Airline pricing behavior under limited inter-modal competition.
Bergantino, Angela Stefania ; Capozza, Claudia
I. INTRODUCTION
There are three sources of competition in the airline market for
short-haul flights which jointly affect fares. Airlines compete with
other airlines for the same city-pair markets (intra-modal competition).
Moreover, airlines compete with other modes of transport (inter-modal
competition) such as trains, especially high speed trains, and cars,
which have the advantage of allowing travel at any time. Finally,
airlines compete with themselves by setting different fares in different
time periods prior to departure. This pricing strategy is known as
inter-temporal price discrimination (IPD).
Past empirical contributions exploring pricing behavior and
competition in air transportation were not able to control for the
effect of inter-modal competition which, we can expect, affected the
results. This paper differs from existing works as it attempts to study
airline pricing for short-haul flights in contexts with no credible
threat of inter-modal competition in order to shed light on pricing
behavior in response to the pure air-related competition. To this end,
we analyze a market, southern Italy, which definitely shows a highly
limited degree of inter-modal competition. For the connections
considered, in fact, services by alternative modes, including road
transport, require, on average, more than seven times the same traveling
time as airline connections. Thus, for these peripheral areas, air
transport is often the only realistic alternative. We can assume,
therefore, that airline pricing strategies are the straight result of
air-related competition. The pricing behavior of airline companies also
shows high variability of fares per mile that unlikely can be justified
by cost differentials. The fare differentials might, instead, be
considered as evidence of different degrees of market power with the
capacity to determine mark-ups. (1)
In this paper, we address two issues. The first is to measure the
extent to which intra-modal competition determines fares. The second is
to shed light on the inter-temporal profile of fares in order to verify
whether airlines engage in IPD, and whether IPD is of the monopolistic
type or the competitive type. As for the former type, market power is
required to price discriminate as it enhances the ability of firms to
set and maintain higher mark-ups (Tirole 1988). As for the latter type,
market power is not required to sustain price discrimination if
consumers show heterogeneity of brand preferences (Borenstein 1985 and
Holmes 1989) or demand uncertainty about departure time (Dana 1998).
The dataset we use to address the research question is unique. It
covers routes that originate in southern Italy and that operated from
November 2006 to February 2011. Data on fares were collected from
airline websites to replicate consumer behavior when making
reservations. Unlike previous contributions, we simulate the purchase of
round-trip fares instead of one-way fares. In this way, we effectively
replicate the demand side since travelers more often purchase round-trip
tickets rather than one-way tickets. In addition, we precisely recreate
the supply side as we can clearly see if, for each round-trip flight, a
carrier is a feasible alternative for travelers and is an effective
competitor.
Our results on short-haul markets with no alternative modes of
transport show that when there is less intra-modal competition, airlines
set higher fares since they exploit the greater market power arising
from a concentrated market structure. Specifically, a 10% increase of
the market share allows carriers to post up to 6.4% higher fares.
Consistent with the implementation of IPD, we find a non-monotonic
profile of fares--which can be roughly approximated by a 7-curve--with a
turning point included in the interval of the 43rd to 45th days before
departure. We give new interpretations for the non-monotonicity of
fares' inter-temporal profile, in addition to the existing ones.
Indeed, on the one hand, the non-monotonicity would be the evidence that
airlines exploit consumer-bounded rationality. Actually, a common wisdom
among travelers is "the later you buy, the more you pay for the
ticket," thus price-sensitive consumers tend to buy in advance.
Airlines, aware of this, can extract a greater surplus by setting
moderately higher fares for very-early purchasers who will buy the
tickets believing they are paying the cheapest fares. On the other hand,
a higher fare for very-early purchasers can be seen as a fee for risk
aversion. Indeed, risk-averse travelers usually plan the trip well in
advance as they would like to reduce travel uncertainty by making the
best choice in terms of departure day and time. Therefore, airlines, by
posting moderately higher fares at the very beginning, can extract an
additional surplus from risk-averse travelers. Finally, we provide
evidence of a competitive-type IPD, as airlines seem to be more likely
to engage in IPD in more competitive markets.
The remainder of the paper unfolds as follows. In Section II we
survey the relevant literature. In Section III we present the empirical
strategy and in Section IV we give a description of the data. In Section
V we discuss the results, and in Section VI we draw conclusions. The
robustness check is provided in the Appendix.
II. LITERATURE REVIEW
The literature on which the current work is based concerns pricing
in air transportation and the factors influencing it. We initially
review papers that analyze the effect of the airline market structure on
fares, then we focus on works looking at price discrimination and, in
particular, at IPD. We conclude the survey with contributions exploring
the relationship between market structure and price discrimination.
The first to study the impact of market structure on fares was
Borenstein (1989) on the U.S. airline industry. He develops a model
using market share at both route and airport level. Results indicate
that market share, whatever measure adopted, influences the
carrier's ability to raise fares as the dominant presence of an
airline at an airport increases its market share on the routes included
in that airport. However, Evans and Kessides (1993) point out that, when
controlling for inter-route heterogeneity, market share on the route is
no longer relevant in determining fares that are, instead, determined by
carriers' market share at the airports. More recently, some
contributions explore the European airline markets. Unlike the U.S.
market, Carlsson (2004) finds that market power, measured by the
Herfindahl index, does not have a significant effect on fares whereas it
influences flight frequencies. Consistent with this, Giaume and Guillou
(2004) find a negative and, often, non-significant impact of market
concentration for connections from Nice Airport (France) to European
destinations. Bachis and Piga (2007a) measure the effect of market
concentration at the origin airport on fares applied by British
carriers, considering either the route or the citypair level. Their
results reveal the existence of a large degree of substitutability
between the routes within a city-pair. A greater market share at the
route level leads to higher fares, while at city-pair level it does not.
Gaggero and Piga (2010) find that a higher market share and the
Herfindhal index at the city-pair level lead to higher fares on routes
connecting the Republic of Ireland to the United Kingdom. Finally,
Brueckner et al. (2013) provide a comprehensive analysis of competition
and fares in domestic U.S. markets, focusing on the roles of low-cost
carriers (LCCs) and full-service carriers (FSCs). They find that FSC
competition in an airport-pair market has a limited effect on fares,
while competition in a city-pair market has no effect. In contrast, LCC
competition has a strong impact on fares, whether it occurs in
airport-pair markets or in city-pair markets.
Concerning price discrimination, the main difference between static
and IPD is that two different markets are covered in the former case,
whereas the same market is periodically covered in the latter case. In a
theoretical model with two time periods, Lofgren (1971) shows that, for
the same good, a seller applies higher prices to consumers with higher
purchasing power in the first period and lower prices to consumers with
lower purchasing power in the second period. Stokey (1979) implicitly
extends Logfren's framework to continuous periods. She claims that
IPD occurs when goods are "introduced on the market at a high
price, at which time they are bought only by individuals who both value
them very highly and are very impatient. Over time, as the price
declines, consumers to whom the product is less valuable or who are less
impatient make their purchases." (2) In her paper, reference is
made to commodities such as books, movies, computers, and related
programs. The concept, however, has had application to the airline
industry where IPD consists of setting different fares for different
travelers according to how far in advance the ticket is bought. However,
in the airline industry, different from markets for commodities, the
inter-temporal profile of fares is increasing. Using IPD, airlines
exploit travelers' varied willingness to pay and demand uncertainty
about departure time. Price-inelastic consumers, usually business
travelers, most often purchase tickets close to departure date, while
price-elastic consumers, usually leisure travelers, tend to buy tickets
in advance. Travelers' heterogeneity appears to be a necessary
condition to successfully implement price discrimination strategies. In
a theoretical contribution, Alves and Barbot (2009) illustrate that
low-high pricing is a dominant strategy for LCCs only if travelers, on a
given route, show varied willingness to pay. Actually, Gale and Holmes
(1992, 1993) prove that, through advance-purchase discounts, a monopoly
airline can increase the output by smoothing consumers' demand with
weak time preferences over flight times and can extract the surplus of
consumers with strong preferences. More recently, Moller and Watanabe
(2010) investigate further on advance-purchase discounts versus
clearance sales, showing that the former pricing strategy is preferred
to the latter for airline tickets because their value is uncertain to
buyers at the time of purchase, and reselling is costly or difficult to
implement.
The inter-temporal profile of fares has been also empirically
explored. McAfee and te Velde (2007) find out that 1 week before the
departure there is a significant rise in fares, which is on the top of
the rise of 2 weeks before the departure. Bachis and Piga (2007a) show
that fares posted by British LCCs follow an increasing inter-temporal
profile. Instead, Bachis and Piga (2007b), who examine UK connections to
and from Europe, and Alderighi and Piga (2010), who focus on Ryanair
pricing in the UK market, find a U-shaped fare inter-temporal profile.
Gaggero and Piga (2010) show that fares for Ireland-UK connections
follow a 7-curve. Gaggero (2010) argues that there are three categories
of travelers: early bookers and middle bookers, usually leisure
travelers, and late bookers, mostly business travelers. Early bookers
have a slightly inelastic demand. Families planning holidays are, for
instance, willing to pay moderately higher fares to travel during
vacations. Middle bookers exhibit the highest demand elasticity as they
are more flexible and search for the cheapest fares. Late bookers reveal
an inelastic demand. A business traveler typically books the ticket a
few days before departure, with fixed travel dates and destination. As a
result, the fare inter-temporal profile is 7-shaped as it reflects a
pattern opposite to that of travelers' demand elasticity. (3)
One strand of literature explores the relationship between market
structure and price discrimination to find out whether airlines are more
willing to engage in price discrimination strategies when markets are
more or less competitive. Traditionally, market power enhances the
ability of firms to price discriminate. A monopolist can set and
maintain higher mark-ups. (4) In the oligopolistic airline industry,
when competition increases, carriers lose this ability. Markups
associated with the fares paid by the less price-sensitive (business)
travelers decrease and align with the ones of the more price-sensitive
(leisure) travelers. However, Borenstein (1985) and Holmes (1989) show
that market power is not required to sustain price discrimination if
consumers show heterogeneity of brand preferences. Business travelers
prefer the long-run savings given by loyalty programs, while leisure
travelers disregard carriers for short-run savings. Sorting consumers
based on strength of brand preference is a successful strategy, and
competition does not prevent firms from pursuing it. When competition
increases, the mark-ups applied to leisure travelers decrease, whereas
the mark-ups applied to business travelers remain almost unchanged. As a
result, price discrimination increases as competition increases.
Further, Gale (1993) proves that competition to conquer less
time-sensitive travelers is stronger in an oligopoly than in a monopoly.
Competition reduces fares on the lower end of the distribution, thus
enhancing price dispersion. Finally, Dana (1998) shows that price
discrimination, in the form of advance-purchase discounts, does not
require market power to be implemented. Consumers with more certain
demands are willing to buy in advance because the presence of consumers
with less certain demand could lead to an increase in prices.
Some empirical papers consider price dispersion as the result of
price discrimination. Borenstein and Rose (1994) explore the U.S.
airline industry and provide evidence of competitive-type price
discrimination: lower price dispersion arises in more concentrated
markets. Consistent with this, Carbonneau et al. (2004) show that more
competition is correlated with more price dispersion. Later, Gerardi and
Shapiro (2009) revisit Borenstein and Rose's (1994) analysis. They
find the same results when they replicate Borenstein and Rose's
(1994) cross-sectional model. However, they have opposite results when
performing a panel analysis. (5) Indeed, they provide evidence of
monopolistic-type price discrimination: higher price dispersion arises
in more concentrated markets.
Stavins (2001), instead, measures price discrimination through
ticket restrictions. (6) Consistent with Borenstein and Rose (1994), she
provides evidence of competitive-type price discrimination in the U.S.
airline industry: ticket restrictions reduce fares although the effect
is lower for more concentrated markets. Using the cross-sectional model
of Stavins (2001), Giaume and Guillou (2004) attain the same results on
intra-European connections. (7)
Gaggero and Piga (2011) provide a seminal contribution on the
effect of market structure on inter-temporal pricing-dispersion focusing
on the routes connecting Ireland and the United Kingdom. Consistent with
Gerardi and Shapiro (2009), they find that few companies with a large
market share can easily price discriminate.
In contrast to the aforementioned contributions, Hayes and Ross
(1998) find no empirical evidence of price discrimination and market
structure in the U.S. airline industry. Price dispersion is due to peak
load pricing and it is influenced by the characteristics of the carriers
operating on a given route. Consistent with this, Mantin and Koo (2009)
highlight that price dispersion is not affected by the market structure.
Instead, the presence of LCCs among the competitors enhances dispersion
by inducing FSCs to adopt a more aggressive pricing behavior. (8)
III. EMPIRICAL STRATEGY
We define two models. The baseline model accounts for the effect of
market structure and IPD on fares. The extended model allows for IPD to
vary with market structure. (9)
The baseline model:
(1) Ln ([P.sub.it]) = [[beta].sub.0] + [[beta].sub.1] Market
[Structure.sub.i]
+ [[beta].sub.] Booking [Day.sub.t]
+ [[theta].sub.3] Flight [Charcteristics.sub.i]
+ [[theta].sub.4] Control [Dummies.sub.it] +
[u.sub.it]
the extended model:
(2) Ln ([P.sub.it]) = [[beta].sub.0] + [[beta].sub.1] Market
[Structure.sub.i]
+ [[beta].sub.] Booking [Day.sub.t]
x Booking [Day.sub.t])
+ [[theta].sub.3] Flight [Charcteristics.sub.i]
+ [[theta].sub.4] Control [Dummies.sub.it] +
[u.sub.it]
where i indexes the round-trip flight and t indexes the time. Each
flight i is defined by the route, the carrier, and the date of departure
and return. We have a daily time dimension that goes from 1 to 60.
The dependent variable is the log of the fares. The variable
Booking Day captures the effect of IPD and ranges from 1 to 60. In order
to account for the potential nonlinearity of Booking Day, we also add
Booking Day squared to the model.
We use two indices of market structure at city-pair level (10):
* Market Share, the average share of the daily flights operated by
an airline at the two endpoints of a city-pair; (11) and
* Herfindahl-Hirschman Index (HHI), based on market share.
Flight Characteristics includes the following variables:
* Holiday is a peak-period dummy equal to 1 for flights occurring
during summer holidays, winter holidays, bank holidays, and public
holidays, 0 otherwise;
* LCC is a carrier dummy equal to 1 for flights provided by LCCs, 0
otherwise.
Control dummies are:
* Route dummies to capture route-specific effects, demand and cost
(or price) differences;
* Year dummies to account for macroeconomic factors equally
affecting all flights in each year;
* Month dummies to capture seasonal effects;
* Departure Time and Return Time, two sets of four categorical
dummies capturing the effect of the takeoff time: morning (6:00-10:00),
midday (10:00-14:00), afternoon (14:00-18:00), and evening
(18:00-24:00); (12) and
* Stay dummies to control for the length of stay (i.e., how many
days elapse between departure and return).
Finally, [u.sub.it] is the composite error term, where [u.sub.it] =
[[alpha].sub.i] + [[epsilon].sub.it]. Specifically, [[alpha].sub.i] is
the unobserved heterogeneity and [[epsilon].sub.it] is the idiosyncratic
error term. Standard errors are clustered at flight level as
observations on flights are not likely to be independent over time.
We want to estimate coefficients of time-invariant variables,
therefore we use the random effects (RE) generalized least square (GLS)
estimator. The RE GLS estimator, to be consistent, requires the
assumption that the right-hand side variables are not correlated with
the unobserved heterogeneity a,. Performing the robust Hausman
specification error test using the method of Wooldridge (2002), after
each regression, we can test the validity of that assumption and, hence,
the consistency of RE GLS estimates. (13)
We assume that the market structure is exogenous. Basically, we
agree with Stavins (2001), claiming that elements such as "entry
barriers prevent new carriers from entering city-pair routes (e.g.,
limited gate access, incumbent airlines' hub-and-spoke systems, and
scale economies in network size)." (14) Moreover, in the European
Union there are the "grandfather rights": an airline that held
and used a slot last year is entitled to do so again in the same season
the following year. In the short run, then, market structure can be
assumed to be fixed.
However, one might argue that capacity is always designed to
accommodate planned demand. Even in a slotted-constrained airport, if
demand is clearly below capacity, carriers adjust capacity and
reallocate flights between routes. Therefore, even if the overall
airport capacity is fixed, route level capacity is not necessarily
fixed, even in the short run. If this applies in our sample, Market
Share and HHI are potentially correlated with [[epsilon].sub.it].
Therefore, we also employ the generalized method of moments (GMM)
estimator to obtain coefficients. We use instruments designed by
Borenstein (1989) and largely adopted in the related literature. (15)
Market Share is instrumented with GENP and Log(Distance), while HHI is
instrumented with QHHI and Log(Distance).
GENP is the observed carrier's geometric mean of enplanements
at the endpoints divided by the sum across all carriers of the geometric
mean of each carrier's enplanements at the endpoint airports:
(3) GENP = [square root of [ENP.sub.k,1]] *
[ENP.sub.k,2]/[summation][square root of [ENP.sub.j,1] * [[ENP.sub.j,2]]
where k is the observed airline and j refers to all airlines.
QHHI is the square of the market share fitted value plus the
rescaled sum of the squares of all other carriers' shares:
(4) QHHI = [??] + (HHI - MS)/(1 - [MS.sup.2]) (1 - [[??].sup.2])
where MS stands for the Market Share and MS is the fitted value of
MS from the first-stage regression.
Log(Distance) is the logarithm of the distance in kilometers
between the two route endpoints.
In the extended model we add the interaction between Booking Day
and Market Share or HHI. The interaction could be endogenous too, thus
we include, as an additional instrument, the interaction between Booking
Day and GENP or QHHI, respectively.
IV. DATA COLLECTION
Data on fares were collected to replicate real travelers'
behavior when making reservations. First, we identify plausible
round-trips, then we retrieve data directly from airlines' website
by simulating reservations. (16) We observe fares daily, starting
generally at 60 booking days before departure. However, for some
round-trip flights we have less than 60 observed fares, thus the panel
is unbalanced. We define a dataset comprised of 19,605 observations on
427 round-trip flights from November 2006 to February 2011. Our sample
includes 10 city pairs (see Table 1) and 11 airline companies.
We consider both FSCs and LCCs (see Table 2); thus we choose the
basic services (no add-ons) to make carriers' supply effectively
comparable.
We simulate the purchase of round-trip tickets, which gives us
several advantages. Firstly, we effectively replicate the consumer
behavior because travelers mostly purchase round-trip tickets rather
than one-way tickets. (17) In addition to that, we precisely recreate
the market structure as we can clearly see whether, for each round-trip
flight, a given carrier is a feasible alternative for travelers and an
effective competitor. The use of round-trip fares also allows us to
account for peak periods and to verify whether airlines adjust the
pricing behavior during phases of greater travel demand. Further,
one-way ticket pricing differs depending on carrier type. For FSCs, a
round-trip fare is lower than the sum of the corresponding two one-way
fares. This pricing policy is not adopted by LCCs. To avoid distortions,
previous contributions, using one-way fares, limit the empirical
analysis to LCCs or to a few carriers. Instead, we do not encounter this
problem and we are able to carry out a market analysis and compare
pricing behavior of all carrier types. In Table 3 we provide descriptive
statistics.
Our data sample has a good deal of variation in terms of both fares
and market structure indices. In fact, we observe either monopolistic or
more competitive markets.
Further, in Table 4, we report the average fares per mile posted by
the incumbent airline providing services for the city pair included in
the empirical analysis.
From each origin, connections to Rome appear to be comparatively
more expensive than connections to Milan, even though point-to-point
distances to Rome are shorter than point-to-point distances to Milan.
This could be only partially explained by the cost of fuel. For
short-haul flights, approximately 35% of fuel is used on the takeoff.
Thus, the cost function is strictly decreasing with distance. However,
differences in fares do not seem to reflect only differences in costs,
but, instead, would suggest that the incumbent airline applies different
markups to different connections. This preliminary evidence motivates an
in-depth discussion on fares' determinants.
It is worth looking at Figure 1 that shows that the relationship
between average posted fares and days prior to departure seems to be
non-monotonic.
Airlines set the initial level of fares, subject to slight changes
for, roughly, 15 days, and then fares sharply decrease to the minimum
level. Henceforth, airlines increase fares up to the departure day. The
increment becomes steeper in the last 15 days before departure. We look
into this in depth when presenting regression results. Figure 2 shows
the density distribution of fares. The mass of values is concentrated
between 50 and 200 Euros.
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
V. RESULTS
In each regression table we report both the results from RE GLS
estimator and GMM estimator. (18) The results of the robust Hausman
specification error test do not lead to the rejection of the null
hypothesis that the RE GLS estimator is consistent. (19) Concerning GMM
estimations, we report the results of some tests. The first one concerns
the non-weakness of instruments. For all the regressions, the
Kleibergen-Paap rk statistic-the robust analog of the Cragg-Donald
statistic-is far greater than the critical value; therefore, the null of
the weakness of instruments is strongly rejected. (20) The second one is
the Hansen J test for the validity of the population moment conditions.
For all the regressions, we fail to reject the null hypothesis, that the
over-identifying restriction is valid, thus the instruments are not
correlated with the error term. Finally, the third one is the exogeneity
test for market structure variables. We fail to reject the null
hypothesis of exogeneity of either Market Share or HHI for all the
specifications. GMM estimates are also very close to the RE GLS
estimates. This underlines the robustness of the results.
Estimation results reported in the tables contained in this section
are organized as follows: columns (1), (2), (5), and (6) report
regressions' output using the variable Market Share, while columns
(3), (4), (7), and (8) report regressions' output using the
variable HHI.
Table 5 shows the results of the Baseline Model. Market Share and
HHI have a positive and highly significant impact on fares. According to
RE GLS estimates, holding constant other variables, a 10% increase in
Market Share leads to 6.4% higher fares, and a 10% increase of HHI leads
to 5.7% higher fares. Results are similar to GMM estimates. Indeed, a
10% increase in Market Share results in 6.9% higher fares, and a 10%
increase of HHI results in 8% higher fares.
Estimations are done, at first, with only the variable Booking Day.
Its coefficient is negative and significant, meaning that airlines do
engage in IPD. Indeed, fares posted the day before appear to be 1.41%
lower. We then include Booking Day squared to the regression equation to
check for the nonlinearity, as the graphical investigation suggests. The
coefficient of Booking Day squared is positive and highly significant.
Booking Day has a negative effect on fares until the turning point is
reached. Beyond that day, it has a positive impact on fares. In the
nonlinear case, the marginal effect of Booking Day on fares is dependent
on the level of Booking Day. [partial derivative]ln([P.sub.it])/[partial
derivative]Booking [Day.sub.t] = - 0.0353 + 2 X (0.0004) Booking
[Day.sub.t]. We compute the marginal effect for given values of Booking
Day which indicates how fares vary with respect to fares posted a day
early.
As shown in Table 6, from the 45th day before departure, fares
posted a day before are no longer cheaper. The marginal effects are not
statistically different from zero at days 43rd, 44th, and 45th before
departure, indicating, thus, that the minimum of the J-curve occurs in
the interval of 43rd to 45th days.
The non-monotonicity of the fare's intertemporal profile has
received various interpretations in the literature. As stated earlier in
the literature review, Gaggero (2010) explains that the non-monotonicity
is determined by travelers' demand elasticity. Indeed, he
identifies three categories of travelers with different demand
elasticity and purchase timing. Early bookers have a slightly inelastic
demand. For instance, families are willing to pay moderately higher
fares to travel during vacations. Middle bookers exhibit the highest
demand elasticity as they are more flexible and search for the cheapest
fares looking at different holiday destinations. Late bookers reveal an
inelastic demand, such as the business traveler typically booking the
ticket a few days before departure, with fixed travel dates and
destination. Instead, Bilotkach et al. (2012) claim that fare drops
occur when the actual demand is below the expectation. Therefore,
airlines might periodically reduce fares in order to respond to the need
of raising the load factor.
Although we share the previous arguments, we propose two new
interpretations. We assume, in fact, that the J-shaped fare distribution
might also be generated by other factors. Travelers generally believe
that they can save money by buying in advance. Therefore, setting
moderately higher fares for very-early purchasers seems to be a good
pricing strategy for airlines as travelers will buy the tickets anyway,
believing to be paying the cheapest fare. With this line of reasoning,
the non-monotonicity might be seen as the result of consumer-bounded
rationality.
Furthermore, travelers are also heterogeneous in terms of risk
aversion. There are risk-averse travelers who do strongly prefer to plan
the trip well in advance in order to make the best choice in terms of
departure day and time, thus reducing the overall travel uncertainty.
Therefore, higher fares at the very beginning of the inter-temporal
distribution might be considered as a fee for risk-averse travelers from
whom the airline can obtain an additional surplus.
Coefficients of the control variables are those one might expect.
The coefficient of Holiday is positive and significant. During peak
periods, airlines exploit the greater travel demand and set 21%-24%
higher fares than off-peak periods. The LLC's coefficient is
negative and significant. (21)
In regressions with Market Share, LCCs appear to price 23% lower
than FSCs, while in regressions with HHI as the predictor, LCCs appear
to price 41% lower than FSCs. The different impact is due to the
coexistence of Market Share and LCC in the same regressions. Actually,
Market Share takes lower values when a carrier is a low-cost one, thus
it already captures the effect on fares induced by LCC.
Table 7 shows the results of the Extended Model I. Booking Day is
still negative and significant, while its interaction with Market Share
or HHI is positive and significant. The negative impact of Booking Day
reduces in less competitive markets, therefore competition does not
prevent airlines from using IPD strategies.
The marginal effect of Booking Day is now given by:
* [partial derivative]ln ([P.sub.it]) /[partial derivative]Booking
[Day.sub.t] = -0.0375 + 2 (0.0004) Booking [Day.sub.t] + (0.0001) Market
[Share.sup.i],
* [partial derivative]ln ([P.sub.it]) /[partial derivative]Booking
[Day.sub.t] = -0.0374 + 2 (0.0004) Booking [Day.sub.t] + (0.00004)
HH[I.sub.i].
In Table 8, we report the marginal effects for values of Booking
Day, setting Market Share and HHI equal to the sample mean. We compare
these results with those obtained from the baseline regression (no
interaction).
In less competitive city-pair markets, the J-curve appears to be
flattened. Differences between fares posted on different booking days
are less pronounced. This finding is in favor of competitive-type price
discrimination, in line with Borenstein and Rose (1994), Stavins (2001),
and Giaume and Guillou (2004), and in contrast to Gerardi and Shapiro
(2009) and Gaggero and Piga (2011).
Table 9 illustrates the results of the Extended Model II by which
we investigate IPD further. We test whether airlines adjust their
pricing behavior during phases of a greater travel demand. To this end,
we add to the regression equation, the interaction between Booking Day
and Holiday, which has a positive and significant impact on fares.
The effect of Booking Day on fares for peak periods is 0.56%-0.64%
lower than for off-peak periods. Basically this is due to two facts. On
the one hand, the greater travel demand allows airlines to decrease IPD
because they can sell all the seats with no need of discounted fares. On
the other hand, during holidays, travelers are more homogeneous, as
people journey mainly for tourism. IPD, being based on the heterogeneity
of travelers, becomes less effective. Furthermore, we focus on IPD
strategies implemented by LCCs. To this end we employ the interaction
between the Booking Day and LCC, which has a negative impact on fares.
The effect of Booking Day on posted fares is 0.34%-0.42% higher for LCCs
than FSCs. LCCs engage in a stronger IPD, in line with the more
aggressive pricing behavior of LCCs.
VI. SUMMARY AND CONCLUSIONS
This paper explores pricing in air transportation for short-haul
markets, removing the influence of inter-modal competition. To that end,
we use a unique dataset on the southern Italian market that exhibits
limited inter-modal competition, thus airline pricing strategies are the
straight results of air-related competition.
Basically, we explore two issues. The first is to measure the
extent to which intra-modal competition determines fares. The second is
to shed light on the inter-temporal profile of fares to verify whether
airlines engage in IPD and whether IPD is of a monopolistic type or a
competitive type. Results are robust across regressions. Further, the
robust Hausman test shows that the RE GLS estimator provides consistent
estimates.
We find that airlines exploit their dominant position on a
city-pair market. When the intramodal competition reduces, airlines post
higher fares. Indeed, a 10% increase in Market Share leads to 6.4%
higher fares, and a 10% increase of HHI leads to 5.7% higher fares.
Further, we show that airlines do undertake IPD and that the fare
inter-temporal profile appears to be non-monotonic, resembling a J-curve
with a turning point included in the interval of the 43rd to 45th days
before departure. In addition to the existing interpretation on the
non-monotonicity of fares inter-temporal profile, we set forward two new
possible views. On the one hand, the non-monotonicity would be the
evidence that airlines exploit consumer-bounded rationality. Travelers
generally believe that the later the ticket is bought, the more it costs
and, thus price-sensitive consumers tend to buy in advance. Thus,
airlines can extract a greater surplus by setting moderately higher
fares for very-early purchasers who will buy the tickets believing to
pay the cheapest fares. On the other hand, a higher fare for very-early
purchasers can be interpreted as a fee for risk aversion. Airlines can
extract additional surplus from risk-averse travelers by posting
moderately higher fares at the very beginning of the selling schedule.
The empirical evidence is in favor of competitive-type price
discrimination: a more competitive market structure fosters the
implementation of IPD. Basically, in less competitive city-pair markets,
the J-curve appears to be flattened. Finally, airline-pricing strategies
differ depending on carrier type. LCCs seem to adopt a more aggressive
pricing behavior as, on average, they set lower fares and undertake
stronger IPD strategies.
One might say that price discrimination is only beneficial for
airlines. However, in more competitive markets, airlines charge lower
fares that, together with the IPD, allow them to target larger segments
of demand, which leads to a "democratization" of air travel.
This is very important for areas such as southern Italy where the
intermodal competition is limited.
Developments for future research could be an enlargement of the
territorial coverage in order to compare different exogenously
determined accessibility conditions and, thus, to measure the impact of
air-related competition on accessibility. Moreover, it would be
interesting to analyze the impact of local government subsidies often
granted to low-cost airlines through co-marketing programs on fares and
pricing strategies, thus analyzing the net welfare of the area in
question.
APPENDIX
We have distinguished between carriers of two types: FSCs and LCCs.
Indeed, we have assumed similar operating characteristics and pricing
behavior within types. For the robustness check we verify whether the
results hold when a more detailed distinction is made and carrier
dummies are added to the model. As shown in Table Al, estimates do not
change when we make more specific hypotheses about the behavior of each
carrier.
TABLE Al
RE GLS Estimates with Carrier-Specific Dummies
Baseline Model
Market Share HHI
(1) (2) (3) (4)
Market Structure 0.0068 *** 0.0063 *** 0.0051 *** 0.0051 ***
(0.0012) (0.0011) (0.0009) (0.0009)
Booking Day -0.0141 *** -0.0353 *** -0.0141 *** -0.0353 ***
(0.0005) (0.0013) (0.0005) (0.0013)
Booking Day (2) 0.0004 *** 0.0004 ***
(0.0000) (0.0000)
Market Structure
x Booking Day
Holidays 0.2253 *** 0.2359 *** 0.2307 *** 0.2339 ***
(0.0435) (0.0442) (0.0448) (0.0449)
Hausman Test
Statistic 0.011 1.821 0.065 2.541
Hausman Test p
value 0.916 0.402 0.798 0.281
Observations 19,605 19,605 19,605 19,605
Extended Model I
Market Share HHI
(5) (6) (7) (8)
Market Structure 0.0047 *** 0.0049 *** 0.0036 *** 0.0041 ***
(0.0012) (0.0012) (0.0010) (0.0010)
Booking Day -0.0166 *** -0.0375 *** -0.0171 *** -0.0374 ***
(0.0008) (0.0015) (0.0013) (0.0016)
Booking Day (2) 0.0004 *** 0.0004 ***
(0.0000) (0.0000)
Market Structure 0.0001 *** 0.0001 *** 0.0001 ** 0.0000 **
x Booking Day (0.0000) (0.0000) (0.0000) (0.0000)
Holidays 0.2333 *** 0.2363 *** 0.2318 *** 0.2346 ***
(0.0441) (0.0442) (0.0448) (0.0448)
Hausman Test
Statistic 0.088 2.081 0.119 2.666
Hausman Test p
value 0.957 0.556 0.942 0.446
Observations 19,605 19,605 19,605 19,605
Note: Standard errors (in parentheses) are clustered at flight
level. Control dummies are always included but not reported. * p <
0.1; ** p < 0.05; *** p < 0.01.
ABBREVIATIONS
FSC: Full-Service Carrier
GENP: Geometric Mean of Enplanements
GLS: Generalized Least Square
GMM: Generalized Method of Moments
HH1: Herfindahl-Hirschman Index
IPD: Inter-Temporal Price Discrimination
LCC: Low-Cost Carrier
QHHI: Quasi Herfindahl-Hirschman Index
RE: Random Effects
RPM: Revenue Passenger Miles
doi: 10.1111/ecin.12104
REFERENCES
Abrate, G., G. Fraquelli, and G. Viglia. "Dynamic Pricing
Strategies: Evidence from European Hotels." International Journal
of Hospitality Management, 31 (1). 2012, 160-68.
Alderighi, M, and C. A. Piga. "On-line Booking and Revenue
Management: Evidence from a Low-Cost Airline." Review of Economic
Analysis, 2(3), 2010, 272-86.
Alderighi, M.. A. Cento, P. Nijkamp, and P. Rietveld. "The
Entry of Low-Cost Airlines: Price Competition in the European Airline
Market." Tinbergen Institute Discussion Paper, 04-074/3, 2004.
Alves, C. F., and C. Barbot. "Price Discrimination Strategies
of Low-Cost Carriers." Journal of Transport Economics and Policy,
43(3), 2009, 345-63.
Bachis, E., and C. A. Piga. "Hub Premium, Airport Dominance
and Market Power in the European Airline Industry." Discussion
Paper Series 2007_11, Department of Economics, Loughborough University,
2007a.
--. "Pricing Strategies by European Traditional and Low Cost
Airlines. Or, When Is It the Best Time to Book On Line?," in
Advances in Airline Economics, Volume 2: The Economics of Airline
Institutions, Operations and Marketing, edited by D. Lee. Amsterdam. The
Netherlands: Elsevier, chapter 10, 2007b, 319-44.
Belobaba, P. P. "Air Travel Demand and Airline Seat Inventory
Management." Ph.D. dissertation, Massachusetts Institute of
Technology, 1987.
Bergantino, A. S. "Le Strategic di Prezzo Delle Compagnie
Tradizionali e Delle Low Cost. Implicazioni per i Sistemi Aeroportuali
Minori: il Caso Della Puglia," in Trasporti, Ambiente e Territorio.
La Ricerca di un Nuovo Equilibrio. Milano: Franco Angeli, 2009,77-91.
Bergantino, A. S., and C. Capozza. "Airline Pricing Behaviour
under Limited Intermodal Competition." SERIES Working Paper No. 44,
2012 (other version: SIET Working Paper, 2012).
Bilotkach, V, A. A. Gaggero, and C. A. Piga. "Airline Pricing
under Different Market Conditions: Evidence from European Low-Cost
Carriers." Discussion Paper Series 2012_01, Department of
Economics, Loughborough University, 2012.
Borenstein, S. "Price Discrimination in Free-Entry
Markets." RAND Journal of Economics, 16(3), 1985, 380-97.
--. "Hubs and High Fares: Dominance and Market Power in the
U.S. Airline Industry." RAND Journal of Economics, 20(3), 1989,
344-65.
Borenstein, S., and N. L. Rose. "Competition and Price
Dispersion in the US Airline Industry." Journal of Political
Economy, 102(4), 1994, 653-83.
Brueckner, J. K., D. Lee, and E. S. Singer. "Airline
Competition and Domestic US Airfares: A Comprehensive Reappraisal."
Economics of Transportation, 2(1), 2013, 1-17.
Carbonneau, S., R. P. McAfee, H. Mialon, and S. Mialon. "Price
Discrimination and Market Power." Emory Economics 0413, Department
of Economics, Emory University, Atlanta, 2004.
Carlsson, F. "Prices and Departures in European Domestic
Aviation Markets." Review of Industrial Organization, 24(1), 2004,
37-49.
Dana, J. D. "Advance-Purchase Discounts and Price
Discrimination in Competitive Markets." Journal of Political
Economy, 106(2), 1998, 395-422.
Evans, W. N., and J. N. Kessides. "Localised Market Power in
the U.S. Airline Industry." Review of Economics and Statistics,
75(1), 1993, 66-75.
Fageda, X., J. L. Jimenez, and J. Perdiguero. "Price Rivalry
in Airline Markets: A Study of a Successful Strategy of a Network
Carrier against a Low-Cost Carrier." Journal of Transport
Geography, 19(4), 2011, 658-69.
Gaggero, A. A. Airline Pricing and Competition: The J-Curve of
Airline Fares. Saarbrucken, Germany: Lambert Academic Publishing, 2010.
Gaggero, A. A., and C. A. Piga. "Airline Competition in the
British Isles." Transportation Research Part E, 46(2), 2010,
270-79.
--. "Airline Market Power and Inter-temporal Price
Dispersion." Journal of Industrial Economics, 59(4), 2011, 552-77.
Gale, I. L. "Price Dispersion in a Market with
Advance-Purchases." Review of Industrial Organization, 8(4), 1993,
451-64.
Gale, I. L., and T. J. Holmes. "The Efficiency of
Advance-Purchase Discounts in the Presence of Aggregate Demand
Uncertainty." International Journal of Industrial Organization,
10(3), 1992, 413-37.
--. "Advance-Purchase Discounts and Monopoly Allocation of
Capacity." American Economic Review, 83(1), 1993, 135-46.
Gerardi, K., and A. Shapiro. "Does Competition Reduce Price
Dispersion? New Evidence from the Airline Industry." Journal of
Political Economy, 117(1), 2009, 1-37.
Giaume, S., and S. Guillou. "Price Discrimination and
Concentration in European Airline Markets." Journal of Air
Transport Management, 10(5), 2004, 305-10.
Graham, D. R., D. P. Kaplan, and D. S. Sibley. "Efficiency and
Competition in the Airline Industry." Bell Journal of Economics,
14(1), 1983, 118-38.
Hayes, K., and L. Ross. "Is Airline Price Dispersion the
Result of Careful Planning or Competitive Forces?" Review of
Industrial Organization, 13(5), 1998, 523-42.
Holmes, T. "The Effects of Third-Degree Price Discrimination
in Oligopoly." American Economic Review, 79(1), 1989, 244-50.
Lofgren, K. G. "The Theory of Inter-Temporal Price
Discrimination. An Outline." Swedish Journal of Economics, 73(3),
1971,333-43.
Mantin, B., and B. Koo. "Dynamic Price Dispersion in Airline
Market." Transportation Research Part E, 45(6), 2009, 1020-29.
McAfee. R. P, and V. te Velde "Dynamic Pricing in the Airline
Industry, " in Economics and Information Systems, Vol. I, edited by
T. J. Hendershott. Amsterdam, The Netherlands: Elsevier, 2007.
Moller, M., and M. Watanabe. "Advance Purchase Discounts
versus Clearance Sales." Economic Journal, 120(549), 2010, 1125-48.
Stavins, J. "Price Discrimination in the Airline Market: The
Effect of Market Concentration." Review of Economics and
Statistics, 83(1), 2001, 200-202.
Stock, J. H., and M. Yogo "Testing for Weak Instruments in IV
Regression," in Identification and Inference for Econometric
Models: A Festschrift in Honor of Thomas Rothenberg, edited by W. K. A.
Donald and J. H. Stock. Cambridge: Cambridge University Press, 2005,
80-108.
Stokey, N. "Inter-Temporal Price Discrimination."
Quarterly Journal of Economics, 93(3), 1979, 355-71.
Tirole, J. The Theory of Industrial Organization. Cambridge, MA:
MIT Press, 1988.
Wooldridge. J. M. Econometric Analysis of Cross Section and Panel
Data. Cambridge, MA: MIT Press, 2002.
(1.) A preliminary version of this paper is Bergantino and Capozza
(2012).
(2.) See page 355.
(3.) Abrate et al. (2012) show that in the hotel industry,
hoteliers undertake IPD with two opposite trends. If a room is booked
for the working days, last minute prices are lower. Instead, if a room
is reserved for the weekend, last minute prices are higher.
(4.) See Tirole (1988), chapter 3.
(5.) Gerardi and Shapiro (2009) explain that the panel approach
allows them to estimate the effect of competition by accounting for
changes in the competitive structure of a given route over time rather
than changes in competitive structures across routes.
(6.) Ticket restrictions are the Saturday-night stay-over
requirement and the advance-purchase requirement.
(7.) Besides the ticket restrictions used by Stavins (2001), Giaume
and Guillou (2004) take into account some exogenous segmentation such as
families, age groups, student status, and events.
(8.) Alderighi et al. (2004) find that when an LCC enters a given
route, the FSC incumbent reacts by lowering both leisure and business
fares. Further, Fageda et al. (2011) note that traditional carriers are
progressively adopting the management practices of LCCs. In particular,
FSCs, through their low-cost subsidiaries, are able to price more
aggressively, and hence successfully compete with LCCs.
(9.) The idea of measuring the net effect of price discrimination
from varying the market structure has been inspired by the approach of
Stavins (2001).
(10.) We do not compute market structure variables at route level
because, working with a peripheral area, almost all the carriers could
operate as a monopolist on a given route. We need the city-pair level to
capture the real competition between carriers.
(11.) Market structure would be more appropriately calculated using
revenue passenger miles (RPM), while we base our calculations on the
amount of daily flights. This choice is due, basically, to data
constraint. At the moment, in fact, data on RPM are not available at
flight level for European connections and, thus, the best proxy of
market structure is based on data on the number of flight provided,
which are publicly available. For the same reason, a measure based on
daily flights is used also by Gaggero and Piga (2010, 2011).
(12.) Based on Gaggero and Piga (2011).
(13.) See Wooldridge (2002), 290-91.
(14.) Stavins follows the approach of Graham et al. (1983).
(15.) See, for instance, Borenstein and Rose (1994), Gerardi and
Shapiro (2009), Gaggero and Piga (2010). For a more detailed description
of the instruments see Borenstein (1989), 351-53.
(16.) We avoid any potential distortion on pricing strategies
caused by online travel agencies that could set discounted fares.
(17.) See, for instance, the analysis on airline travel demand
carried out by Belobaba (1987).
(18.) Current data on number of enplanements do not cover the whole
sample of round-trip fares, so GMM estimations are carried out on a
smaller sample.
(19.) We find support for the use of the RE GLS estimator for two
main reasons. The RE GLS estimator is inconsistent if regressors are
correlated with individual-specific effect, in our case the
flight-specific effect. This is the omitted-variables problem one could
try to solve by adding further regressors which might be enough to make
the fixed effect unnecessary. Actually, we include in the regressions a
rich set of control dummies that, given the Hausman test's results,
are able to account for much of the variance in the data. Then, the RE
GLS estimator corresponds to the FE estimator, and t goes to infinity.
In our data sample, we observe each round-trip fare starting from 60
days before departure, thus t = 60 might be fairly consistent as t is
equal to infinity.
(20.) Critical values were computed by Stock and Yogo (2005) for
the Cragg-Donald statistic which assumes i.i.d. errors. Results need to
be interpreted with caution only if the Kleibergen-Paap rk statistic is
close to the critical values.
(21.) This is in line with Bergantino (2009). She highlights that
LCCs post half the fares of FSCs on some Italian connection at small
airports.
ANGELA STEFANIA BERGANTINO and CLAUDIA CAPOZZA *
* The authors would like to thank the anonymous referees for their
helpful and constructive comments that contributed to improving the
final version of the paper. They would also like to thank the editor
Wesley W. Wilson for generous comments and support during the review
process. Special thanks go to Marco Alderighi. Richard Amott. Michele
Bemasconi, Volodymyr Bilotkach, Andre de Palma, Alberto Gaggero, Andrew
Goetz, Kai Hiischelrath, Marc Ivaldi, and Claudio Piga for useful
insights on earlier versions. Finally, the authors also wish to
acknowledge the suggestions made by the participants at 17th ATRS
Conference, 5th ICEEE, 5th CRNI Conference, 39th EARIE Conference, 14th
SIET Conference, RCEA Workshop on "The economics and management of
leisure, travel and tourism" and 32nd AISRe Conference. All
remaining errors are the authors'.
Bergantino: Department of Economics, Management and Business Law,
University of Bari Aldo Moro, Bari 70124, Italy. Phone 0039 080 5049038,
Fax 0039 080 50499149/042, E-mail angelastefania.bergantino@uniba.it
Capozza: Department of Economics, Management and Business Law,
University of Bari Aldo Moro, Bari 70124, Italy. E-mail
claudia.capozza@uniba.it
TABLE 1
List of City-Pair Markets
Origin Destination
Bari Milan
Bari Rome
Brindisi Milan
Brindisi Rome
Catania Milan
Catania Rome
Naples Milan
Naples Rome
Palermo Milan
Palermo Rome
TABLE 2
List of Airline Companies
Full-Service Carriers Low-Cost Carriers
AirOne Alpieagles Meridiana WindJet
Alitalia Blu Express MyAir Ryanair
Lufthansa Easyjet Volare Web
TABLE 3
Descriptive Statistics
Standard
Variables Observations Mean Deviation Minimum Maximum
Fares 19,605 153.80 84.85 11.92 690.49
Market Share 19,605 0.405 0.286 0.065 1
HHI 19,605 0.497 0.203 0.225 1
Booking Day 19,605 24.672 14.889 1 60
Holiday 19,605 0.458 0.498 0 1
LCC 19,605 0.455 0.498 0 1
TABLE 4
Average Round-Trip Fares per Mile Posted by
the Incumbent Airline
Average
Fare per
Connection Mile
BRI-FCO-BRI 0.4260
BRI-LIN-BRI 0.1832
BRI-MXP-BRI 0.2387
BDS-FCO-BDS 0.3086
BDS-LIN-BDS 0.1588
BDS-MXP-BDS 0.1332
PMO-FCO-PMO 0.2548
PMO-LIN-PMO 0.1587
PMO-MXP-PMO 0.1225
CTA-FCO-CTA 0.2594
CTA-MXP-CTA 0.1421
NAP-FCO-NAP 0.8788
NAP-L1N-NAP 0.1976
Note: Data on the distance between the two route end-
points are taken from the Word Airport Codes web site.
BRI, Bari; BDS, Brindisi; CTA, Catania; FCO, Rome
Fiumicino; LIN, Milan Linate; MXP, Milan Malpensa: NAP,
Naples; PMO, Palermo.
TABLE 5
Baseline Model
RE GLS Estimates
Market Share
(1) (2)
Market Structure 0.0064 *** 0.0064 ***
(0.0009) (0.0009)
Booking Day -0.0141 *** -0.0353 ***
(0.0005) (0.0013)
Booking Day (2) 0.0004 ***
(0.0000)
Holidays 0.2082 *** 0.2112 ***
(0.0521) (0.0522)
LCC -0.2249 *** -0.2259 ***
(0.0426) (0.0426)
Hausman test stat. 0.843 2.141
Hausman test p-value 0.359 0.343
Kleibergen-Paap stat.
Hansen J test stat.
Hansen J test p-value
Endogeneity test statistic
Endogeneity test p-value
Observations 19,605 19,605
RE GLS Estimates
HHI
(3) (4)
Market Structure 0.0057 *** 0.0057 ***
(0.0010) (0.0010)
Booking Day -0.0141 *** -0.0353 ***
(0.0005) (0.0013)
Booking Day (2) 0.0004 ***
(0.0000)
Holidays 0.2310 *** 0.2341 ***
(0.0554) (0.0554)
LCC -0.4047 *** -0.4058 ***
(0.0324) (0.0325)
Hausman test stat. 0.085 1.645
Hausman test p-value 0.771 0.439
Kleibergen-Paap stat.
Hansen J test stat.
Hansen J test p-value
Endogeneity test statistic
Endogeneity test p-value
Observations 19,605 19,605
GMM Estimates
Market Share
(5) (6)
Market Structure 0.0068 *** 0.0069 ***
(0.0013) (0.0013)
Booking Day -0.0136 *** -0.0331 ***
(0.0005) (0.0014)
Booking Day (2) 0.0004 ***
(0.0000)
Holidays 0.1836 *** 0.1883 ***
(0.0597) (0.0599)
LCC -0.2481 *** -0.2460 ***
(0.0555) (0.0556)
Hausman test stat.
Hausman test p-value
Kleibergen-Paap stat. 114.9 114.9
Hansen J test stat. 0.064 0.054
Hansen J test p-value 0.800 0.817
Endogeneity test statistic 0.058 0.031
Endogeneity test p-value 0.809 0.860
Observations 16.476 16.476
GMM Estimates
HHI
(7) (8)
Market Structure 0.0079 *** 0.0080 ***
(0.0013) (0.0013)
Booking Day -0.0135 *** -0.0331 ***
(0.0005) (0.0014)
Booking Day (2) 0.0004 ***
(0.0000)
Holidays 0.1990 *** 0.2041 ***
(0.0623) (0.0624)
LCC -0.4281 *** -0.4286 ***
(0.0374) (0.0374)
Hausman test stat.
Hausman test p-value
Kleibergen-Paap stat. 355.2 355.4
Hansen J test stat. 0.048 0.039
Hansen J test p-value 0.827 0.844
Endogeneity test statistic 2.780 2.741
Endogeneity test p-value 0.096 0.098
Observations 16,476 16,476
Notes: Standard errors (in parentheses) are clustered at flight
level. Control dummies are always included but not reported. Stock
and Yogo (2005) critical value is 19.93. * p < 0.1; ** p < 0.05; ***
p < 0.01.
TABLE 6
The Marginal Effect ([beta]) of Booking Day (BD)
on Fares
BD [beta] BD [beta] BD [beta] BD [beta]
5 -0.0313 *** 35 -0.0070 *** 45 0.0011 51 0.0059 ***
(0.0011) (0.0005) (0.0008) (0.0010)
10 -0.0273 *** 40 -0.0030 *** 46 0.0019 ** 52 0.0067 ***
(0.0009) (0.0006) (0.0008) (0.0010)
15 -0.0233 *** 41 -0.0022 *** 47 0.0027 *** 53 0.0075 ***
(0.0008) (0.0006) (0.0008) (0.0011)
20 -0.0192 *** 42 -0.0014 ** 48 0.0035 *** 54 0.0083 ***
(0.0006) (0.0007) (0.0009) (0.0011)
25 -0.0151 *** 43 -0.0006 49 0.0043 *** 55 0.0091 ***
(0.0005) (0.0007) (0.0009) (0.0011)
30 -0.0111 *** 44 -0.0002 50 0.0051 *** 60 0.0132 ***
(0.0004) (0.0007) (0.0009) (0.0011)
TABLE 7
Extended Model I
RE GLS Estimates
Market Share HHI
(1) (2) (3) (4)
Market Structure 0.0049 *** 0.0051 *** 0.0043 *** 0.0047 ***
(0.0010) (0.0010) (0.0011) (0.0011)
Booking Day -0.0166 *** -0.0375 *** -0.0171 *** -0.0374 ***
(0.0008) (0.0015) (0.0013) (0.0016)
Booking Day (2) 0.0004 *** 0.0004 ***
(0.0000) (0.0000)
Market Structure 0.0001 *** 0.0001 *** 0.0001 ** 0.0000 **
x Booking Day (0.0000) (0.0000) (0.0000) (0.0000)
Holidays 0.2088 *** 0.2118 *** 0.2321 *** 0.2348 ***
(0.0521) (0.0522) (0.0554) (0.0554)
LCC -0.2263 *** -0.2271 *** -0.4049 *** -0.4060 ***
(0.0424) (0.0424) (0.0324) (0.0325)
Hausman test
statistic 0.942 2.325 0.109 1.709
Hausman test
p-value 0.624 0.508 0.947 0.635
Kleibergen-Paap
statistic
Hansen J test
statistic
Hansen J test
p-value
Endogeneity
test
statistic
Endogeneity
test
p-value
Observations 19,605 19,605 19,605 19.605
GMM Estimates
Market Share HHI
(5) (6) (7) (8)
Market Structure 0.0055 *** 0.0057 *** 0.0067 *** 0.0068 ***
(0.0014) (0.0013) (0.0013) (0.0013)
Booking Day -0.0159 *** -0.0350 *** -0.0161 *** -0.0354 ***
(0.0011) (0.0016) (0.0014) (0.0018)
Booking Day (2) 0.0004 *** 0.0004 ***
(0.0000) (0.0000)
Market Structure 0.0001 ** 0.0000 ** 0.0000 * 0.0000 *
x Booking Day (0.0000) (0.0000) (0.0000) (0.0000)
Holidays 0.1842 *** 0.1888 *** 0.1995 *** 0.2045 ***
(0.0597) (0.0598) (0.0624) (0.0624)
LCC -0.2472 *** -0.2452 *** -0.4278 *** -0.4283 ***
(0.0554) (0.0554) (0.0373) (0.0374)
Hausman test
statistic
Hausman test
p-value
Kleibergen-Paap
statistic 76.80 76.82 233.8 233.9
Hansen J test
statistic 0.062 0.053 0.043 0.035
Hansen J test
p-value 0.803 0.819 0.835 0.852
Endogeneity
test
statistic 0.658 1.064 3.644 2.810
Endogeneity
test
p-value 0.720 0.587 0.162 0.245
Observations 16,476 16.476 16,476 16,476
Notes: Standard errors (in parentheses) are clustered at flight
level. Control dummies are always included but not reported. Stock
and Yogo (2005) critical value is 14.43. *** p<0.01 ; ** p< 0.05; *
p<0.1.
TABLE 8
The Marginal Effect ([beta]) of Booking Day (BD)
on Fares by a 1 % Increase of Market Share/HHI
[beta] [beta]
BD (No Interaction) (Market Share) [beta] (HHI)
5 -0.0313 *** -0.0311 *** -0.0311 ***
(0.0011) (0.0011) (0.0011)
10 -0.0273 *** -0.0271 *** -0.0271 ***
(0.0009) (0.0009) (0.0009)
15 -0.0233 *** -0.0231 *** -0.0231 ***
(0.0008) (0.0008) (0.0008)
20 -0.0192 *** -0.0191 *** -0.0191 ***
(0.0006) (0.0006) (0.0006)
25 -0.0151 *** -0.0151 *** -0.0151 ***
(0.0005) (0.0005) (0.0005)
30 -0.0111 *** -0.0111 *** -0.0111 ***
(0.0004) (0.0004) (0.0004)
35 -0.0070 *** -0.0070 *** -0.0071 ***
(0.0005) (0.0005) (0.0005)
40 -0.0030 *** -0.0031 *** -0.0031 ***
(0.0006) (0.0006) (0.0006)
45 0.0011 0.0009 0.0009
(0.0008) (0.0006) (0.0007)
50 0.0051 *** 0.0050 *** 0.0049 ***
(0.0009) (0.0009) (0.0009)
55 0.0091 *** 0.0090 *** 0.0089 ***
(0.0011) (0.0011) (0.0011)
60 0.0132 *** 0.0130 *** 0.0129 ***
(0.0011) (0.0013) (0.0013)
TABLE 9
Extended Model II
RE GLS Estimates
Market Share HHI
(1) (2) (3) (4)
Market Structure 0.0064 *** 0.0064 *** 0.0057 *** 0.0057 ***
(0.0009) (0.0009) (0.0010) (0.0010)
Booking Day -0.0154 *** -0.0355 *** -0.0154 *** -0.0355 ***
(0.0009) (0.0015) (0.0009) (0.0015)
Booking Day (2) 0.0004 *** 0.0004 ***
(0.0000) (0.0000)
Holidays 0.0544 0.0763 0.0773 0.0992 *
(0.0572) (0.0564) (0.0602) (0.0594)
Holidays x 0.0064 *** 0.0056 *** 0.0064 *** 0.0056 ***
Booking Day (0.0009) (0.0008) (0.0009) (0.0008)
LCC -0.1279 *** -0.1462 *** -0.3068 *** -0.3255 ***
(0.0476) (0.0465) (0.0378) (0.0364)
LCC x Booking Day -0.0042 *** -0.0034 *** -0.0042 *** -0.0034 ***
(0.0009) (0.0008) (0.0009) (0.0008)
Hausman test
statistic 9.329 10.809 10.505 12.133
Hausman test
p-value 0.025 0.029 0.015 0.016
Kleibergen-Paap
statistics
Hansen J test
statistic
Hansen J test
p-value
Endogeneity test
statistics
Endogeneity test
p-value
Observations 19,605 19,605 19,605 19,605
GMM Estimates
Market Share HHI
(5) (6) (7) (8)
Market Structure 0.0068 *** 0.0069 *** 0.0079 *** 0.0080 ***
(0.0013) (0.0013) (0.0013) (0.0013)
Booking Day -0.0137 *** -0.0323 *** -0.0133 *** -0.0320 ***
(0.0010) (0.0015) (0.0010) (0.0015)
Booking Day (2) 0.0003 *** 0.0003 ***
(0.0000) (0.0000)
Holidays 0.0683 0.0848 0.0880 0.1049
(0.0639) (0.0633) (0.0666) (0.0659)
Holidays x 0.0046 *** 0.0041 *** 0.0044 *** 0.0039 ***
Booking Day (0.0009) (0.0009) (0.0010) (0.0009)
LCC -0.1147 ** -0.1276 ** -0.2855 *** -0.3008 **
(0.0579) (0.0564) (0.0407) (0.0392)
LCC x Booking Day -0.0054 *** -0.0048 *** -0.0057 *** -0.0051 **
(0.0009) (0.0009) (0.0009) (0.0009)
Hausman test
statistic
Hausman test
p-value
Kleibergen-Paap
statistics 115.2 115.2 356.4 356.6
Hansen J test
statistic 0.088 0.074 0.070 0.057
Hansen J test
p-value 0.767 0.786 0.791 0.812
Endogeneity test
statistics 0.032 0.016 3.043 2.967
Endogeneity test
p-value 0.857 0.900 0.081 0.085
Observations 16,476 16,476 16,476 16.476
Notes: Standard errors (in parentheses) are clustered at flight
level. Control dummies are always included but not reported. Stock
and Yogo (2005) critical value is 19.93. * p < 0.1; ** p < 0.05; *** p <
0.01.
