Impact of product differentiation on airline fares
An empirical study on European duopoly routes
Matthieu Fragni`
ere
Master’s Thesis supervised by Dr. Sander Onderstal
Amsterdam School of Economics - MSc ECO - 11390638
Statement of Originality
This document is written by Matthieu Fragni`ere who declares to take full respon-sibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those men-tioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Acknowledgements
I am grateful to many people for their assistance and support throughout this thesis. In particular, I would firstly like to thank my supervisor, Dr. Sander Onderstal, for his kind guidance and the interesting meetings that we had. Secondly, I thank Dr. Maurice Bun for his help on econometric issues. In addition, I would like to thank my family and Natacha Vallette d’Osia for their loving support and encouragements.
Contents
1 Introduction 4
2 Theory 5
2.1 Quantity-quality trade off under vertical differentiation . . . 7 2.2 Hotelling model and location . . . 10
3 Empirical study 13
3.1 Econometric model . . . 13 3.2 Data . . . 16
4 Results and interpretations 18
4.1 Results . . . 18 4.2 Interpretations . . . 21
5 Conclusion 24
1
Introduction
According to Industrial Organizations (IO) models, two firms competing in a duopoly market have the opportunity to increase their profits by differentiating from each other. Differentiation is an act of creating a set of meaningful differences that makes a company’s offers distinctive from those of competitors (Kotler & Keller, 2012). This could be done, for example, by setting two different levels of quality (verti-cal product differentiation). One firm could produce a high quality good, whereas the other could produce at low quality level, as emphasized by Shaked and Sutton (1982). Firms can also differentiate horizontally (spatial differentiation). In the case of airlines, this can be done by setting two different departure times, following the model developed by Hotelling (1929). The main objective of differentiation is to sep-arate the demand in two (in the case of duopoly) segments taking into account that consumers have heterogeneous preferences. It increases market power and allows firms to raise the price on their own segment.
In this study, I empirically analyze how these product differentiation possibili-ties influence prices offered by airlines. Although there are many theoretical papers on product differentiation, with different models and extensions, empirical studies are less common. In this context, the choice of the air transport market is very interesting since it is divided between traditional and low-cost airlines, providing differentiated products (regarding quality, for example). However, in contrast to the theory, there are several examples where traditional airlines reduce the quality of their product in order to reduce costs and compete in prices with low-cost airlines. This is notably the case in Geneva (Switzerland), where Swiss International
Air-lines significantly reduced the quality of its on-board services, getting closer to its
main competitor easyJet, a well-known British low-cost airline. Apart from quality differentiation, airlines could also use schedule (‘spatial’) differentiation to increase prices, by flying at very different time from their competitors. Applying existing models on the airline industry is therefore interesting.
The main objective of this research is to test whether quality and schedule dif-ferentiation (if used) increases flight ticket prices on duopoly routes in the
Euro-pean geographic area, where competition between traditional and low-cost airlines is tough. A positive correlation between the degree of product differentiation and prices suggests that differentiation increases prices and is a way for airlines to in-crease their profits. Moreover, it could also mean that it harms consumers, which may call for government intervention.
As mentioned above, there are few empirical studies on the relation between product differentiation and prices. Clemons, Hann, and Hitt (2002) found that product differentiation accounts for “at least 10% of the actual variation in ticket prices” in the American online travel agencies (OTAs) market (Clemons et al., 2002, p. 544). According to Slade (2004) and Nevo (2001), who respectively studied the U.K. beer market and the American cereals market, product differentiation has a positive impact on prices. For instance, Slade found that approximately 75% of the cost-margin is explained by the ‘differentiation effect’ (Slade, 2004, p. 157). In the driving schools market, Asplund and Sandin (1999) found that prices increase with the distance between markets, which is in line with the above mentioned theory. In the American retail food market, Fik (1988) also found that prices are higher when the distance between competitors is large. On the other hand, according to Cotterill (1986), the distance between competitors has no significant impact on prices, contradicting predictions of the theory. It is therefore interesting to test the validity of product differentiation models on the air transport market.
The paper is structured as follows. In section 2, I review the theory related to vertical and spatial differentiation. In particular, two relevant IO models are described. In section 3, an empirical research is performed using an econometric model in order to test the models. Section 4 provides the results and interpretations. Finally, a conclusion is given in section 5.
2
Theory
In this section, I briefly present the main characteristics of the airline industry. Two vertical and horizontal product differentiation models are precisely defined.
market in 1978 in the United-States and later in Europe (Dresner, Lin, & Windle, 1996). In their study, Dresner et al. (1996) showed that the entry of a low-cost airline on a route significantly reduces the price of tickets. This decrease is also observed on competing routes. There is little information, however, on what tools airlines have been using to keep prices high enough to survive. Differentiation could be one of them. Two IO models are particularly interesting to observe in order to answer this question.
As mentioned above, quality differentiation between airlines fits the ’Quality-quantity trade off under vertical differentiation’ model developed McCannon (2008). Following the theory, when consumers have different tastes for quality, profits of the two airlines are positively correlated with the difference between qualities offered. In fact, one of them should produce the good at the lowest quality level possible, whereas its competitor should produce at the highest one.
Apart from quality differentiation, the spatial differentiation between airlines can be modeled by the well-known ’Hotelling model’ developed by Hotelling (1929). In fact, they are two forces that influence airlines’ choices. On the one hand, they can maximize differentiation by setting very different time-scheduled flights and reduce price competition. Thereby, they would attract different types of consumers (those who do not want to travel at the time offered by the competitor) and be able to raise their prices, without losing consumers to their competitor. This is why, for example, on a particular duopoly route, if an airline decides to fly in the morning, its competitor should fly at the end of the day. On the other hand, airlines can minimize schedule differentiation in order to ’steal’ consumers from competitors (Borenstein & Netz, 1999). However, this would increase price competition and reduce margins. The main message of those two models is that firms have the possibility to raise prices and increase their profits by differentiating. In fact, higher product differentiation reduces cross-price elasticities, which means that consumers are less likely to switch from one good to another because of a change in the price. Therefore, prices of duopoly routes are expected to be higher when airlines differentiate.
In the following sections, two product differentiation models (from the IO the-ory) are defined. In section 2.1, quality (vertical) differentiation is modeled by the
’Quantity-quality trade off under vertical differentiation’ model. In section 2.2, spa-tial differentiation is also modeled, following the ’Hotelling model’. The question of how these models can be applied to the airline industry is also be addressed.
2.1
Quantity-quality trade off under vertical differentiation
The model developed by McCannon (2008) starts with the assumption of two iden-tical (symmetric) firms A and B competing in a duopoly market with differentiated goods. The two firms choose the level of quality of their good and its price (two-stage game). Consumers are uniformly distributed with respect to their taste for quality θ on the interval [θ, ¯θ], with ¯θ > θ:
A consumer has two options: buy the good from firm A or firm B. When consuming from firm i, a consumer j of type θ receives a utility of:
υj(p, s, θ) = θsi− pi
where here si is the quality level of the good and pi is the price paid. If the consumer
decides not to buy, then he/she gets a utility of zero. In the first step of this two-stage game, the two firms simultaneously select the level of quality si, i = A, B from
the interval [s, ¯s], ¯s > s. To simplify, let’s relabel the firms H and L, with sH > sL.
The quality of the good produced by firm H is therefore higher than firm L. After observing sH and sL, each firm chooses its price pi. The cost function of the firms
are defined as follows:
C(qi, si) = cqi+ aqisi = qi(c + asi)
where a > 0 is the magnitude of the trade-off. A high value of a means that producing a high quality good is more costly. The profit function of each firm has the usual form:
Which can be written as:
πi = piqi− qi(c + asi) = qi[pi−(c + asi)]
The method used to solve this game is called backward induction. The first step consists of finding the consumer that is indifferent between buying the good from firm H or firm L. The utility that this consumer gets when buying from firm H is the same as he gets when buying from firm L.
ˆθsH − pH = ˆθsL− pL
Which gives the indifferent consumer:
ˆθ = pH − pL
sH − sL
Consumers located in the segment [ˆθ, ¯θ] prefer to buy the good from firm H and consumers located in the other part [θ, ˆθ] prefer to buy from firm L. Therefore, demands for firm H and firm L are given by the probability that consumers are located on their respective segment. Since consumers are uniformly distributed on the interval [θ, ¯θ], demands are:
qH = P r{θ > ˆθ} = ¯θ − ˆθ
qL= P r{θ < ˆθ} = ˆθ − θ
Substituting qH and qL in the profit function of each firm gives:
πH = (¯θ − pH − pL sH − sL )[pH −(c + asH)] πL= ( pH − pL sH − sL − θ)[pL−(c + asL)]
Assuming that sH and sL are chosen at the first stage, in the second stage the two
firms choose their prices pH and pLby maximizing their own profit function. Taking
the first order conditions gives the resulting prices:
p∗H = c + (sH − sL)2¯θ − θ
3 +
a(2sH + sL)
p∗L= c + (sH − sL)¯θ− 2θ
3 +
a(sH + 2sL)
3 (2)
Both prices increase in H’s quality sH. An increase in sH increases differentiation
and allows both firms to charge higher prices. Also, prices are decreasing in L’s since an increase in sL reduces differentiation. In addition, we can see that quality
is related with the price, through the parameter a. A high value of a (the cost of quality) results in higher prices. Substituting the equilibrium prices p∗
H and p∗L in
the profit functions gives the following equilibrium profits:
πH = (sH − sL) 2¯θ − θ − a 3 !2 πL= (sH − sL) ¯θ− 2θ + a 3 !2
Both profits increase with the difference of quality levels (sH− sL). In addition, the
profit of firm H is increasing in the quality of its product: ∂πH
∂sH > 0, which means
that its best choice is to set sH as high as possible: s∗H = ¯s. Furthermore, the profit
of firm L is decreasing in sL: ∂π∂sL
L <0. Therefore, its best choice is to set sL as low
as possible: s∗
L = s
Interpretation and application to airline industry
There are two main interpretations of the equilibrium resulting from the model. First, as seen in the the profit functions, both profits increase with the difference in qualities provided. In order to maximize profits, firms set their qualities level as such that differentiation is maximized: s∗
H = ¯s and s
∗
L= s. Second, in the equations (1)
and (2) we see that prices are higher when differentiation is important. Therefore, prices are supposed to be higher when firms choose to differentiate (sH 6= sL) (which
is what they should do to maximize their profit) than when they set the same level of quality (sH = sL).
The airline industry, and more globally the air transport market is a complicated market. Many different actors play a role, such as manufacturers (mostly Airbus and Boeing), airlines, airports and final consumers (passengers). However, the previous model can be applied and tested on this market, in particular on duopoly routes, where two airlines compete. In the European area, there are lots of oligopolies,
especially since the deregulation of the market and the growth of low-cost airlines. The product (‘flight’) can be differentiated in regards to quality. For example, low-cost and traditional airlines are likely to offer flights at different quality levels. Therefore, it is interesting to check whether quality differentiation is actually used by airlines and analyze its impact on prices.
Several quality factors provided by airlines can be represented by si, although the
analysis related to a one-dimension version of the model. Following the equilibrium resulting from the development of the model, in duopoly routes, prices are supposed to be higher when the quality of the flight differs by airlines. In fact, there are two main example of quality levels that airlines choose before setting prices: the size of seats and legrooms provided in their airplanes and the size of the hand luggage that passengers can bring on board. Of course, there are other factors, as the distance between the terminal of an airport and the final gate, for instance. However, that type of factors can vary a lot within a particular airline, since airports are all different and airlines are not always able to choose their facilities. In addition to these former points, there are different services provided by some airlines, such as on-board food or coffee, that one could consider as free at first sight. However, the truth is that these services are not free but their price is included in the ticket fare. There are passengers that decide not to consume that type of service, which means that they actually paid for a service they inadvertently choose not to get. Therefore, services that are sometimes ’free’ (traditional airlines) and sometimes charged on-board (low-cost airlines) cannot be included as quality factors such as the size of seats or hand luggage, which passengers do not have to pay for, whatever the airline.
2.2
Hotelling model and location
This section defines the impacts of differentiation are defined according to the Hotelling competition model (Hotelling, 1929), following the method developed by Belleflamme and Peitz (2015, p. 119). The model assumes that consumers are dis-tributed on a linear ’city’ (see Hotelling (1929, p. 45)) that could also be a linear ’day’ in the case of airline competition. This linear day starts at zero and the length of the line is 1. The total mass of consumers is also normalized to 1.
There are two firms (i = A, B) selling their product at price pi. When consuming,
a consumer pays the price of the good and the transportation costs: pi+ ty2, where
t >0 is a measure of differentiation, and y represents the cost of travelling (‘distance
that a consumer has to walk to consume the good’).
Profit functions have the usual form: Πi = (pi − c)qi and the model assumes a
two-stage game:
1. Firms simultaneously choose their location on the line (A,B) 2. Firms simultaneously choose their prices (pA and pB)
The game is solved using backward induction.
Let denote LAand LBthe distance between the extremes of the line and the locations
of the firms A and B. Starting from the left side of the line, location of firm A is
LA∈[0, 1] and location of firm B is 1 − LB ∈ [0, 1]. As in the previous model, the
first step consists of finding the consumer ˆx that is indifferent between firms:
pA+ t(ˆx − LA)2 = pB+ t(ˆx − (1 − LB))2
Which gives the indifferent consumer ˆx: ˆx = LA+1 − L
A− LB
2 +
pB− pA
2t(1 − LB− LA)
The demand for firms A and B are therefore:
DA(pA, pB) = qA= ˆx = LA+ 1 − L A− LB 2 + pB− pA 2t(1 − LA− LB) DB(pA, pB) = qB = 1 − ˆx = LB+ 1 − LA− LB 2 + pA− pB 2t(1 − LA− LB)
Substituting qA and qB into the profit functions gives:
ΠA= (pA− c)DA(pA, pB) = (pA− c) LA+ 1 − LA− LB 2 + pB− pA 2t(1 − LA− LB) ! ΠB = (pB− c)DB(pA, pB) = (pB− c) LB+1 − L A− LB 2 + pA− pB 2t(1 − LA− LB) !
Taking the first order conditions (∂πi
∂pi = 0) and substituting each other gives the
following equilibrium prices:
p∗A(LA, LB) = c + t(1 − LA− LB) 1 + LA− LB 3 (3)
p∗B(LA, LB) = c + t(1 − LA− LB) 1 + LB− LA 3 (4) Substituting equilibrium prices into the profit function of the two firms and in the case where LB = 0, then firm A’s profit is:
ΠA(LA,0) =
t
18(1 − LA)(3 + LA)2
This profit function is decreasing in LA, which means that firm A should set LA= 0.
In fact, the best location in response to the competitor’s is at the opposite end of the line. Therefore, firms will locate at the extremes of the line and differentiation is maximized.
Interpretation and application to the airline industry
Equations (3) and (4) show that prices increase with differentiation. In other words, prices are higher when the distance between the location of the firms is bigger. Indeed, if LA = LB = 0 (both firms locate at the extremes of the line) then prices
are pA = pB = c + t, where t measures the level of differentiation. On the other
hand, if LA + LB = 1 (same location, no differentiation, t = 0) then prices are
pA = pB = c (Bertrand prices). Therefore, firms can increase their profits through
higher prices, by differentiating.
In the air transport market, two airlines offering one flight each on a particular route on the same day should thus schedule their flight such that the difference (in minutes, or hours) between them is maximized. This would let them reach two different parts of the demand and thereby increase the price on their own part without losing consumers to their competitor. In other words, prices are supposed to be higher when the timespan between the flights of airline A and airline B is large. For example, on the route Budapest-Barcelona, where Ryanair and Wizzair are competing, prices are expected to be higher if one of them flies early in the morning and the other late in the evening than if the two of them fly at the same time. It can be assumed that consumers are located on the line [0; 1], which could be seen as a entire day from 00:00 to 23:59, according to the time they would like to take off. Each company then maximizes its profit by choosing the best location (departure time) on this segment.
3
Empirical study
In this section, an empirical study is conducted in order to see whether quality and spatial differentiation impact prices in the air transport market. A regression of the price on different variables is also accomplished, focusing on European-duopoly routes. The econometric model used in this study is defined in section 3.1 and the description of the data is provided in section 3.2.
3.1
Econometric model
The following econometric model is used in this section, where P rice is the log of the price offered by airline j on route i at time slot t:
P riceijt= β0+ β1Lengthi+ β2∆Seatsizei+ β3∆Lugsizei+ β4∆T imeouti
+ β5∆T imereti+ β6Soutij + β7Sretij + β8Booktimeij
+ α1OTt+ α2RTt+ ijt
(5)
The explanatory variables are defined as follows:
• Length: Log of the length of the flight (on route i, in minutes). The sign is supposed to be positive, since long flights are more costly for airlines, which should be reflected in higher prices.
• ∆Seatsize: Log of the difference in the size of the seats provided by the two airlines on route i (in square inch), which represents one of the quality factors. Following the quality differentiation model developed above, the price is supposed to be positively correlated with the difference in quality provided by the two competing airlines. The sign of this variable is therefore likely to be positive.
• ∆Lugsize: Difference in the size of hand luggage accepted by the two airlines on route i (in cm3). This variable is the second quality factor included in the model. For the same reasons as the other quality factor, the sign of this variable is supposed to be positive.
• ∆T imeout: Log of the difference in the departure time of the two competing outward flights on route i (in minutes). This variable is the representation of schedule differentiation. According to the Hotelling model previously analyzed, prices should be higher when this differentiation is large. Thereby, the sign is also supposed to be positive.
• ∆T imeret: Log of the difference in the departure time of the two competing return flights on route i (in minutes). For the same reasons, the sign of this variable should be positive.
• Sout: Log of the number of seats available in the outward flight (airline j on route i). Since the price is supposed to increase when the number of available seats decreases, the sign of this variable is expected to be negative.
• Sret: Log of the number of seats available in the return flight (airline j on route i). The expected sign of this variable is also negative, for the same reasons.
• Booktime: The time of the reservation (on route i with airlines j)
• OT , RT : Binary variables that take the value of 1 if the departure time of the outward flight (OT ) is, respectively, in periods t = 1, 2, 3, 4 or 5 of the day, and zero otherwise. Same for the return flight (RT ). Period 1 (t = 1) corresponds to night (00:00-5:59), t = 2 to morning (6:00-11:59), t = 3 to
afternoon (12:00-16:59), t = 4 to late afternoon (17:00-20:59), and t = 5 to evening (21:00-23:59).
Assumptions on the error term ijt are the following: Errors are normally
dis-tributed, homoscedastic, and there is no serial correlation. Therefore, an Ordinary Least Squares (OLS) estimation of the coefficients is consistent.
In order to interpret the coefficients as elasticities (percentages), the variables in the regression are expressed in logs. An exception is made for ∆Lugsize, because its value in the data set sometimes equals zero, a value that does not make sense in logs.
There is one usual issues that an OLS regression faces when estimating the price of a product: endogeneity, caused by simultaneity. Typically, the quantity sold of a product needs to be included in the model as an explanatory variable, even if depends on the dependant variable (price). This leads to biased OLS estimators, seriously depleting the quality of the estimation. Thereby, in this model one can expect simultaneity since the number of available seats is likely to depend on prices. However, it is reasonable to assume that the price setting process is done by taking into account the number of available seats. Indeed, the number of seats available is determined by previous prices rather than the actual one. Therefore, simultaneity (or endogeneity) can reasonably be left out of the concerns and the OLS estimators should remain unbiased.
The most interesting variables that are tested in this study are ∆T imeout, ∆T imeret, ∆Seatsize and ∆Lugsize. The two first represent spatial differenti-ation by measuring the time between the two outward/return flights. ∆Seatsize and ∆Lugsize represent vertical differentiation by measuring the difference between two quality factors. Positive and significant coefficient would mean that quality and spatial differentiation effectively increase prices of air fares.
Other control variables are included in the regression: The length of the flight (Length) should have a positive impact on the price since it increases operation costs through fuel consumption, for example. In addition, the number of seats in the outward and the return flight available at the moment of the reservation (Sout and Sret) should have an negative impact, since prices increase when the number of available seats decreases.
Flight ticket prices might also depend on operating costs such as the price of the aircraft used and its consumption. However, airlines very often use more than one type of aircraft (except a few low-cost airlines such as easyJet and Ryanair), which also enters in the pricing strategy. Since the price of a Boeing 737 (Family, except
MAX models (not used in the selected routes)) is on average 0.42% more expensive
than an Airbus A320 (Family) (Airbus, 2016; Boeing, 2017), prices are not likely to vary because of the type of aircraft.
included in the model in order to control for variations in the willingness to pay within a day. For example, morning or evening flight might be preferred to other, which could be reflected through higher prices. Five binary variables are therefore included to separate the day in five different parts: night, morning, afternoon, late afternoon and evening.
3.2
Data
The collection of the data was done using the online platform www.kayak.com, which shows the departure time, the length of the flight and the type of aircraft used, per airline. Price data were collected on airlines’ official websites, in order to get ‘pure’ prices and avoid discounted prices that could be displayed on ‘search and book’ online platforms. Another way to collect data would have been to use time series. In this case, variation of prices in time would have been observable. However, European airlines unfortunately do not share these data. Therefore, a ‘consumer’ method was used: start a reservation process for every flight and observe what price airlines offer.
Data were collected on 99 European-duopoly routes (included some destinations in Northern Africa and Turkey, see Table 2 in the appendix). The price of flight ticket is a well known puzzle, since it depends on many factors. One of the most important is the timing (number of days until departure, for example). According to Klee (2013), CEO of the online platform CheapAir.com, the cheapest time to buy flight tickets is 7 weeks before departure. Moreover, several studies showed that the price paid by passengers depends on the day of the reservation. For example, the mobile application software Hooper is able to tell passengers which day of the week is the best to buy their ticket (Hooper, 2017). In order to avoid any variation in the price due to the booking date and the number of days left until departure, data were entirely collected on one day: Saturday, May 6th 2017. The selected flights were all scheduled on Saturdays June 17th (outward) and 24th (return) 2017. This prevents any variation of prices due to difference of demand between days (for example, weekend flights are typically more expensive than others). Moreover, the exact time of the reservation was collected in order to control for variations within
Figure 1: Destinations
the day (through Booktimeij).
The main destinations where airlines fly to on Saturdays from European cities are typically touristic places of Southern Europe. All the selected destinations are shown in figure 1, and all routes are reported in Table 2 of the appendix, along with the 33 airlines operating on them.
The average price is € 235.12 and the minimum was found on the Sofia-Milan route, where the return price proposed by Wizzair was only € 30.36. The median price is € 199.715. All prices concerned the cheapest ticket possible: economy class, no checked in baggage. The average length of flight is about 2 hours and 40 min-utes (159.71 minmin-utes), which corresponds to a flight from Budapest to Barcelona. On a few routes, one of the airlines’ flight was complete, which prevented the ob-servation of the price. Therefore, only one obob-servation (of price) was available on seven routes (Nice-Dublin, Hamburg-Split, Catania-Berlin, Catania-Naples, Naples-Stuttgart, D¨usseldorf-Split and Bordeaux-Palma). However, the competitor’s price (the one that offers one incomplete flight) was still reported, as well as schedule and quality differentiation.
differentiation and control for price variation that might be due to demand inconsis-tencies within a day. The average time between an airline’s flight departure and its competitor’s is 269.78 minutes (outward) and 301.47 minutes (return), respectively around 4 hours 30 minutes and 5 hours.
Regarding quality factors, the volume of hand luggage accepted on board (length x width x height, in cm3) was collected as well as the seat width and ‘pitch’ (width x pitch, in square inches). ‘Seat pitch’ corresponds to ‘the distance from any point on one seat to the exact same point on the seat in front or behind it’ (Seatguru, 2017). On average, the volume of hand luggage accepted is 49217.26cm3 (which
corre-sponds to approximately 49.2 liters) and the difference between airlines is 7391.30cm3
(7,3913 liters). The difference in the size of seats is on average 24.26 square inches (154.83cm2).
The number of available seats in the economy class for the selected flights was also collected at the same time. On average 120.72 seats were available per flight, which corresponds to more or less 3/5 of the total capacity of aircrafts typically used on European routes. In order to get more specific results, the number of available seats and the difference in the departure time were collected (and included in the model) separately for outward and return flights.
4
Results and interpretations
Results of the regression defined above are shown in section 4.1. In section 4.2, comments and several interpretations are also given.
4.1
Results
The results of the regression are shown in column 1 of Table 1. There are some missing values since a few airlines do not give information needed. For example, it is impossible to get the size of hand luggage accepted by Royal Air Maroc. Therefore, the total number of observation is 157.
The coefficient of Length is 0.480 and significant at 0.1% level, which means that an increase in 10% in the length of the flight significantly increases the price
Table 1: Price of airfares
(1) (2) (3) (4)
Price Price Price Price
Length (log) 0.480∗∗∗ 0.524 0.342 0.373 (3.58) (1.32) (0.45) (0.44) ∆Lugsize 0.000582 0.0602 0.000542 0.0526 (0.07) (0.72) (0.06) (0.60) ∆Seatsize (log) 0.0471 -0.0363 0.0462 -0.0745 (1.00) (-0.06) (0.97) (-0.12) ∆Timeout (log) 0.0166 0.0138 0.260 0.189 (0.38) (0.31) (0.39) (0.26) ∆Timeret (log) 0.0249 0.0236 -0.325 -0.246 (0.54) (0.51) (-0.44) (-0.32) Sout (log) -0.377∗ -0.370∗ -0.372∗ -0.366∗ (-2.31) (-2.25) (-2.22) (-2.17) Sret (log) -0.119 -0.126 -0.118 -0.126 (-0.73) (-0.77) (-0.72) (-0.76) Booktime 0.000567∗∗ 0.000540∗ 0.000575∗ 0.000553∗ (2.62) (2.40) (2.59) (2.39) Length×∆Lugsize (log) -0.0122 -0.0106 (-0.71) (-0.59) Length×∆Seatsize (log) 0.0159 0.0237 (0.13) (0.19) Length×∆Timeout (log) -0.0493 -0.0356 (-0.37) (-0.25) Length×∆Timeret (log) 0.0709 0.0545 (0.47) (0.35) Constant 4.014∗∗∗ 3.856∗ 4.652 4.557 (4.97) (2.00) (1.22) (1.10) R2 0.371 0.373 0.372 0.374 Number of observations 157 157 157 157 t statistics in parentheses
Insignificant time dummies not included
by 4.80%. The sign is coherent with the fact that a longer flight is more costly for airlines, especially because of a greater amount of fuel consumed, resulting in higher final price.
The number of available seats on the outward flight at the moment of the reser-vation has a significant negative impact on price on the outward flight (at 5% level) with a coefficient of -0.377, meaning that reduction of 10% of available seats in-creases the price by 3.77%. The coefficient of the return flight is not significant, but its sign is coherent. Therefore, the number of available seats is only significant on the outward flight. This might be surprising, but the answer is probably that on this type of flight, people systematically choose to fly back after one week, which would mean that the number of available seats is approximately the same in both flights. This has been confirmed by the data set: the difference between number of available seats in the outward and the return flight is on average only 16.11 and the correlation between them is high: Corr(Sout, Sret) = 0.8349.
The time of the reservation seems to have a significant (at 1% level) and very light positive impact on the price. Indeed, the sign of the coefficient of Booktime is positive, but very close to zero (0.000567). Thereby, prices are likely to increase slightly during the day. However, this trend could easily vary depending on which routes were checked first. For example, it is could be that prices on routes with low-cost (cheaper) airlines were mainly collected in the morning.
Time fixed effects are not significant, meaning that prices do not vary because of the departure time. They are not significantly higher in a particular period of the day. In order to facilitate the table’s readability, they are not included in Table 1.
The first quality factor that reflects quality differentiation is the difference in the size of the seats (∆Seatsize). Although its sign corresponds to the expected one, its coefficient is not significantly different from zero, which means that we cannot conclude that this type of differentiation increase prices. Airlines use it to differentiate from each other, since the difference is on average 24.26 inches square, but it has no significant impact on prices. The second way for airlines to differentiate regarding quality is the difference in the size of hand luggage. Again, the sign of the corresponding variable (∆Lugsize) seems correct although the coefficient is
not significant. Even if airlines have different policies on hand luggage, a greater differentiation in that quality factor does not significantly increase prices.
Regarding spatial differentiation, both coefficients of ∆T imeout and ∆T imeret are not significant. Even if their sign are positive, schedule differentiation does not increase the price.
In order to test for the robustness of the results, three regressions with additional regressors were executed. Because people might care more about quality when their flight is longer, two interaction terms (Length×∆Lugsize and Length×∆Seatsize) were added in a second regression (see column 2 of Table 1). In a third regression, two other interaction terms were included in order to control for the fact that when flights are longer, passengers could potentially care more about the departure time. Indeed, for long flights, people might prefer to fly in the morning and enjoy their evening at destination. Therefore, Length × ∆T imeout and Length × ∆T imeret were added (see column 3 of Table 1). Finally, in the last column of Table 1, the four interaction terms take part of the regression. The significance of the four variables of interest does not change within all regressions. Indeed, they are all insignificant, in any case. This clearly reinforces the robustness of the results of the study.
In conclusion, all differentiation coefficients do not differ significantly from zero, meaning that differentiation in quality and schedule does not increase prices of flight tickets. Results therefore contradict the two models defined in the first part of this study.
4.2
Interpretations
In this subsection, I describe six different features that can provide answers to the contradicting results found in the previous section.
Saturday differs from other days
Although it is intuitive to think that schedule differentiation should increase prices through a division of the demand into two distinct segments, it might not be the case for Saturday flights. Indeed, since destinations in the sample are mainly touristic cities of Southern Europe, the demand for ‘vacation’ flights could be such that
people do not have strong preferences regarding schedule. Choosing the cheapest flight might be more important than the timetable. Thereby, they are likely to simply choose their flight without taking into account the schedule, which would certainly not be the case for business flights, that mainly occur in working days.
Schedule differentiation is not large enough
It might be the case schedule differentiation is not large enough to weigh in con-sumers’ preferences. Indeed, the average difference in departure time is only about 4.5 to 5 hours, whereas differentiation could be far greater. As a result, by increasing their prices, airlines would lose consumers to their competitors. In other words, the level of spatial differentiation is not strong enough to eliminate price competition, preventing airlines to increase their prices profitably.
Price competition on short-haul flights
According to the results, quality differentiation is likewise insignificant. A reason can be that passengers are not willing to pay higher prices for a better level of quality on short-haul flights (most of the selected flights are less than three hours long).
In addition, there are signs of degradation of quality provided by traditional airlines. According to an article of The Economist (2013), the gap between tradi-tional and low-cost airlines is narrowing. For example, easyJet decided to offer a ‘business’ flexible ticket, while legacy airlines started to reduce the quality of their services. It is notably the case for Swiss International Airlines, which developed a special ‘low-cost’ package (Swiss Economy Light that was first available only from Geneva and then extended to all European flights (Swiss, 2017), in order to reduce prices and compete tougher with low-cost airlines (espacially easyJet, which reached 43.6% of market share in Geneva in 2016 (Renaud, 2017)). A clear understanding of this strategy requires a deep analysis. However, one can reasonably think that
Swiss decided to offer this new package in order to serve a demand certainly more
sensitive to changes in price than changes in quality. Indeed, the quality-elasticity1
of the demand is likely to be small. On short-haul flights passengers probably do not
pay too much attention to quality differences between airlines (apart from people with disabilities or elderly).
Moreover, as Dresner et al. (1996) showed, the growth of low-cost airlines seri-ously decreased yields in the industry. Traditional airlines could thus be forced to reduce the quality of their service to remain profitable.
Complex pricing strategy
It is well known that many factors play a role in the pricing strategy of airlines. For instance, according to a short documentary of CNN (2017), airlines use very complicated algorithms to predict demand variations. The time dimension seems to have an important impact. Although all the data were collected on the same day for flights scheduled on the same date, prices might depend on the number of days passed since the beginning of the sale. For example, low-cost airlines are likely to set low-increasing prices, whereas traditional carriers might set higher-stable price from the beginning.
Moreover, some of the selected airlines offers two classes: economy and business. This combination of classes probably is probably also taken into account in the pricing strategy. In addition, several airlines (mostly traditional) provided food and drinks ‘for free’. This is a cost that low-cost airlines do not face and could explain differences in the pricing process.
Collusion
Prices might be independent from quality and schedule differentiation because of collusion. Indeed, if firms agree (tacitly) on prices, they might have an implicit agreement that prices do not depend on differentiation. In other words, because of collusion, quality and schedule differentiation do not affect prices, which could explain the results found in section 4.1.
Endogeneity of quality and time
Quality and time differentiation might be endogeneous since they depend on the competitiveness of the route. Indeed, on highly competitive routes, airlines
differen-tiate to increase their profits. However, on less competitive routes, they might not (perhaps because of collusion). As a result, even if consumers care about quality and the departure time, differentiation has little impact on prices, reflecting the results of this study.
5
Conclusion
In this study, different quality and schedule differentiation possibilities were used in the airline industry to test two IO models. According to the data, differentiation does not increase prices of flight ticket, which contradicts the models. There are several reasons that might explain this contradiction. First, passengers might not care too much about schedule on Saturday ’vacation’ flights. Secondly, the air transport market seems to be driven by price competition, which schedule differentiation is not strong enough to reduce. Moreover, passengers are probably not too cautious about quality in short-haul flights. In addition, pricing strategies of airlines are complex and many factors are taken into account. Finally, collusion and endogeneity might also have a role to play in the results of this study, by reducing the correlation between differentiation and prices.
However, this study only focused on (short) European-duopoly routes. It would be very interesting to apply such models to long-haul flights. On the one hand, quality differentiation might have a stronger impact on long-haul flight, since con-sumers are more likely to care about quality when they fly for a longer period of time. On the other hand, schedule differentiation might be even less significant in the long-haul flight market, since passengers know that they will need a few hours, or days, to deal with changes in the time zones and tiredness, for example.
In any case, the conduct of traditional airlines, which tend to decrease the quality of their product, suggests that the European air transport market is driven by price competition. It would be also interesting to go further by analyzing more into details the pricing strategy of airlines (which remains relatively secret). This would provide more information on the importance of differentiation in their pricing strategy.
in-crease prices, which reduces consumer surplus. Nonetheless, in this research I found no evidence of such impact on prices. Thereby, on the European airline market, spatial and quality differentiation provide a large diversification of flights, through different schedules and qualities, without increasing the price paid by passengers. Even though consumers do not seem to care for quality on short-haul flights, results suggest that differentiation do not decrease the consumer surplus. Therefore, there is no reason to build restrictive policies, since differentiation does not seem to affect prices but gives more opportunities to consumers.
References
Airbus. (2016, December). New Airbus aircraft list prices for 2016. Retrieved from http://www.airbus.com/presscentre/pressreleases/press-release -detail/detail/new-airbus-aircraft-list-prices-for-2016/
Asplund, M., & Sandin, R. (1999). Competition in interrelated markets: An empir-ical study. International Journal of Industrial Organization, 17 (3), 353–369. Belleflamme, P., & Peitz, M. (2015). Industrial organization: markets and strategies.
Cambridge University Press.
Boeing. (2017). Current prices. Retrieved from http://www.boeing.com/company/ about-bca/#/prices
Borenstein, S., & Netz, J. (1999). Why do all the flights leave at 8 am?: Competition and departure-time differentiation in airline markets. International Journal of
Industrial Organization, 17 (5), 611–640.
Clemons, E. K., Hann, I.-H., & Hitt, L. M. (2002). Price dispersion and differentia-tion in online travel: An empirical investigadifferentia-tion. Management science, 48 (4), 534–549.
CNN. (2017, March). Airline pricing secrets: How carriers come up with fares. Re-trieved 2017-05-25, from http://edition.cnn.com/2016/09/16/aviation/ airline-pricing-secrets/
Cotterill, R. W. (1986). Market power in the retail food industry: Evidence from vermont. The Review of Economics and Statistics, 379–386.
Dresner, M., Lin, J.-S. C., & Windle, R. (1996). The impact of low-cost carriers on airport and route competition. Journal of Transport Economics and Policy,
30(3), 309-328. Retrieved from http://www.jstor.org/stable/20053709
Fik, T. J. (1988). Spatial competition and price reporting in retail food markets.
Economic Geography, 64 (1), 29–44.
Hooper. (2017). Hooper - Aubout. Retrieved from http://www.hopper.com/corp/ about.html
Hotelling, H. (1929). Stability in competition. The Economic Journal, 39 (153), 41-57. Retrieved from http://www.jstor.org/stable/2224214
Klee, J. (2013, January). How far in advance should i book my flight? here’s what
the numbers say. Retrieved 2017-05-29, from https://www.cheapair.com/
blog/travel-tips/how-far-in-advance-should-i-book-my-flight/ Kotler, P., & Keller, K. (2012). Marketing Management (14th ed.). Upper Saddle
River, NJ: Prentice Hall.
McCannon, B. C. (2008). The quality—quantity trade-off. Eastern Economic
Journal, 34 (1), 95-100. Retrieved from http://www.jstor.org/stable/
20642395
Nevo, A. (2001). Measuring market power in the ready-to-eat cereal industry.
Econometrica, 69 (2), 307–342.
Renaud, J. (2017, April). Gen`eve A´eroport d´evoile les parts de march´e des
com-pagnies en 2016. Retrieved 2017-05-25, from http://airinfo.org/2017/04/
04/geneve-aeroport-parts-de-marche-compagnies-2016/
Seatguru. (2017). Frequently Asked Questions: What is Seat Pitch. Retrieved from https://www.seatguru.com/help.php
Shaked, A., & Sutton, J. (1982). Relaxing price competition through product differentiation. The Review of Economic Studies, 49 (1), 3-13. Retrieved from http://www.jstor.org/stable/2297136
Slade, M. E. (2004). Market power and joint dominance in uk brewing. The Journal
of Industrial Economics, 52 (1), 133–163.
Swiss, I. A. (2017). Fare concept for Europe: More choice for you. Retrieved from https://www.swiss.com/no/en/book/offers/new-fares-europe.html The Economist, C. (2013, March). Legacy vs low-cost carriers: Spot the
differ-ence. Retrieved from http://www.economist.com/blogs/gulliver/2013/
Appendix
Table 2: All selected routes and airlines
Departure Destination Airlines
Alicante Billund Ryanair, Norwegian Copenhagen SAS, Norwegian Frankfurt Ryanair, Lufthansa Gothenburg Ryanair, Norwegian Milan easyJet, Ryanair Trondheim SAS, Norwegian Warsaw Ryanair, Wizzair Amsterdam Catania KLM, Transavia
Izmir SunExpress, TUI Split easyJet, KLM Tenerife Transavia, TUI Zagreb KLM, Croatia Airlines Basel Ibiza easyJet, Niki
Berlin Malaga easyJet, Ryanair Bologna Barcelona Ryanair, Vueling
Cagliari Ryanair, Meridiana Eindhoven Ryanair, Transavia Lisbon Ryanair, TAP Bordeaux Marrakech easyJet, TUI
Lisbon easyJet, TAP Palma de Mallorca Volotea, Vueling Bristol Lanzarote easyJet, Ryanair Santa Cruz de Tenerife easyJet, Ryanair Budapest Barcelona Ryanair, Wizzair
Milan Ryanair, Wizzair Catania D¨usseldorf Air Berlin, Eurowings
Berlin Ryanair, Air Berlin Naples easyJet, Alitalia Brussels Brusssels Airlines, TUI Dusseldorf Faro Eurowings, Niki
Lanzarote SunExpress, Niki Las Palmas Eurowings, Niki
Lisbon Eurowings, TAP Malaga Eurowings, Niki Manchester Eurowings, FlyBe
Prague Eurowings, Czech Airlines Split Croatia Airlines
Fuerteventura Dublin Ryanair Aer Lingus Brussels Ryanair, TUI Geneva Catania easyJet, Swiss
Heraklion easyJet, Swiss Olbia easyJet, Swiss
Ibiza easyJet, Etihad Regional Hamburg Catania easyJet, Eurowings
Split easyJet, Eurowings Lyon Sevilla Vueling, Transavia
Vienna easyjet, Austrian Malaga Budapest Wizzair, Ryanair
Cologne Niki, Ryanair Gothenburg SAS, Norwegian Hamburg Niki, Norwegian Marseille Vueling, Ryanair Palma de Mallorca Vueling, Ryanair Stavanger SAS, Norwegian Warsaw Ryanair, Norwegian Marseille Istanbul Turkish Airlines, Pegasus
Rabat Ryanair, Royal Air Maroc Munich Alicante Norwegian, Transavia Nantes Las Palmas Vueling, Volotea
Lisbon Transavia, TAP Palma de Mallorca Vueling, Volotea Porto easyJet, Transavia Naples Berlin easyJet, Air Berlin
Casablanca Royal Air Maroc, Air Arabia Copenhagen SAS, Ryanair
D¨usseldorf Air Berlin, Eurowings Stuttgart Air Berlin, Eurowings Turin Blue Air, Alitalia
Nice Berlin easyJet, Eurowings Dublin Ryanair, Aer Lingus Dusseldorf Air Berlin, Eurowings Helsinki Norwegian, Finnair Olbia Berlin easyJet, Air Berlin Palermo Paris Ryanair, Transavia Palma de Mallorca Dresden Eurowings, Germania
Milan easyJet, Alitalia Oslo SAS, Norwegian Santiago de Compostela Ryanair, Npostrum Pisa Stockholm Ryanair, Norwegian Salzburg London Ryanair, Air Berlin Sevilla Las Palmas Vueling, Ryanair
London easyJet, Ryanair Nantes Vueling, Transavia Brussels Ryanair, Brussels Airlines Split Stockholm SAS, Norwegian
Sofia Birmingham Ryanair, Wizzair Milan Ryanair, Wizzair Rome Ryanair, Alitalia Toulouse Bastia easyJet, Volotea
Casablanca Air Arabia, Royal Air Maroc F`es Air Arabia, Ryanair
Palma de Mallorca Vueling, Volotea Rome easyJet, Alitalia Vienna Barcelona Vueling, Eurowings
Ibiza Niki, Austrian Airlines Malaga Niki, Eurowings Santa Cruz de Tenerife Niki, Eurowings
