Copyright
Severin Borenstein.

Do airlines in Chapter 11 harm their rivals? : bankruptcy and pricing behavior in U.S. airline markets online

. (page 1 of 2)
Online LibrarySeverin BorensteinDo airlines in Chapter 11 harm their rivals? : bankruptcy and pricing behavior in U.S. airline markets → online text (page 1 of 2)
Font size
QR-code for this ebook


HD28






DO AIRLINES IN CHAPTER 11 HARM THEIR RIVALS?:

BANKRUPTCY AND PRICING BEHAVIOR

IN U.S. AIRLINE MARKETS

Severin Borenstein
Nancy L. Rose

Massachusetts Institute of Technology
WP#3798 February 1995



Do Airlines in Chapter 11 Harm Their Rivals?:

Bankruptcy and Pricing Behavior in U.S. Airline Markets

by

Severin Borenstein* axid Nancy L. Rose**

Februajy 1995



Abstract: The behavior of firms in financial distress has attracted considerable academic
emd policy interest in recent years. The turmoil in the U.S. airline industry has triggered
much of the public poUcy discussion, as some observers have argued that airlines in financial
distress, paxticulaxly those operating under Chapter 11 bankruptcy protection, reduce
prices to the point of harming themselves and their competitors. This study investigates
the pricing strategies of bankrupt airlines and their rivals. The data suggest that an
airUne's prices typically decHne somewhat before it files for baxikruptcy protection and
remain slightly depressed over the subsequent two or three quarters. We find no evidence
that competitors of the bankrupt adrline lower their prices, however, nor that they lose
passengers to their bankrupt rival. These restilts indicate that bankrupt carriers do not
harm the financial health of their competitors.

OP TECHNOLOGY

M4y 3 1995

l-IBRARIES



We thank Robert Gertner for helpful comments, and the National Science Foundation,
Alfred P. Sloan Foundation, auad Center for .\dvanced Study in the Behavioral Sciences for
financicd support. A shortened version of this paper is forthcoming in American Economic
Review Papers and Proceedings, May 1995.

* University of California Energy Institute, 2539 Channing Way, Berkeley, CA 94720-5180,
510-642-5145, U.C. Davis, and NBER. email: sj[email protected]

** Sloan School of Management and Department of Economics, M.I.T., Cambridge, MA
02139, 617-253-8956, and NBER. email:[email protected]



Modern economics has generated many theories of the ways in which a firm's financial
condition may affect its conduct in the product market. Though some of these imply
that a company constrained by capital structure or financial distress will compete less
aggressively, a common view among business people is that a firm in financial trouble has
■'nothing to lose" and will slash prices to "generate cash." Perhaps nowhere has this view
been repeated more often than in the airline industry. Managers at major carriers, as well
as a blue-ribbon government task force on the industry's financial woes, have argued that
financially weak airlines, and especially those under Chapter 11 bankruptcy protection,
have cut prices and harmed the financial health of the industry.^

There are a variety of channels through which financiaJ distress, and bankruptcy in
particular, might affect pricing decisions of such firms and their rivals. First, fihng for
bankruptcy protection may directly alter costs or demand for the bankrupt carrier. If
bankrupt carriers are able to lower their marginal costs through abrogation of existing
labor and equipment lease contracts, then filing for bankruptcy may lower a carrier's
preferred price on a route. Similarly, if paissengers perceive a bajikrupt carrier to offer lower
quality service and consequently reduce their demand for its flights, a bankrupt carrier may
want to lower its price. Competing airlines may choose to match these price decreases or
not, depending on how lower prices of the bankrupt carrier affect the residual demand
tacmg other airlines. Second, bajikruptcy may lead an airline to discount future revenues
more heavily. This could imply either higher current prices (ais in models with consumer
switching costs, where lower current prices can be viewed as an investment in meirket
share that generates higher profits in the future) or lower current prices (as in collusion
models, where increases in discount rates may lead to deviations from cooperative pricing
behavior). Third, bankruptcy may alter the strategic position of the firm, committing
it to more aggressive competition (e.g., by inducing a preference for greater risk) or less
aggressive competition {e.g., by increasing liquidity constraints or otherwise constraining
managerial actions). Finally, bankruptcy may invite predatory behavior by financicdly



The Report of the National Commission to Ensure a Strong Competitive Airline Industry states that
"The Commission recognizes that the financial problems facing the airline industry have been caused
by a number of factors and that bankrupt carriers have been one of those factors." (p. 15).



healthy rivals because it may Umit the ability of the bankrupt firm to finance a costly
price war.

Despite considerable theoretical work, there have been few empirical tests of financial
or capital structure effects on product maxket behavior.^ This study contributes to the
empirical evidence by reporting the effect of bankruptcy announcements on pricing behav-
ior in the U.S. airhne industry. We use data from the seven Chapter 11 bankruptcy fihngs
by large U.S. air carriers between 1989 and 1992: Eastern (March 1989), Braniff (Septem-
ber 1989), Continental (December 1990), Pan Am (January 1991), Midway (March 1991),
America West (June 1991), and TWA (January 1992). We focus primarily on the four
largest of these, as measured by the number of affected domestic routes: Eastern, Conti-
nental, America West, and TWA. We find Uttle evidence that bankruptcy per se affects an
airUne's pricing behavior, although financial distress that culminates in a bankruptcy filing
appears to be associated with somewhat lower prices by the distressed carrier. Among the
four major bankruptcies, only one airhne - Eastern - appears to have significantly reduced
its prices subsequent to filing for Chapter 11, and this change may be confounded with the
effects of a contemporaneous strike against the airline. We find no evidence that earners
competing with bankrupt airlines cut their prices on overlapping routes subsequent to a
bankruptcy filing, even for the Eastern bankruptcy.

The analysis in this paper is based on fares recorded in the Department of Trans-
portation's 10% ticket sample (Databank lA) for the 26 quarters from 1987:1 through
1993:2.'' We first detail the pricing behavior of bankrupt airiines and their rivals for each
of the four major bajikruptcy events. Regression ajialysis of price changes for the full set
of bankruptcies follows.

I. Pricing Behavior Around Major Bankruptcy Events

Figures l(a-d) present price trends around the four major airhne bemkruptcies. For



^ See Chevalier, 1994, Chevalier and Scharfstein, 1995, and Kovenock and Phillips, 1995, for examples.

■* Details of the dataset construction are available from the authors, and are virtually the same as those
described in Severm Borenstem and Nancy Rose (1994). A routes was included only if there were at
least 90 passengers recorded in each of the quarters analyzed.



each quarter, we compare the airhne's sampled ticket prices to the average price for all
domestic tickets on routes in the same 100-mile distance block. A value of zero for this
"normalized price" reflects faxes equal to the (distance-adjusted) overall domestic average
price for that quarter. The solid lines in each figure trace the normalized prices for the
bankrupt carriers; the da.shed lines trace the normalized prices averaged across all non-
bamkrupt competitors present on the same routes. The bold vertical Hne indicates the
quarter in which the bankruptcy filing occurred.

Looking first at the effect of a bainkruptcy filing on an airline's own pricing behav-
ior, we note that only Eastern exhibits a marked change in pricing patterns subsequent
to its bankruptcy filing. Prior to its March 1989 bankruptcy amnouncement, Eastern's
normalized prices were stable at about 5% less than industry averages. The decline in its
prices after bankruptcy was large (dropping to 15% to 25% below industry average fares)
and apparently permanent, although this may confoimd bankruptcy effects with those of
a labor strike during the same period.'* This stands in contrast to the three other ma-
jor bankruptcies we analyze. Continental, TWA and America West all appear to have
reduced fares relative to previous trends about 6 months before their bankruptcy filings;
these returned to trend within 6 months for TWA, and within a year for America West.
Continental fcires continued to decline relative to industry average fares through the end of
our sample period. With the exception of Eastern, these figures provide little evidence of
permanent price changes induced by the bankruptcy fiHng, and suggest that financial dis-
tress rather than bankruptcy per ae may be responsible for observed changes in an airline's
price preferences.

The pricing behavior of non-bankrupt carriers on routes served by each bankrupt firm
is represented by the dashed lines in figures l(a-d).' These do not support the contention
that bankrupt airlines have forced down the prices of their competitors. For all four
major bankruptcy events, competitors raised normalized prices in the quarter their rival

The Eastern strike in the second quarter of 1989 virtually shut dovm the airline and created significant
negative press coverage. We cannot distinguish Eastern's change in pricing strategy around the
bankruptcy from the effects of this strike.

All tickets of non-bankrupt carriers on a route are included in these calculations if the bankrupt
carrier has at least a 10% share of traffic on the route during the quarter.



declared bankruptcy; in 3 of the 4 ceises, competitors' prices increcised further in the quarter
following the bankruptcy filing.

While these figures provide a convenient summary of relative price trends, they do
not control for changes in the mix of routes over time. Route changes could pose a partic-
ular problem for this analysis, since the financially distressed airlines tend to exit routes
completely or drop below the 5% route share threshold we use to denote routes on which
a caxrier is "active". We therefore next calculate average changes in distance-adjusted
normalized prices for the a matched set of routes.^ Table 1 reports these results for the
bankrupt airlines. The first row indicates the level of normalized prices two quarters be-
fore the bankruptcy filing; the subsequent rows report changes in prices relative to this
benchmark over the following four quarters.' The number of routes used in each price
caiculation axe bracketed.

The Eastern column, for example, shows that on the 744 Eastern routes in the third
quarter of 1988 (1988:3, two quarters before Eastern filed for bankruptcy), Eastern's prices
were 0.9% below industry average, controlling for distcuice. Eastern's normalized prices in
1989:1, the quarter it filed for bankruptcy protection, had increased by 1.7% for the 593
routes on which it was still active. Data on prices in the quarter immediately following the
bankruptcy filing axe not very informative, since Eastern virtually shut down in 1989:2 (it
remained active on only 22 routes). By 1989:3, however. Eastern had returned to significant
operations (with 564 active routes), and its normalized prices were 16.9% lower than they
had been a year earlier. While Eastern substantially lowered its prices subsequent to its
bankruptcy filing, this is not true of the three other major bankruptcies. Continental,



^ The normalization is now relative to prices of non-bankrupt carriers on routes in the same 100-mile
distance category that have no bankrupt (or soon to be bankrupt) carrier present. This approach still
does not adjust for changes in "traffic mix," such as decreases in average yield for bankrupt carriers
due to a decline in the proportion of higher fare business travelers. For a carrier in bankruptcy, this
effect will lower the average price reported if, as is often suggested, business travelers are the first to
abandon financially distressed airlines. Similarly, this effect could generate increased average prices
for competitors of a bankrupt airline as a result of a "richer" traffic mix.

'' To assure that the bankrupt carrier has been a significant competitor in a route, a route is included
in the analysis if the bankrupt carrier has at least a 10% route share two quarters before the quarter
in which it declares bankruptcy. We calculate price changes on the route whenever the bankrupt
carrier has at least a 5% route share.



-0.009


+0.089


-0.306


-0.054


[744


679


[284


[419]


+0.031


-0.050


-0.118


-0.043


(0.004)
665


(0.004)
674


(0.009)
278


(0.005)
416


+0.017


-0.065


-0.037


-0.053


(0.005)
593


(0.005)
657


(0.009)
[259


(0.008)
412


-0.223


+0.002


+0.034


-0.043


(0.041)
22


(0.007)
659


(0.010)
257


(0.010)
407


-0.169


-0.012


+0.048


-0.000


(0.007)
[564


(0.006)
[657]


(0.015)

241


(0.011)
[380]



Table 1
Average Price Changes of Chapter 11 Airlines

Airline Eastern Continental America West TWA

Relative Price at t-2

Change t-2 to t-1
Change t-2 to t
Change t-2 to t + 1
Change t-2 to t+2

Standard error of average in parentheses. Number of markets included in brackets.

American West, and TWA prices appear to have declined prior to bankruptcy, but these
decHnes were laxgely offset in later queo-ters. For these three airUnes, prices were nearly as
high or higher two quarters following bankruptcy as they had been two quarters prior to
the bemkruptcy filing.

Table 2 presents the results of this analysis for rivals of each bankrupt airline.® The
Continentad column, for example, indicates that on the 602 routes on which Continental
was still active cind faced at least one significant competitor in 1991:2 (two periods after
declaring bankruptcy), passenger-weighted average prices of competitors had risen by 0.6%
over the previous four quarters, relative to routes on which no bemkrupt carriers were
present. Overall, competitors' prices do not tend to decline subsequent to bankruptcy,
although America West's rivals (of which Southwest AirUnes is most significant) appear
to have matched its price cut just prior to its bankruptcy filing. Within two quarters
following each bankruptcy filing, competitors prices are about as high (for Continental



The number of routes in this table are a subset of those in table 1, since for some of the routes the
bankrupt carrier has no significant (5% minimum share) competitor.



Table 2
Price Changes of Competitors of Chapter 11 Airlines

Airline Eastern Continentai America West TWA

Relative Price at t-2 +0.021 -0.009 -0.363 -0.067

[739] [645] [261] [403]

Change t-2 to t-1 +0.001 +0.010 -0.129 +0.015

(0.003) (0.003) (0.015) (0.004)

[658] [633] [251] [399]

Change t-2 to t +0.027 -0.009 -0.018 +0.043

(0.004) (0.005) (0.009) (0.008)

[586] [608] [228] [392]

Change t-2 to t+1 -0.020 +0.024 +0.057 +0.007

(0.023) (0.007) (0.010) (0.007)

[21] [608] [222] [387]

Change t-2 to t+2 +0.071 +0.006 +0.020 +0.005

(0.005) (0.007) (0.010) (0.009)

[561] [602] [207] [359]

Standa.rd error of average in parentheses. Number of markets included in brackets.



and TWA) or higher (for Eastern and America West) than they were two quarters prior
to the bankruptcy filing. As in the figures presented eaxUer, the evidence does not support
the claim that bankrupt airhnes have forced competitors to lower their prices.

These ancdyses average the prices across all rivals, although theoretical models and
some popular accounts suggest that price responses might differ. Models of predatory
behavior, for example, suggest that financially healthy ("deep pocket") carriers are more
Ukely to undertake aggressive pricing against weaker rivals. Similarly, executives at some
airlines have been more outspoken thaji others in blaming the indxistry's financial difficul-
ties on the behavior of financially weak carriers. Table 3 reproduces the last row of Table 2
individually for each of the six largest U.S. airlines that have not entered Chapter 11 since
deregulation. American Airlines, probably the most vocal critic of airiines operating under
Chapter 11 protection, has raised its post-bankruptcy prices relative to industry norms in
3 of the 4 bankruptcies, the 3 cases in which American overlapped on the most routes.
US Air, the financially weakest of the major airlines that have not sought Chapter 11 pro-
tection, raised its prices in response to all 4 bankruptcies, although the average increase is



Table 3
Average Price Changes of Individual Competitors over [-2, +2] Window



Competitor

American



Delta



Northwest



Southwest



United



USAi



ir



Ecistern



+0.090
(0.015)
[205]

+0.068
(0.005)

[452]

+0.105
(0.023)

[71]

-0.438

[if

-0.014
(0.019)

[48]

+0.112
(0.012)
[89]



Bankrupt Airline

Continental America West



+0.037
(0.009)

[421]

+0.046
(0.009)
[327]

-0.018
(0.021)

[114]

-0.028
(0.019)

[68]

-0.082

(0.009)

[252]

+0.072
(0.014)
[134]



-0.051

(0.012)

[119]

+0.063
(0.021)

[98]

+0.368
(0.035)
[30]

+0.012
(0.011)

[44]

-0.039

(0.013)

[120]

+0.020
(0.041)
[19]



TWA



+0.027
(0.014)

[294]

+0.011
(0.012)

[181]

+0.007
(0.025)

[123]

+0.045
(0.026)

[32]

+0.019
(0.014)

[184]

+0.035
(0.022)
[88]



Standard error of average in parentheses. Number of markets included in brackets.



significant in only 2 of these cases. Only United shows a significant price cut in response to
more than one of the four bankruptcies, and Delta, one of the financially strongest czirriers
during this period, exhibits consistently positive price responses. Overall, there appears
to be no relationship between financial health and price responses to rivals' bankruptcies.

A final possibility we investigate is whether bankrupt ctirlines harm competitors by
mcreasing market share through lower prices, service improvements, or special non-price
promotions. We first use matched-route analysis similar to the approach used to analyze
fares in the previous tables.^ Table 4 presents these route average market shares from 2



In this case, however, we use a consistent market share cutoff of 10% for inclusion of a route in order
to avoid a selection bias that would result if the inclusion criterion in period t-2 were different from
the inclusion criterion in later periods.



Table 4
Average Market Share Changes of Chapter 11 Airlines

Airline Eastern Continental America West TWA

Average Share at t-2 0.352 0.409 0.432 0.329

[744] [679] [284] [4191

Change t-2 to t-1 -0.026 +0.019 -0.016 +0.028

(0.003) (0.002) (0.005) (0.004)

[599] [626] [272] [388]

Change t-2 to t -0.116 -0.009 -0.011 +0.016

(0.004) (0.003) (0.006) (0.006)

[438] [572] [256] [381]

Change t-2 to t+1 -0.289 -0.012 -0.046 -0.017

(0.032) (0.004) (0.008) (0.007)

[12] [560] [240] [339]

Change t-2 to t+2 -0.120 -0.003 -0.052 -0.029

(0.008) (0.004) (0.008) (0.007)

[453] [586] [231] [319]

Standard error of average in parentheses. Number of markets included in brackets.
10% minimum route share threshold used in all periods.



quarters before through 2 quarters after the carrier files for Chapter 11 protection. These
averages are weighted by market size, but the results are very similar if the markets are
equzdly weighted.

In all four of the major bankruptcies, the filing carrier's average market share on
routes it continues to serve with a (minimum 10% share) either remains about constant,
in the case of Continental, or decHnes over the bankruptcy period. For example, two
quarters before its bankruptcy declaration, Continental was active on 679 routes, carrying
an average of 40.9% of the passengers. Two quarters after its bankruptcy declaration,
Continental had dropped below a 10% share on 93 of those routes, and its market share
on the remaining 586 decUned by an average of 0.3%. Eastern, America West, and TWA
all a higher proportion of their original routes than did Continental and experienced more
substantial declines in market share on routes they continued to serve subsequent to their
bankruptcy filings.

To investigate whether airUnes declaring bamkruptcy harm competitors by entering
new routes, we excimine the bankrupt carrier's share of trciffic on all "relevant" routes.



0.158


0.117


0.165


0.063


0.131


0.125


0.169


0.072


0.096


0.114


0.170


0.076


0.014


0.112


0.162


0.068


0.063


0.123


0.153


0.063


2263


3053


850


2225



Table 5
Aggregate Market Share of Chapter 11 Airlines in All Relevant Markets

Airline Eastern Continental America West TWA

Share at t-2
Share at t-1
Share at t
Share at t + 1
Share at t-l-2
Number of Markets



defined as ajiy route on which the bcinkrupt carrier had passengers in any quarter of the
[-2, +2] window. Table 5 reports the carrier's share of all traffic on this set of routes in each
of the quarters.^" As expected, Eastern's share drops substantially around the bankruptcy
declaration. For the other three airhnes, share rises just prior to declaring bankruptcy,
consistent with the price cuts that they carry out at that time, but then decUnes after
entering Chapter 11. Overall, Continental's shaire rises sHghtly (0.6 percentage points) from
t-2 to t-l-2, TWA's is unchanged, and America West's share falls sUghtly (1.2 percentage
points). Thus, although competitors of beinkrupt airUnes maintain or increase their prices
subsequent to these bankruptcy filings, competitors' market shares appear to increase or
remadn about constant.

II. Econometric Analysis of Bankruptcy Effects on Airline Prices

In this section, we report estimates from an econometric model of airline prices, using
data on all 7 bankruptcy filings during our sample period of 1988:1 through 1992:4. We
model the quarter-to-quarter chainge in log average price for an airline on a given route
as a function of: current and one-period lagged changes in the passenger-based Herfindahl
index, to control for changes in market structure; the lagged change in the log price for
the airUne-route, to control for the substantial negative serial correlation in price changes;



" These calculations include all routes on which the bankrupt carrier had a minimum 1% market share
in any of the quarters, regardless of route size. The computed shares are for all passengers on these
routes, however, so a very small route has very little influence on the calculation.



fixed time effects, to control for general airline price movements over time; and a set
of bankruptcy measures described below. The model implicitly assumes that exogenous
changes in the remaining variables typically included in cross-sectioned models of airline
price levels - such as airport congestion, market density, network interconnectedness, and
airport dominance - are sufficiently small that they can be excluded from the specification.

We include two sets of indicator variables designed to capture the effects of impending
or recently declared Chapter 11 bankruptcy. The first set measures the average change
in price for a bzmkrupt airline; the second measures the average change in price for non-
bankrupt airhnes on routes with a near-bankrupt or bemkrupt competitor. We scale these
variables so that their coefficients reflect price changes over 4 three-month periods: 180
to 90 days and 90 to days before a bankruptcy filing, and to 90 days and 90 to 180
days after a bankruptcy filing. ^^ The sum of the price changes over these four periods
indicates the net change in price dxiring the window beginning 6 months before and ending
6 months after the date of Chapter 11 fihng. None of the airUnes in our sample exit
Chapter 11 during our observation period, although bankrupt airlines do exit routes or
drop below the 10% share threshold used in the regression analysis.

This model is estimated by unweighted ordinary least squares using the 1777 routes
for which we observed at leaist 300 passengers during each quarter from 88:1 through 92:4.
The standard errors are adjusted for heteroskedasticity and contemporaneous correlation
of residuals across carriers on the same route. Unhke the figtires and previous tables,
observations axe not weighted by the number of passengers in this regression as this would
be likely to induce heteroskedasticity.

The results in table 6 suggest that airlines that file for bankruptcy protection cut their



^^ The first set of variables indicate bankruptcy status for the observed carrier. We define BANK-
RUPT-2 as the mean within a quarter of a dummy variable that is zero until 180 days before the


1

Online LibrarySeverin BorensteinDo airlines in Chapter 11 harm their rivals? : bankruptcy and pricing behavior in U.S. airline markets → online text (page 1 of 2)