DOES AIRLINE COMPETITION WORK IN SHORT-HAUL MARKETS?

1
DOES AIRLINE COMPETITION WORK IN
SHORT-HAUL MARKETS?
Research Unit on Public Policies and Economic
Regulation

• Xavier Fageda
• GARS workshop EU Liberalization of Air Transport
• Cologne, 21-22 November 2005

2
Introduction

• Motivation
• Positive effects of liberalization in the
  European Union (EU) depend on effective
  competition on the route
• Concern Scale advantages major airlines hold in
  their domestic markets as a consequence of their
  dominance of airport access
• Note 2 common features of EU domestic markets
• The majority of routes are short-haul routes
• Allocation of slots based on grandfather rigths
  in a context of airport congestion
• Objective
• To examine airport dominance advantages of a
  whole network in markets characterized by
  short-haul routes and congestion.

3

• Methodology
• We estimate an equation system, which is based
  on theoretical grounds, for the Spanish domestic
  market during 2001 and 2002
• Note The Spain domestic market is representative
  in the EU because is the largest market and
  airport congestion took place in the period
  considered.
• Contributions
• - We do not focus the attention on the hub
  premium but dominance of a whole national
  network.
• Note Dominance of major carriers can be even
  higher in small airports. The low cost
  phenomenon is modest in many domestic markets,
  particularly in low-density routes.
• - Product differentiation is not a
  usual assumption but it is sensible to test
  explicitly the cost and demand advantages of
  airport dominance.

4
Airline competition

• Airline competition depend on demand and supply
  side charact.
• Suply side Density economies versus
  scale economies
• Demand side - Leisure travelers
  versus business travelers
• - Frequent
  Flyer Programs (FFP) imply Switching Costs
• Given that, benefits of airport dominance in
  short-haul markets come from a high flight
  frequency
• - It reduces waiting time and allows
  a better exploitation of FFP
• - It is not necessarily cost
  damaging modest cost diseconomies of smaller
  planes, higher annual utilization of planes and
  crews, lower break-even load factors
• Main determinant of flight frequency in a given
  network (and its size) is number of slots that
  can be used allocation of slots is a crucial
  issue!

5
The empirical model

• Demand
• From a vertical product differentiation model, we
  derive the following log equations
• - Demand equation in the route k
• log(Qk) ?1?11log(Pk)?12log(Sk)?13log(Nk)
  ?14Disland ?t ?1k
• - Market share equation of airline j in route k
• log(MSjk)?2?21(Pjk/Pk)?22log(Sjk/Sk)
  ?24Disland ?t ?2k
• P is price, S is flight frequency, N is
  population of the city-pair and Disland is a
  dummy variable for islands. ?t refers to
  seasonal specific effects
• Note Prices in demand eq. are captured through
  interaction of prices in the economy class
  and a dummy variable for discounts, while prices
  in the MS eq. are captured through a dummy
  variable for discounts. Disland captures
  intermodal competition opportunities and the
  tourism effect.

6
The empirical model

• Main hyphotesis to test in MS equation a
  positive sign in the variable for prices should
  imply price competition to attract leisure
  travelers, while a positive sign in the variable
  for frequency should imply quality competition to
  attract business travelers

7
The empirical model

• Supply
• - Assumption of airlines market behaviour in a
  Cournot framework
• - Inverse market demand function Pk
  F(Qk,Sk,Nk)
• - Cost Function of airline j in market k Cjk
  Cjk (Distk,Qjk(FQjk, equipjk , lfjk), ?j)
• Dist is Distance, FQ is flight frequency,
  equip is aircraft size and w are input prices.
• Note w can be excluded from the empirical
  analysis (airline specific fixed costs).

8
The empirical model

• First order conditions from profit maximization
  by each airline lead to the following pricing
  equation Pjk ?jk(Sjk/Sk, nk)Cjk(Distk, Qjk)
• is the mark-up and n is the number of
  competitors
• Note Under Cournot, prices (in the economy
  unrestricted fare class) as mark-up on marginal
  costs. Prices in the lowest fare class can be
  understood as a discount on prices in the economy
  unrestricted fare class.
• Pricing equation of airline j in market k
  (economy unrestricted fare class)
• log(Pjk)?3?31log(Distk)?32log(Qjk)?33log(Sjk)
  ?34log(HHIk) ?t ?3jk
• HHi is Herfindahl-Hirschman Index
• Discount policy equation
• Ddiscountjk ? ?1log(equipjk/equipk)
  ?2log(APjk/APk) ?3HHIk ?4Dislandk ?t ?jk

9
Main hyphotesis to test in pricing and discount
equations a positive sign in the variable for
airport presence in the pricing and discount
equation should be consistent with competitive
advantages from airport dominance due to the
product differentiation explanation.
The empirical model
10
Data

• - Sample of 35 non-stop oligopoly routes
• - Winter and summer seasons of 2001-2002
• - Demand data does not differentiate between
  connecting and final traffic A possible
  network effect should not bias our results
• - Price data
• A weighted distribution of passengers in the
  different fare classes is not available
• To account for variability in lowest fare
  class, data obtained under homogeneous conditions
• Note Discount as a discrete variable due to
  variability Discounts are considered relevant
  when variable takes a value lower than the
  standard deviation with respect to the mean.
  Equations are also estimated using discounts in a
  continous form.

11
Estimation and results

• - Demand, market share and pricing equation
  estimated as a system (TSLS)
• - Policy discount equation estimated separately
  (logit)

12
Estimation and results
Market share equation (TSLS). Num. observations 215 Instruments for log(Pjk/Pk) log(equipjk/equipk), log(APjk/ APk) Coefficients (White standard errors Robust to heterocedasticity) Market share equation (TSLS). Num. observations 215 Instruments for log(Pjk/Pk) log(equipjk/equipk), log(APjk/ APk) Coefficients (White standard errors Robust to heterocedasticity)
Explanatory Variables Dependent Variable log(MSjk)
Intercept log(Pjk/Pk) log(Sjk/Sk) Dislandjk win01k sum02k -1.37 (0.08) 1.36 (0.31) 0.88 (0.05) 0.21 (0.1) -0.44 (0.17) 0.008 (0.05)
R2adj. F-Statistic 0.72 135.25

1. Significance at the 1 (), 5 (), 10()
Main results coefficients for price discounts
and relative quality are positive. Evidence that
airlines compete both in price and quality to
attract leisure and business passengers,
respectively.
13
Estimation and results
Pricing Equation (TSLS). Num. observations 215 Instruments for log(Qjk) log(Nk), Dislandk Coefficients (White standard errors Robust to heterocedasticity) Pricing Equation (TSLS). Num. observations 215 Instruments for log(Qjk) log(Nk), Dislandk Coefficients (White standard errors Robust to heterocedasticity)
Explanatory Variables Dependent Variable log(Pjk)
Intercept log(Distk) log(Qjk) log(Sjk) log (HHIk) win01k sum02k 3.60 (0.19) 0.43 (0.007) -0.06 (0.01) 0.08 (0.01) 0.03 (0.06) -0.02 (0.01) 0.12 (0.01)
R2adj. F-Statistic 0.95 792.21

1. Significance at the 1 (), 5 (), 10()
14
Estimation and results

• Main results of pricing equation
• coefficients for demand and distance are as
  expected. Evidence of density and distance
  economies
• coefficient for airport presence is positive.
  Evidence of the quality effect of airport
  dominance
• coefficient for intensity of competition (HHI) is
  not significant (non monopoly routes)

15
Estimation and results
Policy Discount Equation (logit). Num. observations 215 Coefficients (White standard errors Robust to heterocedasticity) Policy Discount Equation (logit). Num. observations 215 Coefficients (White standard errors Robust to heterocedasticity)
Explanatory Variables Dependent Variable Ddiscountjk
Intercept log(equipjk/equipk) log(APjk/APk) log (HHIk) Dislandk win01k Sum02k -2.48 (0.78) 2.71 (0.88) 1.24 (0.33) -0.75 (1.10) -0.97 (0.54) 2.53 (0.51) 0.79 (0.54)
Pseudo R2 Wald test (?2) 0.28 48.08
1. Significance at the 1 (), 5 (), 10()
Main results coefficient for airport presence is
positive. Evidence of the cost effect of airport
dominance
16
Concluding remarks

• 1. Aiport dominance advantages in markets
  characterized by short-haul routes and
  congestion
• - Competition in the leisure segment focused on
  prices. Major carriers can take advantage of cost
  economies
• - Competition in the business segment focused on
  quality. Major carriers can take advantge of
  demand economies.
• 2. An airline that controls an airport network
  can
• offer large discounts (higher load factors) and,
  at the same time, a convenient flight schedule
  (higher proportion of business travelers)
  This threatens the competitive position of
  its rivals
• 3. Increases in airport capacity and
  implementation of new rules for slot allocation
  could improve the scope of airline competition