to include the average ancillary fees—
baggage and change/cancellation fees—on which DOT collects revenue data
through its Form 41 P-1.2. Revenue information associated with other ancillary
fees is not identifiable given current reporting requirements and ancillary fees are
not included in the reported airfares. For example, the booking fee charged by
ULCCs for reservations made online or over the phone is likely incurred by the
vast majority of passengers but is not included in the DB1B. As a result, fares
listed in the DB1B for ULCCs are likely significantly lower than passengers’ cost of
purchasing tickets—even if the passenger does not add ancillary services outside
of the booking fee. If the Department tracked such ancillary fees, it would
improve the accuracy of its information regarding the cost of air travel to
passengers.
Limited Data on Ancillary Fees Could
Limit DOT’s Ability To Oversee Airlines’
Competitive Practices
Lack of data on many ancillary fees and their associated revenue could hinder
DOT’s oversight of the airline industry. Effective economic oversight by the
Department is important to ensure the efficiency of our transportation system.
Airlines’ pricing of ancillary services is also an important dimension of airline
competition. However, DOT does not collect data on these prices. This lack of
information could pose challenges to the Department’s understanding of
competitive practices in the industry.
Pricing of ancillary services is an important consideration for antitrust authorities
evaluating prospective mergers in the airline industry. DOJ raised concerns over a
prospective increase in ancillary fees in its complaints filed against the two most
recent mainline carrier mergers. In its complaint filed against the proposed
28
We obtained this version of the DB1B, the Superset, from Airline Data Inc.
EC2020036 37
merger between US Airways and American Airlines, DOJ stated “…industry
consolidation has left fewer, more-similar airlines, making it easier for the
remaining airlines to raise prices, impose new or higher baggage and other
ancillary fees, and reduce capacity and service.” In this complaint, DOJ stated that
even a modest increase in ancillary fees could cost consumers millions.
29
Likewise,
DOJ’s complaint filed against the proposed merger between Virgin Atlantic and
Alaska Airlines stated that the merger would likely result in higher fees.
30
Consequently, airlines’ offerings and pricing of ancillary services represent an
important aspect of competition in the industry. For example, Alaska Airlines
notes that fee pricing is a significant competitive factor in the industry.
31
Also,
growth of ULCCs may exert competitive pressure on mainline carriers, which
influences their product offerings. For example, in 2017, American Airlines
introduced its Basic Economy product to compete with ULCCs.
32
Limited information on the prices paid by passengers for ancillary services could
hamper DOT’s ability to provide adequate information on the flying public’s cost
of air transportation between different communities. The Wendell H. Ford
Aviation Investment and Reform Act
33
requires covered airports to produce a
written competition plan to gain approval for passenger facility charges (PFC) and
as a condition of certain grants. Airports’ competition plans are required to
incorporate information on airfares and how they compare to airfares at other
airports, using DOT data. The Department also releases a quarterly report that
provides information on airfares across city-pair markets. Air carriers differ
substantially in terms of the airports and routes they serve, as well as the share of
revenue they earn from fees for ancillary services. As a result, reported airfares
may closely approximate passengers’ full cost of flying from some communities,
but understate it for communities served by carriers that draw substantial
revenue from ancillary fees.
Without supplementary data on ancillary fees and their associated revenues, the
Department’s airfare data also may not accurately capture changes in the cost of
air travel to the public over time. From 2010 to 2018, airlines introduced new fees
for ancillary services such as seat selection and online booking. If the average
charge incurred by passengers for such ancillary services has risen, comparing
29
Amended Complaint, U.S., et al. v. US Airways Group, Inc., et al., 38 F.Supp.3d 69, No.13-cv-1236 (D.D.C. 2014)
30
Complaint, U.S. v. Alaska Air Group, Inc., et al., No. 16-cv-02377 (D.D.C. June 23, 2017) (unpublished).
31
Alaska Air Group Inc., 2017 Form 10-K, (2018).
32
American Airlines Group Inc., 2017 Form 10-K, (2018).
33
P.L. 106-181, section 155. Covered airports include any commercial service airport that has more than 0.25 percent
of the total number of passenger boardings each year at all such airports and where one or two air carriers control
over 50 percent of passenger boardings.
EC2020036 38
airfares over time may not accurately convey changes in passengers’ cost of
flying over time.
34
Increases in ancillary fees may cause the cost of flying to change, even if airfares
remain the same. For example, we queried Spirt Airlines’ website in both March
and August of 2019. For each query, we selected a round-trip itinerary from
Baltimore/Washington Thurgood Marshall International Airport (BWI) to Boston
Logan International Airport (BOS).
35
For the March query, the total cost to a
purchaser was $106.60. For the August query, the total cost was $112.60. A
$6 increase—from $39.98 to $45.98—in the booking fee was the only component
of the total price that changed. Table 5 below displays the results of the queries.
34
For example, Airlines For America’s webpage lists the average domestic round-trip airfare in the United States over
time. They present both a “Base Fare” as well as an “All-In Fare”. The latter incorporates the average baggage and
change/cancellation fees using data from DOT’s Form 41 Schedule P-1.2. This data shows that while the average
baggage and change/cancellation fees increased from $5.88 to $23.47 from 2007 through 2009, the average baggage
and change/cancellation fees declined slightly from $23.47 to $21.85 from 2009 through 2018. However, because this
data does not account for any other ancillary fees such as seat selection or booking fees, it does not completely
represent the change in costs incurred by passengers from 2009 through 2018.
35
The outbound leg for each query was Flight 1028, which was scheduled to depart BWI around 6 a.m. and arrive in
BOS around 7:30 a.m. The inbound leg for each query was Flight 1027, which was scheduled to depart BOS around
10 p.m. and arrive at BWI around 11:30 p.m. The queried itineraries do not include any ancillary services other than
the booking fee.
EC2020036 39
Table 5. Example of Price Components for Travel on Spirit Airlines, March and
August 2019
Price Type Price Component Query 3/8/2019 Query 8/21/2019 Difference
Total Round Trip
Price
All $106.60 $112.60 $6.00
Flight Price Flight $0.02 $0.02 -
Regulatory Compliance
Charge
$13.02 $13.02 -
Fuel Charge $22.32 $22.32 -
Booking Fee
a
$39.98 $45.98 $6.00
Total $75.34 $81.34 $6.00
Government
Fees and Taxes
Security Fee $11.20 $11.20 -
Segment Fee $8.40 $8.40 -
Passenger Facility Fee $9.00 $9.00 -
Federal Excise Tax $2.66 $2.66 -
Total $31.26 $31.26 -
a
Spirit Airlines refers to this as “Passenger Usage Charge.”
Source: Queries from Spirit Airlines’ webpage on 3/8/2019 and 8/21/2019
In addition, information obtained from Securities and Exchange Commission
(SEC) filings and carriers’ webpages indicates that booking fees have increased
considerably in recent years.
36
For example, by September 2019, typical per
segment online booking fees included: $22.99 for Spirit Airlines, $21 for Frontier
Airlines,
37
and $18 for Allegiant Air. In comparison, per-segment online booking
fees were: $5 for Spirit Airlines in 2010, $0 for Frontier Airlines until 2015, and
$13 for Allegiant in 2018.
GAO has noted that steps are needed to address the limited availability of data
on ancillary fees. In response to a 2010 GAO recommendation,
38
DOT issued a
Notice of Proposed Rulemaking on July 15, 2011, which required carriers to
36
Allegiant Air refers to their booking fee as an “Electronic Carrier Usage Charge.” Frontier Airlines refers to their
booking fee as a “Carrier Interface Charge.” Spirit Airlines refers to their booking fee as a “Passenger Usage Fee.”
37
Frontier Airlines does not list the level of this fee on their webpage. However, online queries typically showed a fee
of $21 per segment, with a lower fee appearing on some discounted itineraries.
38
GAO, Consumers Could Benefit from Better Information about Airline-Imposed Fees and Refundability of Government-
Imposed Taxes and Fees (GAO-10-785), July 2010.
EC2020036 40
report revenues on 19 separate charges for ancillary services.
39
However, DOT
withdrew the proposed rule on December 14, 2017, citing concerns about the
potential reporting burden on the industry—while acknowledging there would be
benefits of collecting and publishing this information.
40
E
ffective economic oversight of competitive practices in the airline industry is
critical to ensuring the efficiency of our transportation system. Airlines’ ancillary
service pricing strategies now represent an important aspect of airline
competition. As a result, the Department’s lack of data on ancillary fees could
hinder the Office of Aviation Analysis’s ability to effectively inform the
Department on issues related to airline competition.
Increased Reliance on Ancillary Fees
Could Impact Airport and Airway Trust
Fund Receipts
Unlike domestic airfares, fees charged by airlines for many ancillary services are
not subject to the 7.5 percent excise tax on transportation of persons by air.
Revenue collected from this excise tax constitutes an important funding source
for the Airport and Airway Trust Fund (AATF). As a result, increased reliance on
ancillary fees, as opposed to revenue from airfare, could result in diminished
AATF receipts.
41
One type of fee in particular—booking fees charged by some
carriers for purchasing tickets through the carrier’s webpage or call center—may
result in foregone AATF revenues of $60.6 million in 2019.
Baggage revenue is the only ancillary revenue that is both identifiable from DOT
data and not subject to the 7.5 percent excise tax. In a 2017 report, GAO used this
information to estimate that an additional $309 million in excise taxes would have
been credited to the AATF in 2016 had baggage fees been subject to the tax.
Because DOT does not separately record revenues associated with other ancillary
fees, it is difficult to determine the scale of foregone AATF receipts that could
39
Federal Register 76-136 (July 2011), pp. 41726-41731. The categories were (1) Booking fees, (2) priority check-in
and security screening, (3) baggage, (4) in-flight medical equipment, (5) in-flight entertainment/internet access,
(6) sleep sets, (7) in-flight food/non-alcoholic drinks, (8) alcoholic drinks, (9) pets, (10) seating assignments,
(11) reservation cancellation and change fees, (12) charges for lost ticket, (13) unaccompanied minor/passenger
assistance fee, (14) frequent flyer points/points acceleration, (15) commissions on travel packages, (16) travel
insurance, (17) duty-free and retail sales, (18) one-time access to lounges, and (19) other.
40
Federal Register 82-239 (December 2017), pp. 58777–58778.
41
Section 122 of the FAA Reauthorization Act of 2018 mandated that the Secretary of Transportation commission an
organization to conduct a study that includes an analysis of airlines’ ancillary fees and their impact on taxable
revenue. The report was released in January 2020 and estimates the impact of baggage fees on excise tax revenues.
The report also recommends that Congress include ancillary fees in the domestic passenger ticket tax. For more
details, see RAND Corporation, U.S. Airport Infrastructure Funding and Financing, 2020.
EC2020036 41
result from airlines’ reliance on ancillary fees more generally. However, the limited
data available from public filings suggests foregone receipts from other ancillary
fees may be significant and increasing.
Public filings indicate seat selection fees represent a significant and growing
revenue source for some carriers. For example, in 2018 JetBlue reported
$274 million in revenue from its “Even More Space” upgrade, a 14.0 percent
increase from 2017. Spirit Airlines reported $180 million in revenue from seat
selection in 2018, a 36.7 percent increase from 2017. While some other carriers
also charge seat selection or seat upgrade fees, we were unable to find
associated revenue levels in public filings. Further, it is not possible to determine
seat selection fees or revenues using DOT data.
42
The limited information available also appears to indicate that booking fee
revenues are sizeable and growing. Spirit Airlines is the only ULCC that has
publicly reported its booking fee revenue, which has grown steadily since 2009. In
2018, it reached over $531 million. Figure 20 is a line graph of Spirit Airlines’
booking fee revenue from 2009 through 2018.
Figure 20. Spirit Airlines’ Booking Fee Revenue ($Millions)
Source: OIG generated from information in Spirit Airlines’ SEC filings
42
Specifically, we are referring to seat selection and upgrade fees. Ticket class upgrades, on the other hand, are
included as part of the airfare and so are included in the prices recorded in the DB1B.
EC2020036 42
More specifically, booking fees can represent a large share of the total amount
paid by purchasers to the air carrier for a ticket, which could significantly limit—
or eliminate entirely—the amount of excise tax collected. For example, as shown
in figure 21, on September 3, 2019, we found Spirit Airlines
43
offering a round trip
between Baltimore and Boston for $64.60 total; $36 represented charges
collected by Spirit Airlines, and $28.60 represented charges collected for
government or airport purposes.
44
However, the $36 collected by Spirit Airlines is
further broken down into two components: $0.02 for the “Flight” and $35.98 for
the “Passenger Usage Charge.” If the airline only collects the 7.5 percent excise
tax on the Flight component, the carrier would not collect any excise taxes on this
$.02 itinerary. Figure 21 is a screenshot of the query described above. Note that
there is no reference here to Federal Excise Tax, which can be seen in our earlier
example from table 5 above.
43
For additional examples, on February 17, 2020, we found Allegiant Air offering a roundtrip itinerary that charged
$1.12 for the flights, a $36 booking fee, and $0.08 in excise taxes. On February 17, 2020, we also found Frontier
Airlines offering a roundtrip itinerary that charged $0.87 for the flights, a $42 booking fee, and $0.06 in excise taxes.
44
The Security Fee, Segment Fee, and Passenger Facility fee listed here are charged based on the number of
segments, and are not affected by the price of the itinerary.
EC2020036 43
Figure 21: Spirit Airlines Itinerary Listing $.02 Flight Cost and
No Collection of Excise Tax
Source: Screenshot from query on Spirit Airlines’ webpage from September 3,
2019
Based on Allegiant Air and Frontier Airlines’ passenger enplanements and the
share of their passengers incurring a booking fee, as well as Spirit Airlines’
reported booking fee revenue, the combined revenue from booking fees earned
by all three ULCCs in 2018 may have been roughly $1 billion. Further, revenue
earned on booking fees could keep increasing if:
• ULCCs continue to increase the share of revenue they earn from booking
fees,
• ULCC growth continues to outpace the industry as a whole, or
• Additional carriers begin charging passengers for booking tickets online.
EC2020036 44
Notably, ULCC online booking fees differ from other ancillary service fees in
several ways. First, consumers are automatically opted-in to the booking fee
when they use ULCCs’ websites to book a ticket—unlike optional fees such as
baggage fees and seat selection fees. To opt out of this fee, consumers must
purchase their ticket at the airport.
45
However, service counters selling the tickets
may have limited hours.
46
Second, when consumers opt out of the fee by
purchasing a ticket at the airport, they may not be offered the same price for
baggage as when booking online. For example, as of June 17, 2019, Allegiant Air
typically charged $18 to $25 per direction for carry-on bags purchased while
booking online; $45 after booking, but prior to departure; and $50 at the airport.
Third, many ancillary services, such as in-flight meals, increase carrier costs.
Because online distribution is likely the ULCCs’ least costly form of distribution,
ULCCs’ cost of distributing tickets is lower for passengers who choose to book
tickets online than for those who book at the airport.
We spoke with Internal Revenue Service (IRS) officials on September 13, 2019, to
discuss the tax treatment of airlines’ booking fees. Specifically, we asked about
booking fees associated with purchasing tickets on the carrier’s webpage or over
the phone. IRS officials stated that they have not made a ruling on the taxability
of such booking fees.
We estimate that booking fees charged by ULCCs may result in foregone AATF
revenue between $60.6 million and $74.5 million in 2019. We computed these
figures using information on each of the ULCCs’ enplanements over the past
12 months, the range of booking fees typically charged by the carriers, and the
share of passengers who purchase a ticket through the carrier’s webpage or call
center. In both estimates, we assume that no foregone revenue is associated with
tickets purchased through a third-party channel or the carrier’s ticket counter.
The difference between the estimates arises from the range in booking fees
charged. If carriers charge a range of booking fees, the high-end estimate
assumes that all passengers purchasing through a direct channel pay the higher
fee—which available information indicates is the typical fee charged. Our low-end
estimate makes the more conservative assumption that half of passengers
purchasing through a direct channel pay the discounted booking fee, whereas
the other half pay the higher fee.
45
Two of the three ULCCs—Frontier Airlines and Spirit Airlines—sell some of their tickets through third-parties, such
as Expedia. Examining a small number of itineraries on these carriers’ webpages and Expedia’s website, we found that
bookings made through Expedia on 07/08/2019 were between $6 and $15 more expensive than on the carriers’
webpages. So, it does not appear that consumers using this third party could have avoided the cost of the booking
fee.
46
For example, as of 6/19/2019, Allegiant’s ticket counter at Orlando Sanford International Airport —Allegiant’s
busiest airport—was open on Wednesdays, Thursdays, and Fridays from 9:00 a.m. to 11:00 a.m. Allegiant’s webpage
notes that airport ticket purchases are typically available for one hour following each scheduled departure.
EC2020036 45
Taxability of ancillary services—including booking fees—lies outside the
mandated authority of the Department. However, the level of foregone revenue
that may result from carriers’ use of booking fees could have a notable impact on
the AATF.
Conclusion
The structure of the airline industry transformed considerably from 2005 through
2017. During this period, the characteristics of airline service to all community
size groups have also evolved. Small- and medium-sized communities have
experienced the greatest percent changes according to a range of measures.
EAS-subsidized service now accounts for a greater share of small community
flights. Further, the impacts of airline industry dynamics underscore the
continuing need for the Department to collect and analyze adequate data to
accurately capture the industry’s effects on all communities and travelers.
Recommendations
To enhance the Department’s analytical and advisory capability with respect to
monitoring the cost of airline service to the flying public, we recommend that:
1. The Bureau of Transportation Statistics issue a Reporting Directive
clarifying that air carriers are to include booking fees, along with any/all
fees required to board the aircraft, in the fare line item reported to the
Office of Airline Information’s Origin and Destination Survey.
To improve the Department’s ability to assess competitive conditions in the
airline industry and to monitor risks to the Airport and Airway Trust Fund, we
recommend that the Assistant Secretary for Aviation and International Affairs
direct:
2. The Office of Aviation Analysis to develop a process to regularly collect,
maintain, and use information from airlines’ website disclosures of all fees
charged for optional or ancillary services as a screening mechanism for
significant changes in these fees. For each mainline carrier and posted fee,
this information should include—but not necessarily be limited to—
identification of the type of each service and its price (or price range).
To ensure that airlines and airline passengers are treated equitably in the
collection of air transportation excise taxes and to support the integrity of the
Airport and Airway Trust Fund, we recommend that:
EC2020036 46
3. The Department request a Revenue Ruling or policy statement from the
Department of Treasury regarding the taxation of airline booking fees
and, if appropriate, that the Department of Treasury take action to assess
the relevant tax. If the Department of Treasury finds that these fees are
taxable—and assuming no change in the conditions underlying our
calculation of their impact on the Airport and Airway Trust Fund in 2019—
this could conservatively result in $60.6 million in funds put to better use
in every year following the determination.
Agency Comments and OIG Response
We provided DOT with our draft report on February 26, 2020, and received its
formal response on April 28, 2020. DOT’s response is included in its entirety as an
appendix to this report.
The Office of Inspector General holds all of its work to the highest standards of
evidence, and the evidence supporting each report is independently reviewed for
sufficiency, appropriateness, and reasonableness. The audit objective for this
report was to detail recent aviation industry trends, particularly as they relate to
small- and medium-sized communities. Developments in the airlines’ treatment
of ancillary fees constitute an important aviation industry trend with significant
potential impacts for the Department and the traveling public. The Department’s
statement that online booking fees—the focus of the report’s analysis of ancillary
fees—are primarily charged by airlines serving mostly larger communities is
inaccurate. For example, one of the three carriers that charges this fee draws a far
greater share of its passengers from smaller communities than any other carrier.
47
In its response, the Department criticizes the integrity of our methodology and
quantitative analysis because, in its view, the report’s conclusions and
recommendations relating to ancillary fees go beyond the stated audit objective.
However, for the reasons stated here and in the report itself, ancillary fee
concerns fit well within the audit scope. Moreover, the Department had ample
opportunity to criticize our audit work on its merits. However, after OIG
addressed their concerns about our initial draft, DOT officials did not further
question our evidence or analysis in the several meetings held to discuss
subsequent drafts.
Still, the Department concurs with recommendation 1. Its actions in this regard
will significantly improve the accuracy of the effective ticket prices reported by
47
In 2017, 41.3 percent of Allegiant Air’s enplanements were in smaller communities. For comparison, 18.3 percent of
legacy carriers’ enplanements were in smaller communities.
EC2020036 47
carriers that charge passengers substantial ancillary fees to board an aircraft. This
information is critical to the Department’s ability to assess the status of airline
competition.
Similarly, we recommend that the Department develop a process to regularly
collect and maintain the ancillary fee information airlines disclose on their
websites. However, the nonconcurrence with recommendation 2 limits DOT’s
ability to ensure that airlines comply with the reporting directive it will issue in
response to recommendation 1, as well as limit its awareness of trends affecting
airline competition. The Department states that it already monitors changes in
the airline industry, including ancillary fees. However, its current monitoring
practices failed to detect that online booking fees had become a substantial
revenue source for ULCCs—totaling around $1 billion in 2018—until notified by
OIG in the course of this audit. Recommendation 2 constitutes what we believe to
be the minimal action the Department can take to ensure it is aware of future
significant changes in ancillary fees. Otherwise, the Department faces the risk that
its ticket price data will be an inaccurate source of information about costs to
airline passengers.
Finally, we reiterate that the purpose of recommendation 3 is to obtain
clarification on an issue that puts the AATF at risk. While it is not clear whether
online booking fees are taxable under current law, ULCCs have reallocated an
increasing share of the total boarding cost to online booking fees. Our report
presents a dramatic, but not singular, example in which a carrier charged $35.98
for using its website to book a ticket and $0.02 for the ticket itself. This illustrates
that carriers may be able to entirely avoid collecting the ticket tax by treating
nearly the entire value of the purchase as a booking fee. In this example, the cost
breakdown did not list any taxes as 7.5 percent of $0.02 rounds to $0.00. Absent
a ruling or policy statement on the taxability of online booking fees—carriers may
be able to effectively opt out of collecting the ticket tax. We cannot predict that
this practice will be implemented on a broader scale, but the possibility exposes
the AATF to a considerable, if not existential, threat.
Recommendation 3 asks the Department to ask Treasury whether booking fees
are currently taxable under the ticket tax but does not recommend that the
Department itself take any specific position. We believe clarification and
resolution of this issue is necessary to either properly recoup ticket tax proceeds
or conduct long-term solvency planning for the AATF. As the beneficiary of the
ticket tax, it is appropriate for DOT to inform the Treasury about industry trends
that may affect the collection or application of that tax, as well as the solvency of
the AATF.
EC2020036 48
Actions Required
We consider recommendation 1 resolved but open pending completion of DOT’s
planned actions. We consider recommendations 2 and 3 open and unresolved
and request that DOT reconsider its position and provide us with its revised
response within 30 days of the date of this report in accordance with DOT Order
8000.1C.
Exhibit A. Scope and Methodology 49
Exhibit A. Scope and Methodology
We conducted this performance audit between January 2018 and February 2020
in accordance with generally accepted Government auditing standards as
prescribed by the Comptroller General of the United States. Those standards
require that we plan and perform the audit to obtain sufficient, appropriate
evidence to provide a reasonable basis for our findings and conclusions based on
our audit objectives. We believe that the evidence obtained provides a
reasonable basis for our findings and conclusions based on our audit objectives.
Our objectives for this self-initiated audit were to (1) detail recent aviation
industry trends, particularly as they relate to service to small- and medium-sized
communities; and (2) provide a descriptive analysis of factors associated with
changes in airline service to small- and medium-sized communities. Note that
this report addresses only the first audit objective, since we plan to address the
second objective in a later report. Specifically, we detail trends in airline service
levels; numbers of passengers flown; airline service quality, including connectivity;
and airline competition, including prices paid by airline passengers—particularly
as they relate to small and medium-sized communities.
To meet objective one, we analyzed Census and multiple DOT datasets that
highlighted changes in activity, competition, prices, and service quality from 2005
through 2017. We reviewed airline industry research conducted by Government
agencies as well as academic economists and transportation researchers with a
focus on articles that analyzed competitive practices and service to smaller
communities. In addition, we interviewed representatives of the Regional Airline
Association (RAA), Airlines for America, and the Air Line Pilots Association. We
also met with DOT officials to obtain information on key drivers of commercial air
service practices and to understand the Department’s role in monitoring and
regulating the commercial air service industry.
Importantly, we used the 2010 U.S. Census data to define and separate
communities into five size groups—small, medium-small, medium, medium-large,
and large—based on Census’s statistical areas and population within the
contiguous United States. We used the Department’s Bureau of Transportation
Statistics (BTS) T-100 and the Department’s list of Essential Air Service (EAS)
recipients and transportation facilities to assess changes in departures and
destinations, passenger and seats totals, and connectivity. We used the
Department’s Airline Service Quality Performance (ASQP)—in addition to the EAS
and facility data—to assess changes in service quality, such as the rate of
cancelations, on-time performance, and the minutes of delay associated with late
flights.
Exhibit A. Scope and Methodology 50
We used a pre-processed version (SS1B) of the Department’s origin-destination
survey data (DB1B), which removes the excise tax and segment PFC fees, to
assess changes in competitive conditions. In addition, we used the T-100, the
Department’s list of transportation facilities and list of changes in airline
ownership, RAA annual reports and Security and Exchange Commission (SEC)
filings to assess changes in competitive conditions.
We assessed changes in the relative cost of flying from smaller communities by
utilizing the SS1B, the Department’s facility data, BTS T-100, and BTS Form
41 P1.2, which collects quarterly airline financial data. We ran regressions on the
data to generate price indices over the sample period to assess changes in the
relative cost of flying. We assessed changes in booking fees and foregone AATF
revenue by utilizing archived and current air carrier websites, SEC filings and BTS
T-100. We used the Department’s guidance and BTS Form 41 P1.2 data to assess
the potential impacts of insufficient ancillary fee data.
In the following sections, we detail our definition of communities and community
size groups, data preparation, and data analysis. First, we discuss our process for
defining communities and their corresponding size groups. Second, we outline
the datasets used in this audit and provide additional detail on our data
preparation. Third, we detail our methodology for our connectivity, market
structure, and foregone tax revenue computations as well as for constructing our
price indices.
Defining Communities and Size Groups
In this section, we detail our definition of communities, our mapping of airports
into their communities, and our definition of community size groups.
Defining Communities and Their Airports
To construct airports’ catchment areas, we followed Wittman’s 2014 report on air
service accessibility,
48
and used Primary Statistical Areas (PSA). We refer to the
constructed catchment areas throughout our report as communities. PSAs are
defined based on two census definitions—Core Based Statistical Areas (CBSA)
and Combined Statistical Areas (CSA). The CBSA represents a county or set of
counties
49
with at least one urbanized area or cluster with a population of at least
10,000, plus adjacent counties with significant social and economic integration
48
Michael D. Wittman, An Assessment of Air Service Accessibility in U.S. Metropolitan Regions, 2007–2012 (Report No.
ICAT-2014-02), 2014.
49
The term counties is used here to refer to counties or county-equivalents. For example, Louisiana is divided into
parishes rather than counties, and we treat parishes as a county-equivalent.
Exhibit A. Scope and Methodology 51
with the core county based on commuting ties.
50
The CSA is a higher level of
aggregation, which consists of two or more CBSAs that have a significant
employment interchange. For CBSAs within a CSA, the PSA is defined identically
to the CSA. Also, for CBSAs that do not lie within a CSA, the PSA is defined
identically to the CBSA. For counties that are not located within a CBSA—those
without an urban cluster with a population of at least 10,000—we define the PSA
as the county. Figure 22 is a flowchart that illustrates our process for assigning
counties to their PSA.
Figure 22. Construction of PSAs
Source: OIG generated
We used information from Census and FAA to determine which airports lie within
each PSA. The Census information listed counties alongside their corresponding
CBSA or CSA, which we used to construct the PSAs. FAA’s information included
the county of each airport in the United States, which we used to assign airports
to PSAs.
51
W
e decided to conduct our analysis at the community level—rather than at the
airport level—because smaller airports do not always fall within smaller
communities, and the impact of changes in service on passengers at a smaller
airport may differ if there are alternative airports nearby. For example, the closest
airport to Worcester, MA, is Worcester Regional Airport (ORH). However,
Worcester is also located around 50 miles from Boston Logan International
Airport (BOS) and T. F. Green International Airport (PVD) in Providence, RI. On the
other hand, the closest airport to Knoxville, TN, is McGhee Tyson Airport (TYS).
The nearest alternative airports are Tri-Cities Regional Airport (TRI) and
Chattanooga Metropolitan Airport (CHA), both around 100 miles away. Because
50
CBSAs may correspond to either a Metropolitan Statistical Area or a Micropolitan Statistical Area.
51
We were unable to merge the census data into the airport data for a small group of airports. Because these airports
represented a miniscule share of passenger enplanements over our timeframe—0.005 percent—we dropped these
airports.
Exhibit A. Scope and Methodology 52
prospective travelers from Worcester have alternative airports nearby, while
prospective travelers from Knoxville do not, change in service at the local airport
may not have as great an impact on travelers from Worcester as it has on
travelers from Knoxville. Therefore, our definition groups ORH with other airports
in the Boston-Worcester-Providence CSA. On the other hand, TYS is not grouped
with any other airports. See exhibit F for a list of our multiple airport PSAs.
While there are several possible ways to define communities, we chose our
definition for two major reasons. First, we wanted to define communities in a
manner that enabled us to cover all airports in the contiguous United States—
without requiring us to make ad-hoc assessments of individual airports. Second,
we valued a definition that—to the extent possible without an ad hoc
characterization—aligned with airports’ catchment areas. Our evidence suggests
that these two considerations are reasonably addressed by our definition.
Defining Community Size Groups
We constructed five community size groups—small, medium-small, medium,
medium-large, and large—based on community population data from the U.S.
Census Bureau. By construction, the combined populations of communities within
each of these size groups represents roughly 20 percent of the population of the
contiguous 48 United States. This definition is conceptually similar to a
categorization of routes that was used in GAO’s 2014 report on airline
competition,
52
in which each route size group accounts for roughly 20 percent of
passenger enplanements.
53
W
e encoded communities by sorting communities by their population. Beginning
with the largest community in the country (New York-Newark, NY-NJ-CT-PA) and
proceeding iteratively to the community with the next highest population, we
classified communities as large until the cumulative population of these
communities was approximately 20 percent. At this point, we classified the next-
largest community as medium-large and similarly proceeded to label the next
largest communities as medium-large until the cumulative population of
medium-large and large communities combined was approximately 40 percent.
We continued this process to code medium, medium-small, and small
communities until all the communities were classified into one of the five size
groups. Therefore, the entire population of the contiguous 48 United States was
accounted for in one of the size groups.
52
GAO, Airline Competition: The Average Number of Competitors in Markets Serving the Majority of Passengers Has
Changed Little in Recent Years, but Stakeholders Voice Concerns about Competition (GAO-14-515), June 2014.
53
We chose the number of groups—five—to align with GAO’s definition and for exposition.
Exhibit A. Scope and Methodology 53
We considered a few other factors when deciding upon our categorization of
communities into size groups. First, we wanted to define categories in a manner
that would not require us to make any ad hoc assessments of individual
communities. The process described in the preceding paragraph satisfies this
criteria, as communities were categorized based on Census data. Second, we
valued a definition that did not directly use information about the communities’
airline service—such as enplanements—to categorize communities. Our
categorization achieves this by using information on populations without using
any information on airline service.
54
Third, we wanted a definition that would
reasonably align with terminology used in the industry and the Department. We
compared our definition to terminology used by airlines in public presentations
and filings, as well as to documentation for communities that received a Small
Community Air Service Development Program grant from DOT. In both cases, we
determined our terminology is largely consistent.
55
Data Preparation
In this section, we discuss the sources and preparation of the data used in the
report.
T-100 Database
The T-100 data was downloaded from the BTS online portal. This database
reports monthly air carrier traffic information from certified U.S. air carriers. The
data includes monthly information on air traffic. This includes the origin,
destination, operating carrier, number of departures performed, passengers, and
seats.
We prepared the T-100 data with a few additional filters and restrictions. We
restricted the data by dropping flights with either an origin or destination outside
the contiguous United States. Additionally, we filtered the data by retaining only
flights that represent scheduled passenger service, and dropping observations
with zero recorded departures or potentially erroneous passenger data. We also
restricted the data to airports that had at least 2,500 enplanements in at least
1 year between 2005 and 2017, and flights that either had at least 5 average
54
Airline service may indirectly affect this grouping through its impact on a community’s population.
55
Between 2005 and 2016, 92 percent of communities which received a Small Community Air Service Development
Program grant from the USDOT are categorized by our approach as small or medium-small. In a February 2018
management presentation, Allegiant listed examples of cities and their size categorization. The example in their
smallest origin category (“tiny”) is classified by our algorithm into our smallest origin category (S); the example in their
second-smallest origin category (“small”) is classified into our second-smallest category (MS); and the example in their
third-smallest origin category (“mid-size”) is classified into our third-smallest category (M).
Exhibit A. Scope and Methodology 54
passengers or those that had between 2 and 5 average passengers with at least
8 departures per month.
SS1B Database
We obtained our ticket price data—the Superset 1B (SS1B)—from Airline Data,
Inc. This data is a pre-cleaned version of DOT’s Origin and Destination Survey
(DB1B). The DB1B contains quarterly data on a 10 percent sample of airline tickets
from reporting carriers. Specifically, the DB1B is collected from carriers that
operate any aircraft that are designed for a maximum seating capacity of more
than 60 seats. The SS1B data is produced primarily using the DB1B data, and is
cross-validated with the T-100. Additionally, the SS1B data pre-filters fares
56
and
removes excise taxes.
We further prepared the SS1B data in a few additional ways. We dropped flights
with either an origin or destination outside the contiguous United States, and
also dropped open-jaw itineraries.
57
For the price index analysis, we augmented
the data by merging in information on ancillary revenues from BTS Form
Schedule P-1.2 so that the reported prices include these average charges.
Specifically, we computed for each quarter and carrier the per-passenger revenue
associated with two fields—baggage and change/cancellation fees—and merged
these two fields into the SS1B. Finally, we deflated the ticket prices listed in the
SS1B to a base period of the first quarter of 2005 using the Consumer Price Index
from the Bureau of Labor Statistics.
Additional Datasets
Table 6 below lists additional data sets we used, a brief explanation of each data
set, and how these data sets were used in our report.
56
Specifically, observations are filtered if they have a one-way price below $25 because these fares historically
represented purchases made with frequent flyer points.
57
Open-jaw itineraries are those where a passenger returns from a different airport than their outbound destination.
Exhibit A. Scope and Methodology 55
Table 6. Description of Additional Data Used
Data
Source
Explanation
Uses
Airport locations DOT/FAA
Information extracted from
FAA’s Airport Data and
Contact Information query
tool, including airport codes,
addresses, and geographic
coordinates.
Used with PSA boundary data to
encode each airport into its PSA.
PSA boundaries U.S. Census Bureau For each county or county
equivalent, lists its CBSA
and/or CSA, if applicable, as
defined by the Office of
Management and Budget in
August, 2017.
Used with airport location data to
encode each airport into its PSA.
PSA populations U.S. Census Bureau Data on the population of
each county or county
equivalent from the 2010
Census.
Used to classify PSAs into a size
group.
BTS Form 41 Schedule
P-1.2
DOT/BTS Quarterly financial data
provided for select US airlines,
including information on
baggage and reservation
change/cancellation fees.
Used to adjust price index data
to include the average baggage
and change/cancellation fees for
each carrier-quarter.
EAS contracts DOT/Office of the
Secretary
Information on each EAS
contract from 2005 through
2018, including origins and
destinations.
Used to define EAS communities
and routes with an EAS subsidy
over our period of analysis.
Airline ownership DOT/Air Carrier Fitness
Division
Information on ownership
changes recorded by DOT’s
Air Carrier Fitness Division.
Used to define airlines’ holding
companies over time.
Airline Service Quality
Performance
DOT/BTS Flight-level data on delays
and cancellations, which
includes whether the flight
was cancelled and the delay
in minutes.
Used to compute delays and
cancellation rates in the service
quality section.
Source: OIG generated
Data Analysis
In this section, we detail our connectivity measures, market structure measures,
and the estimation of our price indices. In addition, we describe our calculation of
foregone tax revenue associated with booking fees.
Exhibit A. Scope and Methodology 56
Connectivity
We used the T-100 data from 2005 through 2017 to construct a community-level
measure of connectivity based on Wittman and Swelbar’s Airport Connectivity
Quality Index (ACQI).
58
The ACQI computes airport connectivity based on the
frequency of available scheduled flights, the quantity and quality of nonstop
destinations, and the quantity and quality of connecting destinations.
We made two adjustments to the ACQI, and refer to our measure as the
Community Connectivity Quality Index (CCQI). First, we defined connectivity at
the community level rather than the airport level to align with our interest in
studying airline service across communities, rather than airports. As shown in
exhibit F, there are several multi-airport communities in our data, and accounting
for each of a passenger’s airport options—rather than just a single airport—is
important to measure connectivity in those multi-airport communities. Second,
the quality measures for destinations, whether direct or indirect, are based on the
community size group’s share of total enplanements. In contrast, Wittman and
Swelbar compute this parameter using each FAA hub type’s share of total
enplanements. Accounting for these changes, the CCQI for community
p
is
computed as
= (
)
+ (
)
where
i I
denotes a nonstop destination and
j J
denotes a destination that can
be reached by connecting through a nonstop destination (i.e., a one-stop
destination).
The quality of nonstop destinations is represented by the first summation, and
includes:
1.
f
pc
,
which is the average number of daily scheduled flights per destination
from PSA
p
to community of size
c
;
59
2.
d
pc
,
which is the number of nonstop destinations of size
c
served from PSA
p
; and
58
Michael D. Wittman and William Swelbar, “Capacity Discipline and the Consolidation of Airport Connectivity in the
United States,” Transportation Research Record: Journal of the Transportation Research Board, vol. 2449, no. 1 (2014),
pp. 72–78.
59
Community sizes are defined in the same manner as described in the first section of this exhibit.
Exhibit A. Scope and Methodology 57
3.
w
c
, which is the weight attached to destination size
c
as determined by its
enplanement share, as described in the final paragraph of this subsection.
Connecting service is represented by the second summation, and is multiplied by
a scaling factor to account for the differential impact of connecting versus
nonstop service on connectivity. This piece of the computation includes:
4.
d
pc
, which is the number of connecting destinations of size
served from
PSA
p
;
5.
w
c
, which is the weight attached to connecting destination of size
; and
6. , which is a scaling factor for connecting versus nonstop service.
The CCQI includes three parameters that are defined outside the model—,
w
c
,
and
w
. First, we chose the scaling factor on connecting service () based on
literature regarding the ACQI and the Quality Service Index, which is a model
used by airlines to compute market share based on path quality. In Quality
Service Index models, this parameter has generally fallen between 0.03 and 0.20,
and = 0.125 is used in the ACQI parameter. We followed the ACQI and
specified = 0.125 in our CCQI. We defined the relative quality of a destination
airport, denoted by
w
for connecting destinations and
w
for nonstop
destinations, based on the domestic share of enplanements of the PSA size
group.
60
Specifically, we computed the average domestic enplanement share of a
PSA within each size group, and then normalized the weights relative to the
average enplanement share of a large PSA. Table 7 below lists the weights
assigned by community size group.
Table 7. Weights Assigned by Community Size Group
Community Size Group
Weight (
)
Large 1.000
Medium-Large 0.632
Medium 0.218
Medium-Small 0.026
Small 0.003
Source: OIG-generated
60
The ACQI computes these weights based on the share of enplanements by FAA hub type.
Exhibit A. Scope and Methodology 58
We computed the average change in connectivity for each of our community size
groups over time by computing the passenger-weighted average connectivity in
each community size group for each year from 2005 through 2017. This is given by
=
, where
pax
pt
represents a community’s passengers in
year
t
, and we sum over all communities
p
in community size group
c
. We then
normalized the CCQI of each community size group in each year based on its
2005 CCQI:
=
,
By construction, for any community size group
c
,
,
= 1. In any year, the
normalized CCQI score of a community size group
c
can be converted into a
percentage change since the baseline year of 2005 as %
=
,
,
.
Market Structure
To conduct our market structure and competition analysis, we began by taking
additional steps to further prepare the SS1B database. Specifically, we dropped
interline itineraries—which are those with multiple marketing carriers on the
same itinerary.
61
We also used information on airline ownership to encode the
holding company of each airline over time, and—unless otherwise noted—
conducted our analysis at the level of holding companies rather than individual
airlines. The data was aggregated to the level of holding company–route–year
prior to conducting the analysis. We used the term route to refer to the
origin community–destination community pair throughout this section.
W
e computed the HHI for each route-year combination using holding
companies’ squared market shares of passengers. Letting
t
represent the year,
j J
the holding company (e.g., “American Airlines Group”), and
r R
the route,
the HHI for each route-year combination is given by
= 10,000
After computing the HHI for each route in our data, we aggregated the data to
the community size group–year level, by computing the passenger-weighted HHI
for each community size group in each year. Letting
c
represent a community size
group (e.g., small communities) and
pax
rt
represent the number of passengers on
61
We retained itineraries which have multiple operating carriers, as long as they had one marketing carrier. For
example, a one-stop itinerary which is marketed by Alaska Airlines with one leg operated by Alaska Airlines and the
other operated by Horizon Air would not be dropped from our data.
Exhibit A. Scope and Methodology 59
a route in a given year, the passenger-weighted HHI for a community size group
in a given year is computed by
=
By passenger-weighting in this manner, the HHI presented represents the
average HHI faced by a passenger within a community size group in a given year.
For example, consider a simple scenario with only two routes in small
communities: Bozeman Yellowstone International Airport (BZN) to Denver
International Airport (DEN)—with 200 passengers and an HHI of 10,000—and
Duluth International Airport (DLH) to O’Hare International Airport (ORD)—with
100 passengers and an HHI of 2,500. The passenger-weighted HHI in this
example is given by
10,000 +
2,500 = 7,500. The route with more
passengers is given greater weight. As a result, the HHI lies closer to the HHI of
BZN to DEN than the HHI of DLH to ORD.
62
Our computations for effective competitors involve an identical passenger-
weighting to our computations for the HHI. As a result, this measure represents
the average number of effective competitors faced by a passenger within a
community size group in a given year. Specifically, the route-level effective
competitors was given by
=
1[
0.05]
where 1
[
.
]
represented an
indicator function. The community size group’s passenger-weighted number of
effective competitors in a given year was then given by
=
Continuing with the earlier example, assume that the route from BZN to DEN with
200 passengers has 1 effective competitor, while the route from DLH to ORD with
100 passengers has 2 effective competitors. The passenger-weighted effective
competitor in this example is given by:
1 +
2 1.33. Once again,
because the route with more passengers was given greater weight, the number of
effective competitors lies closer to the 1 effective competitor of the BZN to DEN
route than to the 2 effective competitors of the DLH to ORD route.
Price Indices
We estimated our hedonic price indices using techniques outlined by Aizcorbe’s
2014 guide to price index and hedonic techniques.
63
The methodology behind
estimation of hedonic price indices provides an explicit way to control for
62
We did not filter out routes with low passenger counts throughout the report. We ran a sensitivity check of our
analysis where we dropped low passenger routes. Because routes with few passengers receive a relatively small
weight in both the HHI and effective competitor computations, our results did not change notably.
63
Ana M. Aizcorbe, A Practical Guide to Price Index and Hedonic Techniques (Oxford University Press, 2014).
Exhibit A. Scope and Methodology 60
changes in product characteristics when constructing a price index. Specifically,
we estimated a dummy variable hedonic price index for each community size
group by estimating a regression of prices on variables including dummy
variables for each year-community size group combination. We also included
product characteristics in the regression to account for changes in product
characteristics over time. These characteristics included the type of airline (e.g.,
legacy carrier); the number of segments flown to reach the destination; the
geodesic distance; the circuity of the itinerary, defined as the ratio of an
itinerary’s total distance to that of a nonstop itinerary on the same route; and the
number of seats on the plane relative to the typical number of seats a particular
aircraft historically contained, in order to account for changes in seat density
during this period.
Prior to running our regression, we aggregated the data to the level of origin-
destination-marketing carrier-product-year-quarter. Letting
t
represent the year,
m
represent a specific product defined by a set of characteristics
X
kmt
, and
D
mct
represent a dummy variable equal to 1 if product
m
is in community group
c
during time period
t
, we estimated a passenger-weighted ordinary least squares
regression of the form
64
ln
= +
+
+
where
P
mt
represents product
m
’s price at time
t
. By construction, the coefficient
ct
was allowed to vary across community size group and time. The price index for
community group
c
at time
t
is then given by e
ct
, and can be interpreted relative
to the price of a product in a large community at time
t = 2006
(the omitted
category). In the report, we present the price premium relative to that of a large
community for each community size group and each year. For year
t
and
community size group
, the price premium was computed as
where
L
denotes the large community group. This represents the difference—in
absolute percentage points—between the prices in community group
at time
and the prices in large communities at time
.
W
e adjusted the prices used in the price index to partly account for changes in
ancillary fees throughout our period of analysis. We did this by using the Form 41
Schedule P-1.2 data to compute, for each quarter and each carrier, the per-
passenger average baggage and change/cancellation fees, and added these
averages to the associated ticket prices.
There are two notable shortcomings, which arise in our ancillary fee adjustment,
that result from data limitations. First, this calculation implicitly assumes that the
64
We ran the regression in semilog form for two primary reasons. First, the semilog model can accommodate
characteristics that may be equal to zero whereas a log-log model cannot. Second, it is more likely that the errors are
homoscedastic in a semilog model than in a linear model.
Exhibit A. Scope and Methodology 61
average per-passenger baggage and change/cancellation fees do not vary across
community size groups. For example, if passengers in small communities across
all airlines were more likely than passengers in large communities to pay
baggage fees, our adjustment would underestimate the impact of these fees on
passengers in small communities.
65
Absent more granular data on ancillary fees,
it is not feasible for us to test this assumption. Second, because other ancillary
fees are not separately identifiable, fees such as seat assignment fees and
booking fees were not included in this adjustment.
Foregone Tax Revenue
We estimated the foregone tax revenue—which could result if tax is not collected
on ULCCs’ booking fees—using information on the level of these carriers’
booking fees, the share of booking across their various distribution channels, and
their passenger enplanements. We computed two estimates—a high-end
estimate and a low-end estimate—using two different sets of assumptions. In our
recommendation, we report the more conservative estimate.
For each of our calculations, we used information from BTS on the number of
domestic passenger enplanements for the three ULCCs from August 2018
through July 2019—the most recent twelve months of available data. We used
data on booking fees as of September, 2019. Also, two carriers—Allegiant Air and
Spirit Airlines —listed the typical levels of this fee on their websites or other
documentation. While Frontier Airlines notes the existence of this fee on their
website, we were unable to find the level of this fee listed. Thus, the level of
Frontier’s booking fee was determined through online queries for flights of
varying prices. We also used information compiled from SEC filings to determine
the share of passengers who either book through the carrier’s website or over the
phone. Notably, Frontier Airlines and Spirit Airlines allow passengers to book
through third-party entities such as Expedia, and we are unable to determine
whether the booking fees apply to fares booked through such third-party
channels.
66
For both estimates, we assumed that the foregone revenue was only
associated with bookings made through the carrier’s website or their call center.
65
This assumption differs from the possibility that ancillary fees differ across community size groups due to
differences in airlines’ strategies towards different size communities. Because we merge this information in at the
carrier-quarter level, any differences arising from differences in the carriers which serve each route are accounted for
in our adjustment. Specifically, we cannot account for differences in ancillary fees within-airline but across community
size group.
66
The most recently available data regarding whether any revenue from third-party bookings is recognized by the
carriers as a booking fee comes from Spirit Airline’s 2017 10-K filing. In that filing, Spirit Airlines reports that the
booking fee is “charged for tickets sold through the Company’s primary sales distribution channels. The primary sales
distribution channels for which passenger usage fees are charged include sales through the Company’s website, sales
through the third-party provided call center and sales through travel agents; the Company does not charge a
passenger usage fee for sales made at its airport ticket counters.”
Exhibit A. Scope and Methodology 62
Table 8 below lists the information we used to compute our estimate of foregone
AATF revenue.
Table 8. Information Used to Compute Foregone AATF Revenue
Airline
Information Type
Value
Allegiant Air
Domestic enplanements, 08/2018–
07/2019
14,528,629
Higher online booking fee $18
Lower online booking fee $18
Share booking through carrier’s
webpage
0.938
Frontier Airlines Domestic enplanements, 08/2018–
07/2019
20,005,002
Higher online booking fee $21
Lower online booking fee $10
Share booking through carrier’s
webpage or call center
0.630
Spirit Airlines Domestic enplanements, 08/2018–
07/2019
28,858,678
Higher online booking fee $22.99
Lower online booking fee $11.99
Share booking through carrier’s
webpage or call center
0.729
Source: Enplanements from BTS, booking fees from carriers’ webpages, and share of bookings by
distribution channel from carriers’ SEC filings
For the high-end estimate, we assume that all passengers pay the higher of the
two booking fees listed in the table above. The estimates shown in table 9 below
are produced by computing 7.5 percent of the product of each carrier’s domestic
enplanements and its higher online booking fee, multiplied by the share
purchased through the website or call center. For the low-end estimate, we
assume that for the two carriers that charge a varying booking fee, 50 percent of
enplanements carry the higher of the two booking fees listed and 50 percent
carry the lower of the two. The rest of the computation is done in the same was
as our high-end estimate. Allegiant Air’s booking fee does not vary across
itineraries, so their low-end estimate is equivalent to their high-end estimate.
Frontier Airlines and Spirit Airlines attach a lower booking fee to certain
reservations, which explains the difference between the high-end and low-end
estimates of the foregone revenue associated with these carriers. Table 9 lists the
Exhibit A. Scope and Methodology 63
estimated foregone revenue for each carrier in the low-end and high-end
scenarios.
Table 9. Estimated Foregone Revenue
Carrier(s)
Low-End Estimate
High-End Estimate
Allegiant Air $18.4 million $18.4 million
Frontier Airlines $14.7 million $19.8 million
Spirit Airlines $27.6 million $36.3 million
Total $60.6 million $74.5 million
Source: OIG analysis of data compiled from BTS, carriers’ SEC filings, and carriers’
webpages
We believe our estimates represent a conservative estimate of foregone tax
revenue during 2019 for a few reasons.
67
First, this computation uses information
on enplanements over the past 12 months, which includes August 2018 through
December 2018. Because all three ULCCs have seen increasing passenger
enplanements in recent years, it is likely that passenger enplanements for August
2019 through December 2019 will exceed those from August 2018 through
December 2018. Second, the low-end assumes that half of passengers traveling
on Frontier Airlines and Spirit Airlines pay the lower booking fee, but available
information suggests the higher booking fee is more common. For example,
Spirit notes that the $22.99 per-segment fee applies to most reservations.
68
Third,
available information suggests the share of passengers who pay a booking fee
could be higher than the figures used in our calculation. Allegiant Air lists
93.8 percent of its revenue as earned through its webpage—we use this figure for
our calculation—but of the remaining 6.2 percent likely includes a significant
number of passengers who purchased tickets over the phone, thus incurring a
booking fee.
69
Further, we used the most recently available data for Frontier
Airlines’ and Spirit Airlines’ share of bookings through their webpage or over the
phone. For each carrier, this data shows an increasing share of bookings through
67
In addition to the computational assumptions listed here, this estimate does not account for changes in traveler
behavior which could result from requiring carriers to collect the excise tax on booking fees. Imposing this tax would
raise the relative price of travel on a ULCC. This could result in substitution of passengers from a ULCC to a non-ULCC
as well as substitution of passengers across modes of transportation. Accounting for substitution across carriers
would increase estimates of foregone revenue, while accounting for substitution across modes would decrease
estimates of foregone revenue. We do not have empirical evidence regarding which of these effects would
predominate, and determining which effect predominates is outside the scope of this report.
68
Spirit Airlines, General Terms and Conditions, August 2019.
69
The 6.2 percent of revenue associated with passengers which do not book through Allegiant’s webpage ostensibly
includes those who book over the phone or at the airport. While the latter do not incur a booking fee, it is possible
these represent a relatively small share of this 6.2 percent. For example, Spirit Airlines’ 2018 10-K filing shows 5.8
percent of distribution came through their call center.
Exhibit A. Scope and Methodology 64
the carrier’s webpage or call center. Therefore, the share of passengers booking
through the carrier’s webpage or call center may be higher in 2019 than it was
when these carriers most recently reported the data.
70
70
Spirit Airlines most recently reported this data in 2018. From 2017 to 2018, the share of passengers booking
through either Spirit’s website or their call center rose from 71.6 percent to 72.9 percent. Frontier Airlines most
recently reported this data in 2016. From 2015 to 2016, the share of passengers booking through Frontier’s website,
mobile application, or another direct channel rose from 58 percent to 63 percent.
Exhibit B. Organizations Visited or Contacted 65
Exhibit B. Organizations Visited or Contacted
Department of Transportation
Aviation Consumer Protection Division
Bureau of Transportation Statistics
Office of the Secretary
Office of Aviation Analysis, Air Carrier Fitness Division
Office of Aviation Analysis, Competition and Policy Analysis Division
Office of Aviation Analysis, Essential Air Service and Domestic Analysis Division
Office of Aviation Analysis, Small Community Air Service Development Program
Other Organizations
Airlines for America
Air Line Pilots Association
Internal Revenue Service
Government Accountability Office
Regional Airline Association
Treasury Inspector General for Tax Administration
Exhibit C. List of Acronyms 66
Exhibit C. List of Acronyms
AATF Airport and Airway Trust Fund
ACQI Airport Connectivity Quality Index
BTS Bureau of Transportation Statistics
CBSA Core Based Statistical Area
CCQI Community Connectivity Quality Index
CSA Combined Statistical Area
DB1B Origin and Destination Survey
DOJ Department of Justice
DOT Department of Transportation
EAS Essential Air Service
FAA Federal Aviation Administration
FTC Federal Trade Commission
GAO Government Accountability Office
HHI Hirschman-Herfindahl Index
ICAT International Center for Air Transportation
IRS Internal Revenue Service
L Large (Community)
LCC Low-cost carrier
M Medium (Community)
MIT Massachusetts Institute of Technology
ML Medium-Large (Community)
MS Medium-Small (Community)
OIG Office of Inspector General
PFC Passenger Facility Charge
PSA Primary Statistical Area
S Small (Community)
SEC U.S. Securities and Exchange Commission
SS1B Superset 1B
ULCC Ultra-low-cost carrier
Exhibit D. Major Contributors to This Report 67
Exhibit D. Major Contributors to This Report
BETTY KRIER CHIEF ECONOMIST
JERROD SHARPE SENIOR ECONOMIST
BRAD SHRAGO SENIOR ECONOMIST
TOM DENOMME AUDIT EXPERT
SETH KAUFMAN DEPUTY CHIEF COUNSEL
CELESTE BORJAS ASSOCIATE COUNSEL
SUSAN CROOK-WILSON WRITER-EDITOR
JANE LUSAKA WRITER-EDITOR
Exhibit E. Categorization of Select Airlines 68
Exhibit E. Categorization of Select Airlines
Carrier
Code
Segment
Category
Detailed Category
American Airlines Inc. AA Mainline Legacy Not Applicable
Continental Air Lines Inc. CO Mainline Legacy Not Applicable
Delta Air Lines Inc. DL Mainline Legacy Not Applicable
Northwest Airlines Inc. NW Mainline Legacy Not Applicable
United Air Lines Inc. UA Mainline Legacy Not Applicable
US Airways Inc. US Mainline Legacy Not Applicable
AirTran Airways Corporation FL Mainline Non-Legacy LCC/Other
Alaska Airlines, Inc. AS Mainline Non-Legacy LCC/Other
America West Airlines Inc. HP Mainline Non-Legacy LCC/Other**
JetBlue Airways B6 Mainline Non-Legacy LCC/Other
Midwest Airlines, Inc. YX* Mainline Non-Legacy LCC/Other
Southwest Airlines Co. WN Mainline Non-Legacy LCC/Other
Sun Country Airlines SY Mainline Non-Legacy LCC/Other
Virgin America VX Mainline Non-Legacy LCC/Other
Allegiant Air G4 Mainline Non-Legacy ULCC
Frontier Airlines Inc. F9 Mainline Non-Legacy ULCC***
Spirit Air Lines NK Mainline Non-Legacy ULCC
Air Wisconsin Airlines Corp ZW Regional Not Reportable Not Applicable
Atlantic Southeast Airlines EV* Regional Not Reportable Not Applicable
Chautauqua Airlines Inc. RP Regional Not Reportable Not Applicable
Colgan Air 9L Regional Not Reportable Not Applicable
Comair Inc. OH* Regional Not Reportable Not Applicable
Commutair C5 Regional Not Reportable Not Applicable
Compass Airlines CP Regional Not Reportable Not Applicable
Endeavor Air Inc. 9E Regional Not Reportable Not Applicable
Envoy Air MQ Regional Not Reportable Not Applicable
ExpressJet Airlines Inc. EV*/XE Regional Not Reportable Not Applicable
Exhibit E. Categorization of Select Airlines 69
Carrier
Code
Segment
Category
Detailed Category
Freedom Airlines F8 Regional Not Reportable Not Applicable
GoJet Airlines LLC G7 Regional Not Reportable Not Applicable
Horizon Air QX Regional Not Reportable Not Applicable
Mesa Airlines Inc. YV Regional Not Reportable Not Applicable
Mesaba Airlines XJ Regional Not Reportable Not Applicable
Piedmont Airlines PT Regional Not Reportable Not Applicable
PSA Airlines Inc. OH* Regional Not Reportable Not Applicable
Republic Airlines YX* Regional Not Reportable Not Applicable
Shuttle America Corp. S5 Regional Not Reportable Not Applicable
SkyWest Airlines Inc. OO Regional Not Reportable Not Applicable
Trans States Airlines AX Regional Not Reportable Not Applicable
Note: We do not report ownership status of regional airlines because this information may not be publically
available for all carriers.
* Code was used for different carriers at different times between 2006 and 2017
** America West Airlines Inc. only operated under the holding company of US Airways during the period of our
price and market structure analysis
*** Frontier Airlines transitioned to the ULCC model later in this period
Source: OIG analysis of DOT data. Mainline carriers are listed if they were the marketing carriers for at least
0.1 percent of passengers from 2006 through 2017. Regional carriers are listed if they were the operating carriers
for at least 0.1 percent of passengers from 2006 through 2017.
Exhibit F. List of Multi-Airport PSAs 70
Exhibit F. List of Multi-Airport PSAs
PSA
PSA
Size
Population
Airports
Boston-Worcester-Providence, MA-
RI-NH-CT
ML 7,893,376
BED, BID, BOS, EWB, HYA, MHT, ORH, PSM, PVC, PVD,
WST
New York-Newark, NY-NJ-CT-PA L 23,076,664 ABE, EWR, FRG, HPN, HVN, ISP, JFK, LGA, MMU, SWF, TTN
Los Angeles-Long Beach, CA L 17,877,006 BUR, LAX, LGB, ONT, OXR, PMD, PSP, SNA
Las Vegas-Henderson, NV-AZ M 2,195,401 BLD, HII, IFP, IGM, LAS, VGT
San Jose-San Francisco-Oakland, CA L 8,153,696 CCR, OAK, SCK, SFO, SJC, STS
Miami-Fort Lauderdale-Port St.
Lucie, FL
ML 6,166,766 FLL, MIA, PBI, VRB
Washington-Baltimore-Arlington,
DC-MD-VA-WV-PA
L 9,051,961 BWI, DCA, HGR, IAD
Albuquerque-Santa Fe-Las Vegas,
NM
MS 1,146,049 ABQ, LAM, SAF
Chicago-Naperville, IL-IN-WI L 9,840,929 GYY, MDW, ORD
Cleveland-Akron-Canton, OH M 3,515,646 BKL, CAK, CLE
Dallas-Fort Worth, TX-OK ML 6,851,398 AFW, DAL, DFW
Flagstaff, AZ S 134,421 FLG, GCN, PGA
Houston-The Woodlands, TX ML 6,114,562 EFD, HOU, IAH
Modesto-Merced, CA MS 770,246 MCE, MER, MOD
Orlando-Deltona-Daytona Beach, FL M 2,818,120 DAB, MCO, SFB
Philadelphia-Reading-Camden, PA-
NJ-DE-MD
ML 7,067,807 ACY, ILG, PHL
San Juan, WA S 15,769 ESD, FRD, S31
Seattle-Tacoma, WA ML 4,274,767 BFI, OKH, SEA
Salt Lake City-Provo-Orem, UT M 2,271,696 OGD, PVU, SLC
Atlanta--Athens-Clarke County--
Sandy Springs, GA
ML 5,910,296 AHN, ATL
Bakersfield, CA MS 839,631 BFL, IYK
Buffalo-Cheektowaga, NY MS 1,215,826 BUF, IAG
Bozeman, MT S 89,513 BZN, WYS
Cincinnati-Wilmington-Maysville,
OH-KY-IN
M 2,174,110 CVG, LUK
Columbus-Marion-Zanesville, OH M 2,308,509 CMH, LCK
Exhibit F. List of Multi-Airport PSAs 71
PSA
PSA
Size
Population
Airports
Charlotte-Concord, NC-SC M 2,375,675 CLT, JQF
Charleston-Huntington-Ashland,
WV-OH-KY
MS 708,228 CRW, HTS
Detroit-Warren-Ann Arbor, MI ML 5,318,744 DTW, FNT
Duluth, MN-WI S 279,771 DLH, HIB
Edwards-Glenwood Springs, CO S 125,734 ASE, EGE
Grand Rapids-Wyoming-Muskegon,
MI
MS 1,379,237 GRR, MKG
Brownsville-Harlingen-Raymondville,
TX
MS 4,28,354 BRO, HRL
Jacksonville-St. Marys-Palatka, FL-
GA
M 1,470,473 JAX, SGJ
Key West, FL S 73,090 EYW, MTH
Memphis-Forrest City, TN-MS-AR MS 1,353,087 MEM, UTA
Minneapolis-St. Paul, MN-WI ML 3,684,928 MSP, STC
Ogdensburg-Massena, NY S 111,944 MSS, OGS
Phoenix-Mesa-Scottsdale, AZ ML 4,192,887 AZA, PHX
Pittsburgh-New Castle-Weirton, PA-
OH-WV
M 2,660,727 LBE, PIT
Portland-Vancouver-Salem, OR-WA M 2,921,408 PDX, SLE
Cape Coral-Fort Myers-Naples, FL MS 940,274 APF, RSW
Santa Maria-Santa Barbara, CA MS 423,895 SBA, SMX
San Diego-Carlsbad, CA M 3,095,313 CLD, SAN
Springfield-Branson, MO MS 520,589 BBG, SGF
North Port-Sarasota, FL MS 897,121 PGD, SRQ
St. Louis-St. Charles-Farmington,
MO-IL
M 2,892,497 BLV, STL
Tampa-St. Petersburg-Clearwater, FL M 2,783,243 PIE, TPA
Virginia Beach-Norfolk, VA-NC M 1,779,243 ORF, PHF
Source: OIG analysis of U.S. Census Bureau and FAA data
Appendix. Agency Comments 72
Appendix. Agency Comments
U.S. Department
of Transportation
Deputy Assistant Secretary 1200 New Jersey Avenue, S.E.
Washington, D.C. 20590
Office of the Secretary
of Transportation
MEMORANDUM
Date: April 27, 2020
Subje
ct: INFORMATION: Management Response to Office of Inspector General (OIG) Draft
Report on Airline Service to Small- and Medium-Sized Communities
From: David E. Short
Deputy Assistant Secretary of Transportation for Aviation and International Affairs
To: Ch
arles A Ward, Assistant Inspector General for
Audit Operations and Special Reviews
The stated objective of the OIG draft report is “to detail recent aviation industry trends, particularly as
they relate to service to small- and medium-sized communities.” The Department neither endorses the
conclusions reached in this self-initiated audit nor has it verified the data, methodology, or quantitative
analysis in the report. OIG did not provide sufficient, appropriate evidence to provide a reasonable basis
for its findings and recommendations based on its audit objectives. Specifically, the report’s
recommendations focus on concerns relating to ancillary fees, yet the linkage between ancillary fees and
service to smaller communities is tangential at best, since the fees on which the report focuses are
primarily charged by airlines serving mostly larger communities.
Upon review of the OIG’s draft report, we concur with Recommendation 1 to issue a Reporting
Directive clarifying that air carriers are to include booking fees, along with any/all fees required to
board the aircraft, in the fare line item reported to the Office of Airline Information’s Origin and
Destination Survey (“O & D Survey”). The Survey is the primary source of ticketed itinerary price
information for scheduled airline services in the United States. Recommendation 1 highlights
inconsistent reporting by some airlines who have failed to include all of the charges and fees a
passenger must reasonably pay to board the aircraft. The Department strives to ensure the accuracy
of the Survey as the industry evolves, and will accordingly complete actions to implement
Recommendation 1 by December 31, 2020.
The Department does not concur with Recommendation 2 to assess competitive conditions in the
airline industry and to monitor risks to the Airport and Airway Trust Fund (AATF) by developing a
process to regularly collect, maintain and use information from airlines’ website disclosures of all
fees charged for optional or ancillary services as a screening mechanism for significant changes in
these fees. As part of its ongoing mandate, the Department already monitors changes in the airline
industry, including ancillary fees, product unbundling and, now re-bundling, to ensure that the
Appendix. Agency Comments 73
Department’s analysis of airline competition, as well as its policies, remain consistent with
commercial developments. The implementation of Recommendation 1 will address the
first concern OIG has raised, which focuses on how ancillary fees influence competition in
the airline industry. The O&D Survey is the primary source of information on the cost of
passenger air travel throughout the national air transportation system. By clarifying that
passenger fares reported in the Survey include all charges that a passenger must reasonably
pay to board the aircraft, the Department will ensure that the basic cost of air transportation
will be fully accounted for despite dynamically changing industry trends to unbundle
service offerings, or (more recently) to re- bundle the product.
The second concern OIG identified as a basis for Recommendation 2 is the risk ancillary fees,
which are not subject to airline ticket taxes, pose to the Airport and Airway Trust Fund. As the
draft report acknowledges, the Department is not responsible for determining taxable charges that
fund the AATF nor for forecasting taxable airline revenues. The U.S. Treasury is responsible for
forecasting tax receipts that go into the Airport and Airway Trust Fund as well as determining
which airline charges are subject to airline ticket taxes. In general, it is the U.S. Treasury’s
decision to determine whether there is a need to monitor airlines’ ancillary fees to determine risks
to their forecasting methodologies.
The Department also does not concur with Recommendation 3 for the Department to request a
Revenue Ruling or policy statement from the Department of Treasury regarding the taxation of
airline booking fees. The predicate for the request for investigation is OIG’s own study and the
Department believes OIG is best suited to represent its own findings and recommendations to the
Treasury Department directly. The FAA monitors receipts due to the AATF based on the
parameters established by the Treasury Department. The Department believes that this division of
responsibilities is appropriate and eliminates potential conflicts of interest that could occur among
the Department’s primary stakeholders. For example, while increasing the scope of airline charges
to be taxed could increase receipts under the AATF, increased taxation could be implemented in
ways that discriminate among airline business models, or could reduce overall demand for air
transportation (especially on the margins where some airlines, like ultra-low-cost carriers, compete
for passengers who would otherwise not travel). The Department believes that the current structure
for administering the AATF is appropriate and is not inclined to take actions inconsistent with the
existing organizational framework and division of responsibilities among the agencies involved.
We appreciate this opportunity to respond to the OIG draft report. Please contact Madeline M.
Chulumovich, Director, Audit Relations and Program Improvement at (202) 366-6512 if you have
any questions or require additional information about these comments.
Our Mission
OIG conducts audits and investigations on
behalf of the American public to improve the
performance and integrity of DOT’s programs
to ensure a safe, efficient, and effective
national transportation system.
