Journal of Air Transport Management 76 (2019) 56–66

Contents lists available at ScienceDirect

Journal of Air Transport Management

journal homepage: www.elsevier.com/locate/jairtraman

Effect of concentration in airline market on Spanish airport technical T

efficiency
Soraya Hidalgo-Gallego∗, Ingrid Mateo-Mantecón
Department of Economics, Universidad de Cantabria, Av. Los Castros s/n, 39005, Santander, Cantabria, Spain

ARTICLE INFO ABSTRACT

Keywords: This paper tries to measure the effect of airline market concentration on airport technical efficiency. With this
Market concentration aim, a Stochastic Frontier Analysis (SFA) is applied, estimating with a one-step procedure technical efficiency
Technical efficiency and the effect of airlines' market power on that efficiency. Results suggest that when airports are lowly con-
Airports centrated increasing concentration reduces such inefficiencies. However, this effect practically disappears when
Airlines
airports are highly concentrated. Additionally, the government's policies to improve airports' performance seems
Stochastic frontier
to have a positive effect in the short run.

1. Introduction Saphiro (2009) show that airline competition increases price disper-
sion. On the other hand, Cao et al. (2017), Greenfield (2014) and
Airport activities are characterized by the vertical structure of their Mazzeo (2003) revise the topic of the effect of airline concentration on
market in which infrastructure providers and carriers are separated. In flight delays. They find a negative effect of concentration on airlines'
this way, airport supplies services directly to final consumers (cargo on-time performance.
and passengers terminals) and indirectly providing airlines inputs such However, to the best of our knowledge, only Ha et al. (2013) ana-
as terminals, runaways or storage areas (Ha et al., 2013). This vertical lyze the effect of the concentration in the airline market (downstream
structure implies the linkage between both upstream market (airport market) on airport efficiency (upstream market)1. This paper aims to
facilities) and downstream market, meaning that what happens in one contribute to the analysis of the relationship between airline market
of them likely has effects on the other one. In this line, we want to study concentration and airport infrastructure efficiency. To do this, we es-
if the structure of the downstream market (airline market) could affect timate a stochastic frontier function to a sample of 41 Spanish airports
the performance of its infrastructure provider (the airport). that allows jointly estimating in a one-step procedure the airport
In literature, it is possible to find several studies that analyze the technical efficiency and the effect of airline market concentration on
effects of airline market concentration. These works mainly focus on the such efficiency.
impact of airline market structure on airlines' prices and quality (flight The relationship between airline concentration and airport effi-
delays). With respect to the relationship between airlines' concentration ciency seems to be complex, so it is possible to find in literature justi-
and the fares that they charge, Brueckner et al. (2013), Boresnstein fications for both a positive and a negative correlation between them.
(1990), Evans and Kessides (1993) and Graham et al. (1983) find that Ha et al. (2013) enumerate arguments both for and against con-
the higher airline concentration in a given airport, the higher fees that centration based on a simple theoretical model. First, authors show that
they charge. In a similar line, Borenstein and Rose (1994), Gaggero and if congestion exists and there are not other alternative modes, when
Piga (2011), Giaume and Guillou (2004) and Stavins (2001) find a airlines have some market power a double marginalization problem
negative relationship between market concentration and price disper- appears. Therefore, passengers have to pay more than under perfect
sion, so a concentrated airline market reduce price discrimination competition market structure. This reduces demand for airports ser-
among consumers. In contrast, the results obtained by Gerardi and vices, arising over-invest problems, and so, inefficiencies. However, the

∗
Corresponding author.
E-mail addresses: hidalgos@unican.es (S. Hidalgo-Gallego), mateoi@unican.es (I. Mateo-Mantecón).
1
The text associated with this footnote should be: It is easier to find studies that analyze the effect of airport competition on airport efficiency (Pavlyuk, 2009,
2010; Scotti et al., 2012; Adler and Liebert, 2014; D'Alfonso et al., 2015). The results achieved by these authors seem to be contradictory. They show that the effect of
competition on airport efficiency can be either negative or positive so that likely other factors could affect this relationship. Merkert and Mangia (2014) go a step
further by including other transport infrastructure (surface infrastructure) as a potential competitor of a given airport.

https://doi.org/10.1016/j.jairtraman.2019.02.003
Received 25 April 2018; Received in revised form 1 February 2019; Accepted 1 February 2019
Available online 25 February 2019
0969-6997/ © 2019 Elsevier Ltd. All rights reserved.
S. Hidalgo-Gallego and I. Mateo-Mantecón Journal of Air Transport Management 76 (2019) 56–66

Table 1
Spanish airport classification.
Group Airports

Group 1 Madrid (MAD); Barcelona (BCN)

Group 2 Alicante (ALC); Gran Canaria (LPA); Tenerife Sur (TFS); Málaga (AGP); Palma de Mallorca (PMI)
Group 3 Bilbao (BIO); Fuerteventura (FUE); Girona (GRO); Ibiza (IBZ); Lanzarote (ACE); Menorca (MAH); Santiago (SCQ); Sevilla (SVQ); Tenerife Norte (TFN); Valencia (VLC)
Group 4 Almería (LEI); Asturias (OVD); Coruña (LCG); Granada-Jaén (GRX); Jerez (XRY); La Palma (SPC); Murcia (MUV); Reus (REU); Santander (SDR); Vigo (VGO); Zaragoza
(ZAZ)
Group 5 Albacete (ABC); Badajoz (BJZ); Burgos (RGS); El Hierro (VDE); La Gomera (GMZ); León (LEN); Logroño (RJL); Melilla (MLN); Salamanca (SLM); San Sebastián (EAS);
Pamplona (PNA); Vitoria (VIT); Valladolid (VLL)

above mentioned double marginalization problem could be total or year. This plan tries to adapt the services supply to the actual demand,
partially solved by a long-term collaboration between airlines and air- reducing the variable costs of those airports (CNMC, 2014). The second
ports. This collaboration is easier to reach when airports render their measure is a reduction of Aena's workforce by 11%, which directly
services to a reduced number of airlines than if many small airlines are affect technical efficiency. Another notable fact that occurs during our
using the same infrastructure (Basso, 2008). Second, their model also sample period is the growth of low-cost companies' market share. In
shows that the higher market share of an airline operator, the more general, from 2012 low-cost companies become the dominant airlines
incentives to reduce congestion, which could improve airport effi- in most Spanish airports. Therefore, on the one hand, the measures
ciency. carried out by the government will affect efficiency, while the growth of
Other arguments that explain a positive relation between airline the low-cost companies' market shares could have affect concentration
concentration and airports' efficiency could be the following ones. First, indexes.
a high concentration in the downstream market could favor better co- Aena (2012) classifies the Spanish airports into five groups ac-
ordination between airlines and airports, allowing airlines to be in- cording to the levels of traffic and the aeronautical charges. This clas-
volved in the decision-making of the facility. This kind of coordination sification has changed over the sample period. However, to achieve
or collaboration could improve input allocation and reduce passenger some degree of homogeneity, it has used for the whole sample. Table 1
traffic uncertainty. Second, Fu et al. (2011) point out long-term con- shows airports' distribution in those groups.
tracts between airports and their dominant airline guarantee airlines'
services for a long period reducing airports' uncertainty about their - Group 1 adds the results of the biggest airports in Spain, Madrid and
future revenues. Finally, Cachon and Lariviere (2005) state that vertical Barcelona airports; they moved more than 41 and 37 million pas-
cooperation could improve efficiency, as we stated above, this co- sengers in 2014, respectively. In each of these airports, more than 80
operation is less complicated with a few airlines with high market airlines are offering regular flights to more than 200 different des-
shares. tinations (national and international). However, nearly 50% of those
The structure of the paper is the following one. In Section 2, the traffics are operated by three companies in Madrid (Iberia, Ryanair
Spanish airport system is briefly described. In Section 3, the metho- and Air Europa) and two in Barcelona (Vueling and Ryanair).
dology applied and the econometric specification are displayed. In - Group 2 includes airports specialized in tourism, with traffic levels
section 4, data sources and variables' definition are explained. Finally, generally higher than 7 million passengers per year. In the airports
Sections 5 and 6 show the results and main conclusions, respectively. belonging to this group, there are two or three airlines companies
that attend the 40% of their traffic, approximately, except Gran
Canaria where most of the traffic is inter-island traffic attended by
2. Spanish airport system
numerous little companies. Additionally, in most airports in this
group, the dominant airline is a low-cost company.
Until 2015, Aena was a public firm, dependent upon the Ministry of
- Group 3 comprises those airports located in large cities or tourist
Transport. Aena manages 46 Spanish airports and two heliports of na-
destinations that generally move between 2 and 7 million passen-
tional interest on a centralized basis. Additionally, through its sub-
gers yearly. From the point of view of the airline market shares, this
sidiary Aena International, operates the infrastructure management of
is a quite heterogeneous group. On the one hand, it is possible to
16 airports in four other countries (United Kingdom, Mexico, Jamaica,
find airports in which one or two airlines service most of the pas-
and Colombia). Overall, Aena moved more than 105 million passengers
senger traffic, for instance, Ryanair's market share in Gerona is 95%;
in 2014 (Aena, 2014; and Hidalgo-Gallego et al., 2017). In 2015, Aena
or, in Seville, Ryanair and Vueling joint market share is 75%. On the
was the object of a privatization process in which 49% of the company's
other hand, there are airports without a clear dominant airline as
capital went into private hands, while the Spanish government main-
Fuerteventura and Lanzarote (the dominant airline's market share of
tained the remaining 51%. However, this model of privatization has not
in these airports is less than the 20%).
been the only one that was proposed for Aena. The previous govern-
- Group 4 lists the airports that move between 0.5 and 2 million
ment proposed the privatization of the two main hub airports in Spain,
passengers. The average number of airlines operating in an airport
Madrid and Barcelona. After the general elections in 2012, the newly
of this group is 10. One or two of these airlines have a high market
elected government opts for a privatization model that maintains the
share (over 60%), with the rest sharing the other 40%. Most of the
centralized management of the system and the decision power in public
dominant airlines in the airports belonging this group are low-cost
hands. A series of measures accompanied both privatization policies. In
companies with market shares that can get 95%. This fact shows the
the first stage, when the plan was to privatize the airports of Madrid
importance of these companies for regional or peripheral airports.
and Barcelona, the aim was to improve the efficiency of these airports
- Group 5 agglutinates airports with traffic levels lower than 0.5
to make them more attractive to investors. However, when the objec-
million passengers. In these airports, a few airline companies can be
tive was to privatize 49% of the company as a whole, efforts were
found, in some cases, just one. In ten of the thirteen airports that
aimed at improving the efficiency of smaller airports with lower traffic
constitute this group, the dominant airline is Air Nostrum with
volumes. Among these measures, two must be highlighted because of
market shares in the region of 90%. The exceptions are: On one
their effects on technical efficiency. The first one is the Airport Effi-
hand, Valladolid with Ryanair as dominant airline following by Air
ciency Plan (Plan de Eficiencia Aeroportuaria) that affects the smaller
Nostrum; and, on the other hand, La Gomera and El Hierro (both
airports, those belonging to group 5 with less of 500,000 passengers per

57
S. Hidalgo-Gallego and I. Mateo-Mantecón Journal of Air Transport Management 76 (2019) 56–66

Fig. 1. Spanish airport system. Source: Hidalgo-Gallego et al. (2017).

located in the Canary Islands) where one and two local airlines population (Dargay and Hanly, 2001; Njegovan. 2006, Sivrikaya and
companies, respectively, present the higher market shares. Tunç, 2013).
The input-oriented distance function (DIit ) is defined as the max-
Fig. 1 shows the spatial situation of the 41 Spanish airports covered imum scalar by which all inputs can be proportionally divided without
in this study. We can observe at first glance the importance of tourism. varying the level of output (Coelli et al., 2005). Formally, the input
In fact, 11 airports are located in the Canary and Balearic Islands. distance function is defined on the input set, L ( qit ) , as:

3. Methodology and empirical specification x it

DIit ( x it , qit ) = max : L ( qit )
(1)
3.1. Methodology
where the input set L ( qit ) represents the set of all input vectors, x it , that
This research attempts to analyze how airline concentration affects
can produce the output vector qit , (Coelli et al., 2005).
airport inefficiency. Early research evaluated the effects of external
A value of the distance function equal to one means that firms
variables on efficiency by carrying out a two-steps model. First, in-
produce efficiently in a technical way, values over one implies the ex-
efficiency was estimated. Then, in a second stage, the inefficiencies
istence of technical inefficiencies. In logarithmic terms, the distance
were regressed on external explanatory factors to evaluate their effect
function can be specified as equation (2).
on these indices. When stochastic frontier methods are applied, this
approach presents problems of biased results (Wang and Schmidt,
lnDIit = lnDI (xit , qit ) + i + vit ;
2002) and consistency (Chang and Tovar, 2014). To solve these pro-
blems, methods based on simultaneous estimation arise, approaches i = 1,2, …, N ;
such as the ones proposed by Battese and Coelli (1995) and Greene t = 1,2, …, N (2)
(2005) that allows expressing inefficiency as a function of firm-specific
factors. where DI ( x it , qit ) is the input-oriented distance function, x it is the
Following studies mentioned above, in this research, airport effi- vector of inputs, qit is the vector of outputs, i relates to the ith airport
ciency and external factors' effects are estimated in the same procedure. and tth relates to the time period. The error component vit represents the
Specifically, we apply the methodology proposed by Greene (2005). It random disturbance identically, independent and normally distributed
allows taking out time-invariant heterogeneity form the inefficiency with zero mean; and i represents the firm-specific effect of the ith air-
component, separating inefficiency and unobserved heterogeneity port.
among airports. Equation (2) can be transformed into equation (3), so inefficiencies
An input-oriented distance function is specified to estimate airports' are included in the term uit (uit = lnDIit ) . Additionally, it is assumed that
inefficiencies. The input orientation has been adopted because the inefficiencies follow a non-negative truncated normal distribution
Spanish airport system is a high-regulated state-owned sector, which (uit ∼iid N+ (µit , uit )) . Furthermore, exogenous factors could affect in-
has no too much control over traffics (Tovar and Martin-Cejas, 2009). efficiency in two ways: firstly, they may affect the position of the in-
This control belongs to airlines (Abrate and Erbetta, 2010), being the efficiency distribution (4), i.e., they affect the mean of the distribution;
air ticket price one of the main determinants of air transport demand second, they may scale inefficiency distribution by affecting its variance
with other exogenous factors such as economic growth, distance or (5) (Belotti et al., 2012).

58
S. Hidalgo-Gallego and I. Mateo-Mantecón Journal of Air Transport Management 76 (2019) 56–66

ln1 = lnDI (x it , qit ) + i + vit uit ; relates to the ith airport and tth relates to the time period. The error
components vit represent random disturbances identically, independent
i = 1,2, …, N ;
and normally distributed with zero mean; while uit represents technical
t = 1,2, …, N (3) inefficiencies identically, independent distributed following a truncated
normal distribution with mean µit ; and i collect the ith airport-specific
µit = µ hit effect.
(4) Symmetry is also required so:

rj = jr ; mn = nm ; rm = mr (8)
uit = u hit
(5) Variables in the distance function are divided by their geometric
means so that first-order coefficients can be interpreted as elasticities at
On the one hand, equation (4) shows that differences in the mean
the sample mean.
values of the truncated normal distribution among airports exist, being
Finally, equations (9) and (10) allow us to determine the effect of
function of a collection of external variables hit . On the other hand, exogenous variables on the distribution of inefficiency. These equations
are estimated jointly with the former distance function.
equation (5) shows how heteroskedasticity in inefficiency distribution

is allowed and what the impact of those external factors hit on it is. µit = 1 hit + µit (9)

Equations (3)–(5) are estimated in one-step procedure by Maximum

Likelihood Estimation (MLE). Additionally, Wang (2002) propose a uit = exp 1 hit + it )
(10)
formulation to compute non-monotonic effects of the exogenous vari-
ables contained in 4 and 5 on efficiency. This procedure can be im- where µit represents the mean values of inefficiency distributions, uit is
plemented after estimating (3)–(5). the heteroscedasticity variance of the inefficiency distribution and hit is
Finally, technical efficiency is obtained following Jondrow et al. a vector of exogenous factors that affect both, the location and the scale
(1982) who gets the predicted value of inefficiency from the conditional of the inefficiency distribution. Finally, µit and it are random vari-
expectation of uit (equation (6)). ables not necessarily identically distributed.
TEit = E [exp ( uit )| it ] (6)
4. Data
where it = vit uit .
4.1. Variables
3.2. Econometric specification
The above empirical specification has been estimated with data
Input-oriented distance function needs to satisfy some regularity from a sample of 41 Spanish airports for the period 2009–2014. In
properties such as homogeneity of degree 1, non-increasing in outputs, 2015, Aena was partially privatized (49% of its capital was sold to
non-decreasing in inputs and concave in inputs. The first one can be private investors; the other 51% was kept in Spanish Government
imposed with some functional forms (Cobb-Douglas and Translog) by hands). From that year, airports' financial data was not published
some restrictions over the parameters. Otherwise, the others have to anymore. Thus, we have not statistical information to carry out the
been checked once the estimation is performed. analysis from that date. Aena manages 46 airports and two heliports in
We adopt the translog functional form for the input-oriented dis- Spain, so five airports (Córdoba, Madrid-Cuatro Vientos, Son Bonet, and
tance function (7) because it allows imposing homogeneity of degree Huesca) and the two heliports have not been included in the sample.
one, is easy to calculate and does not place restriction on input sub- The excluded airports present the particularity of not providing regular
stitution. Additionally, in order to be able to estimate equation (3), flights or being too small. Hence, the sample consists of 41 airports
homogeneity of degree one of the distance function in variable inputs is observed during six years, i.e., 246 observations. Financial and traffic
enforced by normalizing the function with respect one input: data were obtained from Aena, while other statistical information was
R 1 M
1
R 1 R 1 gathered from the Spanish Statistical Office (INE). This set of data was
lnxsit = i + r lnx rit + m lnqmit + rj lnx rit lnx jit used before in Hidalgo-Gallego et al. (2017). Table 2 shows the main
2
r=1 m=1 r=1 j=1
descriptive statistics of the variables included in the model.
M M R 1 M
1 1
+ mn lnqmit lnqnit + rm lnx rit lnqmit + vit
2 m = 1 n= 1 2 r=1 m=1 4.1.1. Output
From the output side, we include two outputs in the input-oriented
uit (7)
distance function: workload units (wlu ) and commercial revenues
where x rit = xrit
xsit , x r is the rth normalized input, qm is the mth output, i (com ).

Table 2
Descriptive statistics of the variables.
Source: Own elaboration based on Aena's data
Variable Units Mean St.Dev Min Max

xL Workers 235.6 273.4 13 1713

xC Millions of constant euros year 2011 150.2 391.2 3.3 2877.1
xIC Millions of constant euros year 2011 24.6 56.7 0.9 311.46
qWLU Passenger and hundred tons 4,880,243 9,607,788 1211 5.36E+07
qCOM Millions of constant euros year 2011 16.1 34.8 1.00E-05 211.38
ratcargwlu Tons per workload units 4.24 16.2 0 98.35
HHI Continuous variable between 0 and 1 0.33 0.26 0.05 0.97

59
S. Hidalgo-Gallego and I. Mateo-Mantecón Journal of Air Transport Management 76 (2019) 56–66

Workload units are used as a proxy of aeronautical services pro- concentration of industries, and in turn is a handy tool to identify the
vided by Aena's airports. A workload unit is a passenger or 100 kg of degree of concentration in any economic system, also for the airport
cargo (Salazar de la Cruz, 2013). So this variable is defined as follow: system. It is part of the Hannah-Kay index family.
cargo It is defined as the sum of the squares of the market share of each
wlu = passengers + company. In our case, this participation is measured by the number of
100 (11)
passengers of each airline divided by the total passengers3 at each
On the other hand, commercial (non-aeronautical) revenues (com ) airport, as equation (13) shows:
approximates the level of commercial activity developed by an airport.
Because of the heterogeneity of these activities and the lack of price L

information we include a financial variable to approximate airport HHIit = Slit2

l=1 (13)
outputs instead the physical units of this output (Sarkis, 2000; Pacheco
and Fernández, 2003; Oum et al., 2008; Barros, 2008; Barros and Dieke, where S is the market share of the lth airline company in the ith airport
2008; Abrate and Erbetta, 2010). in ltth period.
Herfindahl Indices range from zero to one: zero represents a case of
4.1.2. Inputs and their prices perfect competition in which companies share the market equally, and
One of the main problems with the data published by Aena is that it one implies a monopoly in which a single company controls the entire
does not provide information about the physical units of inputs used in market. Reviewing the existing literature in market concentration, we
airport operations. Because of this, we have estimated input quantities have found that some organisms dedicated to the control and legislation
by dividing input expenses by input prices. However, Aena neither of the markets have established different ranges of these indices to
provides information about the prices at which it hires its inputs, so we describe the competitive environment in a given industry. In this study,
have had to approximate them following previous literature as it is we follow the classification proposed by the U.S. Department of Justice
explained below 2: and the Federal Trade Commission (2010). This organism defined the
Labour price (Bottasso and Conti, 2012): the price of labour is ob- following ranges:
tained from the Annual Labour Cost Survey (Encuesta Anual de Costes An HHI index less than 0.01 indicates a highly competitive industry.
Laborales) published by the Spanish Statistical Office (INE) as the An HHI index less than 0.15 indicates a non-concentrated industry.
average wage paid to the employees in activities related to warehousing An HHI index between 0.15 and 0.25 indicates an industry with a
and transport in Spain. Because of this wage is a national average, it has moderate grade of concentration.
been weighted by a ratio between regional and national general labour An HHI index higher than 0.25 indicates the existence of a high
cost per worker, to collect differences in labour cost among Spanish concentration in the market.
regions. By dividing labour cost by our estimated labour prices, we
obtain a proxy of the number of workers at each airport.
Capital price: to estimate the capital price we follow the procedure 4.2. Airline market concentration
proposed by the OECD (2001), which allows us to calculate the price of
capital at a macroeconomic level, based on the perpetual inventory In Fig. 2 the evolution of the means of the Herfindahl indexes by the
method of Jorgenson and Griliches (1967). Therefore, we can define the different airport groups are presented (GI, GII, GIII, GIV and GV), also
price of capital (r) as equation (12) states: the mean of the Spanish airport system (AENA) from 2009 to 2014.
We can notice that there is certain general stability between the first
r = bip (rate + dep) (12) and the last year under study. The mean of the system and the groups 4
where bip is a building index price of public works (obtained from the and 5 present an HHI index higher than 0.25 indicating the existence of
reports of Confederación Nacional de la Construcción, SEOPAN), rate is a high concentration level in the market. However, group 1 and 2 has
the long-term interest rate, and dep is the depreciation rate. The de- an HHI index inferior to 0.15 indicating a non-concentrated industry. In
preciation rate (dep ) is obtained by dividing the annual depreciation addition, between those extremes appears group 3 with an HHI index
expenditures of each airport by their fixed assets. between 0.15 and 0.25 indicating a moderate grade of concentration.
Intermediate consumption price: we approximated this price by the However, there is great variability within each group as we can see in
purchasing power parity index at a regional level proposed by Costa Table 3. Table 3 shows the descriptive statistics of these indices by
et al. (2015) in order to collect regional differences in prices. The group.
temporal evolution of this index follows the growth of the national Regarding group 1 the lowest level of concentration correspond to
consumer price index (INE, several years). Barcelona whereas the highest to Madrid. This group presents a parti-
As we said before by dividing input costs by their respective prices, cularity concerning the evolution of these indices. During our sample
we obtain proxies of input quantities that take into account prices period, Barcelona has increased its level of traffic concentration while
evolution. in Madrid it has been reduced. Group 2 shows the lower levels of
Additionally, a control variable related to traffic composition is concentration in both its minimum and maximum, besides the varia-
included in the input-oriented distance function estimation. This vari- bility within this group is also the most moderate. Finally, groups 3, 4
able (ratcargwlu ) shows the weight of cargo in the total workload units and 5 show a significant dispersion in the concentration of the airports
moved by a given airport in a given period. that formed these groups. In the three groups, it is possible to find lowly
or highly concentrated airports, being their variability too much higher
than in groups 1 and 2. Therefore, there is no clear relation between the
4.1.3. Exogenous factors of technical efficiency
size of the airport and the level of concentration.
To end this section, we describe the exogenous variable include in
our model. As we said before, we try to explain the effect of airline
market concentration on airports' (infrastructure managers) perfor- 3
The information used to calculate these indices comes from Aena. Aena has
mance. Therefore, we include in our model a measure of market con- an extensive database about the traffics of her airports. This database allows
centration obtained by the Herfindahl and Hirschman indices. arranging information over different criteria. In this sense, we have extract
Herfindahl-Hirschman Index (HHI) is often used to measure the traffic information of each airport in each year classified by the airline com-
panies which chapter these traffics. To do this, we have to treat, analyze and
obtain the statistical information from 246 different reports, one per observa-
2
For a more extended explanation see Hidalgo-Gallego et al. (2017). tion.

60
S. Hidalgo-Gallego and I. Mateo-Mantecón Journal of Air Transport Management 76 (2019) 56–66

Table 4
Maximum likelihood estimates of the model.
Variable Coefficient Std.Error z P-Value

Stochastic distance frontier model

ln xL 0.546 0.07 7.83 0
ln x C 0.105 0.06 1.75 0.08
ln xIC 0.349 0.071 4.89 0
lnqWLU −0.156 0.086 −1.81 0.07
lnqCOM −0.171 0.071 −2.42 0.016
ln xLlnxL 0.129 0.111 1.17 0.241
ln x C lnx C 0.596 0.105 0.57 0.571
ln xIC lnxIC −0.471 0.171 −2.74 0.006
lnqWLU lnqWLU 0.011 0.042 0.28 0.781
lnqCOM lnqCOM −0.007 0.013 −0.52 0.606
ln xLlnx C −0.33 0.095 −3.47 0.001
ln xL ln xIC 0.2 0.115 1.75 0.08
ln x C ln xIC 0.27 0.104 2.61 0.009
ln xL ln lnqWLU 0.045 0.053 0.84 0.402
Fig. 2. HHI from 2009 to 2014 by airport group and the evolution of the System ln xL lnqCOM −0.006 0.051 −0.12 0.903
(AENA). Source: Own elaboration based on Aena's data ln x C ln lnqWLU −0.165 0.053 −3.12 0.002
ln x C lnqCOM 0.095 0.044 2.16 0.031
ln xIC ln lnqWLU 0.12 0.046 2.6 0.009
Table 3 ln xIC lnqCOM −0.088 0.05 −1.78 0.076
Descriptive statistics of the concentration indices by group of airport. lnqWLU lnqCOM −0.032 0.038 −0.85 0.394
Source: Own elaboration based on Aena's data year2009 −0.657 0.042 −1.56 0.12
year2010 −0.091 0.04 −2.26 0.024
Group Mean Std.Deviatiom Minimum Maximum
year2011 −0.178 0.344 −5.17 0
year2012 −0.26 0.034 −7.64 0
1 0.14 0.04 0.07 0.21
year2013 −0.016 0.03 −0.55 0.582
2 0.09 0.03 0.05 0.15
lnratcargwlu −0.026 0.008 −3.26 0.001
3 0.21 0.22 0.05 0.91
4 0.32 0.17 0.11 0.97
5 0.56 0.24 0.06 0.97 * Significant at 10%, ** Significant at 5% and *** Significant at 1%.

Table 5
5. Results Maximum likelihood estimates of the model (cont.).
Variable Coefficient Std.Error z P-Value
5.1. Distance function estimation
Inefficient effects model
The input-oriented distance function (equations (7), (9) and (10)) µ
Herfindahl −4.078 0.826 −4.94 0
has been estimated using the maximum likelihood technique. As we
said before and in order to exploit the panel data structure, a true fixed
u
Herfindahl −4.606 0.706 −6.52 0
effects model has been applied which allows us to take into account the v
possible existence of time-constant unobserved effects. A Wald test of Const −4.876 0.246 −19.8 0
joint significance is performed for the second order coefficients to test Variance parameters
0.538
the suitability of the Translog functional form versus the Cobb Douglas. E ( u)
0.087 0.011 8.12 0
The test result shows that we can reject the hypothesis that all second v

order parameters are equal to zero at a level of significance of 99% * Significant at 10%, ** Significant at 5% and *** Significant at 1%.
(ϰ2 = 27.96) which suggests rejecting the hypothesis of constant sub-
stitution. The Cobb-Douglas estimation is collected in Tables 7 and 8
(Annex 1). efficient frontier), but it disappears in the last years. This fact could be
The result of the estimation of the input distance function jointly related to a general traffic recovery after a valley period that coincides
with the effects of concentration on efficiency are displayed in Tables 4 with the economic crisis.
and 5. Firstly, regarding the coefficients of the distance function, it can
be seen that all the first order parameters are statistically significant 5.2. Technical efficiency
and present the required signs theoretically. We observe that the esti-
mation satisfies the regularity conditions above mentioned: the input- The results obtained from the previous estimation show that, on
oriented distance function is non-decreasing and quasi-concave in average, Spanish airports operate at efficiency levels of 91.52%, which
variable inputs prices, non-decreasing in outputs and homogeneous of implies that they can reduce proportionally the input used by an 8.48%
degree one in inputs. A positive sign associated to the inputs means that and keep producing the same level of output. Airports of Gerona, La
an increase in the use of inputs given all other variables unchanged Gomera, Melilla and Badajoz present the higher levels of technical ef-
implies a loss of technical efficiency, i.e., an increase of the distance ficiency; they are relatively small and highly concentrated airports. On
from the technical optimum. Conversely, a negative sign related to the other hand, Vitoria is the most technical inefficient airport. This
outputs implies an improvement of technical efficiency when outputs airport was built to move both, passengers and cargo. However, in the
grow while other variables do not change (a decrease of the distance). years of the sample, the movements of passengers in this airport are
Additionally, the sum of the first-order output parameters is less residual (just some private or non-regular flights), which could make
than one in absolute value, indicating the presence of increasing returns the infrastructure allocated for passengers underutilized. Finally, the
to scale. Year-specific dummy variables collect technical change. These most important Spanish airports regarding passengers -Madrid,
dummies show that technological change occurs in the first years of the Barcelona and Palma de Mallorca-show levels of efficiency of 90.91%,
sample (these negative parameters indicates an approach to the 87.45% and 87.79%, respectively.

61
S. Hidalgo-Gallego and I. Mateo-Mantecón Journal of Air Transport Management 76 (2019) 56–66

efficiency, a second estimation (Tables 9 and 10 in Annex 2) have been

carried out. In this estimation, we have added a dummy variable
(efplan ) jointly with the concentration level to explain airports' in-
efficiencies. This dummy takes values equal to one for the years after
2012, to collect the effect of the Spanish government's measures during
that period. On the one hand, the value and sign of the parameters of
this second estimation are quite similar to the first one, so the robust-
ness of estimation is checked; on the other hand, as we expected, the
sign of the coefficient associated with this dummy variable shows that
these measures have contributed to reducing technical inefficiency.

5.3. Effect of concentration on airport efficiency levels

Finally, in this subsection, we analyze the effect of airline market

concentration on airports' efficiency. As we mentioned above, com-
paring Figs. 2 and 3 shows a positive correlation between concentration
and efficiency, which meets with the result of the estimation of the
Fig. 3. Evolution of the efficiency levels by groups of airports (means of tech- inefficient model in Table 5. The results of this model indicates that, on
nical efficiency of each group vs Aena). the one hand, concentration reduce inefficiencies due to it affects the
mean of the inefficiency distribution negatively; and, on the other hand,
the level of concentration reduces the uncertain about passenger traffics
Regarding the efficiency of each group of airports, group 5 (95.6%) by reducing the variance of inefficiency distribution4. Therefore, results
presents the highest level of efficiency on average. Airports collected in show a positive relationship between concentration and efficiency. This
group 5, except Vitoria, present levels of efficiency over 91.52% (the result could be explained, as we pointed out above, by dominant air-
average of the sample). In this group, the smallest airports in terms of lines incentives to reduce congestion, better coordination between
passenger traffic are collected. Secondly, group 4 present an average airlines and airports, or long-term contracts between airports and air-
level of efficiency close to the average of the whole sample (92.8%). lines. (1) Regarding the relation between concentration and congestion,
This group comprises airports located in small tourist destinations or it is observed that the airports in the sample facing more important
medium-size cities. Thirdly, we find group 1, corresponding with the congestion problems (Madrid, Barcelona or some airports located in
hub airports in the system (Madrid and Barcelona), with an average important tourist destinations) present low levels of concentration. On
level of efficiency of 89.18% (below the sample's average). Finally, the the other hand, these airports have the highest volumes of traffic, which
most inefficient groups are group 3 (88.4%) and group 2 (84.9%), both is a significant source of congestion. Therefore, in the case of Spanish
formed by airports located in important tourist destinations or im- airports, the effect of concentration on congestion is unclear. (2) Sec-
portant cities regarding population and business. Those airports located ondly, long-term contracts between airports and airlines reduce the
in tourist destination are characterized by the temporality of their airport's uncertainty. These contracts are crucial for those small or
traffics, which could affect the efficient use of their resources, especially medium-sized airports (belonging to groups 4 and 5) that are highly
in their valley periods, when the volume of traffic decreases con- concentrated. During our sample period, in order to get the permanence
siderably (more than 50%) with respect peak months. Additionally, if of a given route o airline at a given airport, the regional and local
we compare these results with Fig. 2, it is possible to check that the government signed financial agreements with the dominant airlines to
more concentrated airports, the more efficient ones, showing a positive contract promotion services. In this way, public funds were officially
correlation between concentration and efficiency. used to pay airlines for advertisement services, when the final aim of
Fig. 3 shows the evolution of the annual technical efficiency these transfers was that these airlines kept their services in that airport
throughout our sample period for the whole system (Aena) and at group for a given period (Núñez-Sánchez, 2015). Air Nostrum and Ryanair
level. Group 5 is the more efficient group during the entire period, were the companies that received most public funds, being these air-
followed by group 4 that occupies the second position. The fourth po- lines the dominant and, in some cases, the only one operating in those
sition is holding by group 3. Group 1 is the most inefficient group in the regional airports (CNC, 2011). (3) Finally, and in a similar line to the
first year of the sample but from 2011 become the third most techni- previous, given that airports' technical efficiency is being analysed, i.e.,
cally efficient group. The gains of efficiency achieved by the hub air- the volume of inputs hired by airports to produce their output, in this
ports (group 1) from 2009 to 2011 could be attributed to two facts that case, to move passengers and cargo; the better coordination between
co-occur in that period. The first one is the recovery of their traffic after airport and the dominant airline or airlines, the more optimal input use
the economic and financial crisis. The second one is the first attempt of by the first.
privatization addressed to these two airports, in which Spanish gov- However, this result is partially in conflict with that achieved by Ha
ernment tries to improve Madrid and Barcelona airports' performance et al. (2013). These authors find a U-shaped relationship between air-
to increase their attractiveness for future private investors (Hidalgo- port efficiency and airlines concentration. In this line, authors show
Gallego et al., 2017). In 2012, it can be seen that there is a general loss that when the level of concentration is low, increasing it improves in-
of efficiency. In this period, with a new political party in the govern- frastructure efficiency. Nevertheless, if the level of concentration is
ment, the original idea of privatizing Madrid and Barcelona is changed, high, the more concentration, the higher airport inefficiencies. To
and a new process starts with the objective of privatizing a share (49%) check, how the effect of concentration on efficiency varies across dif-
of the whole system in 2015. With this new schedule of privatization, a ferent levels of concentration we have estimated the non-monotonic
series of measures are carried out to improve the system efficiency, efficiency effects following Wang (2002).
especially, smaller airports belonging to group 5. Additionally, the re-
duction of Aena's workforce seems to be a significant impact in airports 4
In Annex 1, we present an additional estimation. In this estimation, we
efficiency because from 2012 technical efficiency grows, mainly in the measure the effect of concentration using a Gini index instead of the Herfindahl
large and medium airports (groups 2 and 3). index. The estimated effect of concentration on technical efficiency is quite
In line with the idea of these measures have improved technical similar in both estimation, so the robustness of the result is checked

62
S. Hidalgo-Gallego and I. Mateo-Mantecón Journal of Air Transport Management 76 (2019) 56–66

Table 6 infrastructure technical efficiency. A Stochastic Frontier Approach is

Marginal effects of concentration. used to estimate jointly, in a one-step procedure (Greene, 2005), air-
Group Over the mean Over the variance ports' technical efficiency and the effect of airlines' concentration on
such efficiency. The main advantages of the applied methodology are
Mean Std. Err Mean Std. Err the following ones. First, the estimation of both, technical efficiency
and concentration effects in just one-step avoids biased and consistency
1 −0.653 0.031 −0.154 0.014
2 −0.801 0.018 −0.228 0.01 problems. Second, using a fixed effects estimator allows collecting the
3 −0.588 0.032 −0.143 0.011 unobserved heterogeneity among Spanish airports. Finally, this proce-
4 −0.367 0.022 −0.059 0.006 dure enables calculating the non-monotonic marginal effects of con-
5 −0.186 0.024 −0.027 0.007 centration on technical efficiency. To the best of our knowledge, only
Ha et al. (2013) have analysed the relationship between airport effi-
ciency and airline market concentration before, using a different pro-
cedure.
The main results achieved could be meaningful to policymakers and
are the following ones. On the one hand, the airline markets in Spanish
airports are quite heterogeneous regarding concentration. Secondly, we
find a positive but non-monotonic relation between airlines' con-
centration and airports' efficiency. The estimation of marginal effects of
concentration on efficiency shows that the gains of efficiency related
with increasing concentration are reduced along market concentration
grows. Therefore, the benefits of increasing concentration are higher in
lowly concentrated airports than in those with greater airline market
power. Finally, we find that the different policies and measures carried
out to improve airports performance seem to be effect on technical
efficiency in the short run.
The results suggest that an improvement in technical efficiency
could be carried out through a review of the number of companies and
their participation in airports' traffics. However, this kind of measures
should be taken in an individualized way. Spanish airport system is
managed in a centralized way, but airport policies should take into
account the specific characteristics of each airport. Our results show
Fig. 4. Marginal effects of concentration on technical inefficiency.
that policies seem to be more effective when they are addressed to a
particular group of airports, for example, the Airport Efficiency Plan for
the regional airports.
These effects allow us to improve the comprehension of the re- Finally, it is worth to point out the following. Technical efficiency
lationship between airport inefficiency and airlines concentration, and analysis is helpful to orientate airport managers in the optimal use of
so obtain results more informative (Wang, 2002). Two variables having inputs given a technology. However, in some cases, from a managerial
a non-monotonic relationship implies that their values have not to be point of view, the analysis of airport cost or profitability efficiency
the same along the sample, but they can be related to a different extent could be more relevant (Merkert and Assaf, 2015). For instance, during
in different parts of the sample. Regarding the effect of concentration our sample period, Spanish airports are managed by Aena, a public-
on inefficiency, it may not be the same for the different levels of con- owned entity, through a subsidy system. In this period, minimizing cost
centration. Table 6 displays the average marginal effects of airline or maximizing profit did not seem to be an essential issue for Aena's
market concentration on the two statistics of efficiency, E (uit ) and managers. The primary efforts were focused on increasing airports'
V (uit ) by group of airports. It can be checked that the highest marginal traffic and adapt the volume of inputs to the level of output. This be-
effect corresponds to group 2 (this group presents the lowest level of haviour may justify the estimation of technical efficiency because it
concentration on average), followed by group 1 (the second less con- does not assume cost minimization or profit maximization. However,
centrated group) and so. These results suggest that the effect of con- from 2015, this entity becomes in partially private-owned one. In this
centration on the distribution of efficiency is higher with lower levels of new stage, the maximization of profits becomes more relevant than in
concentration. This idea is supported by Fig. 4. Fig. 4 shows the relation the previous, mainly for private investors. Therefore, if the sample
between concentration and its marginal effects on the inefficiency and period were extended to the present, this new analysis should try to
uncertainty in passenger traffics, respectively. Increasing the level of include the estimation of a cost or profit frontier, which would depend
concentration when the airline market in a given airport is little con- on the availability of statistical information on the prices of inputs and
centrated reduces inefficiencies and uncertainties to a greater extent outputs.
than when the airport is highly concentrated. In the future, an interesting research line, if data were available,
would be to measure the effect of the privatization on airline market
6. Conclusions concentration and the effects on airport efficiency.

This paper tries to contribute to the analysis of the vertical relations

in the air sector. There is a vast literature that studies the vertical re- Acknowledgment
lationships between airlines and airport infrastructure providers from
different points of view (theoretical, descriptive, empirical …). This Authors want to express their gratitude to Ramón Núñez-Sánchez
study estimates the effect of airline market concentration on airport for his ideas and suggestions.

63
S. Hidalgo-Gallego and I. Mateo-Mantecón Journal of Air Transport Management 76 (2019) 56–66

ANNEX 1

Table 7
Maximum likelihood estimates of other models

Variable Cobb-Douglas Translog Gini index

Coefficient Std.Error Coefficient Std.Error

ln xL 0.623*** 0.043 0.544*** 0.044

ln x C 0.158*** 0.041 0.375*** 0.047
ln xIC 0.219*** 0.039 0.081*** 0.034
lnqWLU −0.164*** 0.041 −0.299*** 0.06
lnqCOM 0.010 0.150 −0.209*** 0.05
ln xLlnxL – 0.086 0.071
ln x C lnx C – −0.561*** 0.116
ln xIC lnxIC – −0.030 0.087
lnqWLU lnqWLU – −0.136*** 0.033
lnqCOM lnqCOM – −0.053*** 0.012
ln xLlnx C – 0.22*** 0.072
ln xL ln xIC – −0.081** 0.034
ln x C ln xIC – −0.306*** 0.072
ln xL ln lnqWLU – 0.141*** 0.026
ln xL lnqCOM – −0.07** 0.024
ln x C ln lnqWLU – 0.180*** 0.038
ln x C lnqCOM – −0.155*** 0.038
ln xIC ln lnqWLU – −0.321*** 0.045
ln xIC lnqCOM – 0.225*** 0.039
lnqWLU lnqCOM – 0.042* 0.023
year2009 −0.065** 0.035 0.154*** 0.027
year2010 −0.123*** 0.034 0.1114*** 0.025
year2011 −0.191*** 0.032 0.071** 0.028
year2012 −0.277*** 0.031 – –
year2013 −0.013 0.029 0.222*** 0.027
year2014 – – 0.230*** 0.026
0.011** 0.006 −0.015*** 0.005

* Significant at 10%, ** Significant at 5% and *** Significant at 1%.

Table 8
Maximum likelihood estimates of other models (cont.)

Variable Cobb-Douglas HHI Translog Gini index

Coefficient Std.Error Coefficient Std.Error

Inefficient effects model

µ
Concentration −2.874*** 0.631 −7.04*** 1.752
u
Concentration −5.325*** 0.786 −4.035*** 0.789
v
Const −4.941*** 0.17 −4.977*** 0.183
Variance parameters
E ( u) 0.499 0.343
v 0.085*** 0.007 0.083*** 0.008

* Significant at 10%, ** Significant at 5% and *** Significant at 1%.

ANNEX 2

Table 9
Maximum likelihood estimates of the model 2

Variable Coefficient Std.Error z P-Value

Stochastic distance frontier model

ln xL 0.573 0.084 6.82 0
ln x C 0.108 0.091 1.18 0.239
ln xIC 0.319 0.071 4.48 0
lnqWLU −0.188 0.091 −2.07 0.038
lnqCOM −0.278 0.1 −2.78 0.005
ln xLlnxL 0.035 0.141 0.25 0.803
ln x C lnx C −0.194 0.143 −1.36 0.174
(continued on next page)

64
S. Hidalgo-Gallego and I. Mateo-Mantecón Journal of Air Transport Management 76 (2019) 56–66

Table 9 (continued)

Variable Coefficient Std.Error z P-Value

ln xIC lnxIC −0.223 0.172 −1.3 0.195

lnqWLU lnqWLU 0.043 0.042 1.03 0.448
lnqCOM lnqCOM −0.012 0.017 −0.76 0.591
ln xLlnx C −0.032 0.121 −0.26 0.792
ln xL ln xIC −0.003 0.123 −0.03 0.977
ln x C ln xIC 0.226 0.116 1.94 0.052
ln xL ln lnqWLU −0.051 0.084 −0.6 0.548
ln xL lnqCOM 0.072 0.086 0.83 0.405
ln x C ln lnqWLU −0.111 0.055 −2.02 0.044
ln x C lnqCOM 0.062 0.056 1.11 0.269
ln xIC ln lnqWLU 0.161 0.07 2.31 0.021
ln xIC lnqCOM −0.134 0.071 −1.88 0.061
lnqWLU lnqCOM −0.037 0.038 −0.98 0.325
trend −0.11 0.034 −3.24 0.001
trend² 0.028 0.01 2.85 0.004

* Significant at 10%, ** Significant at 5% and *** Significant at 1%.

Table 10
Maximum likelihood estimates of the mode 2 (cont.)

Variable Coefficient Std.Error z P-Value

Inefficient effects model

µ
Herfindahl −4.464 1.271 −3.51 0.000
Efplan −1.36 0.601 −2.27 0.0023
u
Herfindahl −2.718 1.156 −2.35 0.019
Efplan −1.136 0.485 −2.34 0.019
v
Const −4.557 0.361 −12.62 0.000
Variance parameters
E ( u) 0.535
v 0.101 0.012 8.12 0.000

* Significant at 10%, ** Significant at 5% and *** Significant at 1%.

References Organ. 51 (1), 43–73.

Chang, V., Tovar, B., 2014. Efficiency and productivity changes for Peruvian and Chilean
ports terminals: a parametric distance functions approach. Transport Pol. 31, 83–94.
Abrate, G., Erbetta, F., 2010. Efficiency and patterns of service mix in airport companies: CNC, 2011. III Informe Anual Sobre Ayudas Pú blicas en España. Comisión Nacional de la
an input distance function approach. Transport. Res. E Logist. Transport. Rev. 46 (5), Competencia, Madrid, pp. 46–102.
693–708. CNMC, 2014. El sector aeroportuario en España: situación actual y recomendaciones de
Aena, 2012. Guía de tarifas. Avaliable in. http://www.aena.es/csee/ccurl/609/847/ liberalización. Avaliable in: http://www2.congreso.gob.pe/sicr/cendocbib/con4_
guia-tarifas-aena-aeropuertos-2012.pdf. uibd.nsf/DD5FCB6426B003AB05257DEF005A8A72/$FILE/Estudio_Aeroportuario_
Aena, 2014. Guía de tarifas. Avaliable in. http://www.aena.es/csee/ccurl/163/444/Guia España.pdf.
%20tarifas%20aena%20aeropuertos%202014_ed%20octubre.pdf. Coelli, T.J., Rao, D.S.P., O'Donnell, C.J., Battese, G.E., 2005. An Introduction to Efficiency
Adler, N., Liebert, V., 2014. Joint impact of competition, ownership form and economic and Productivity Analysis. Springer Science & Business Media.
regulation on airport performance and pricing. Transport. Res. Pol. Pract. 64, Costa, A., García, J., López, X., Raymong, J.Ll, 2015. Estimación de las paridades de poder
92–109. adquisitivo para las comunidades autónomas españolas. Generalitat de Catalunya,
Barros, C.P., 2008. Technical change and productivity growth in airports: a case study. Departament d’Economia y Coneixement, Monografies (17). Avaliable in. http://
Transport. Res. Pol. Pract. 42 (5), 818–832. www.reunionesdeestudiosregionales.org/Reus2015/htdocs/pdf/p1402.pdf.
Barros, C.P., Dieke, P.U., 2008. Measuring the economic efficiency of airports: a Simar– D'Alfonso, T., Daraio, C., Nastasi, A., 2015. Competition and efficiency in the Italian
Wilson methodology analysis. Transport. Res. E Logist. Transport. Rev. 44 (6), airport system: new insights from a conditional nonparametric frontier analysis.
1039–1051. Transport. Res. E Logist. Transport. Rev. 80, 20–38.
Basso, L.J., 2008. Airport deregulation: effects on pricing and capacity. Int. J. Ind. Organ. Dargay, J., Hanly, M., 2001. May). The determinants of the demand for international air
26 (4), 1015–1031. travel to and from the UK. In: 9th World Conference on Transport Research,
Battese, G.E., Coelli, T.J., 1995. A model for technical inefficiency effects in a stochastic Edinburgh, Scotland.
frontier production function for panel data. Empir. Econ. 20 (2), 325–332. Evans, W.N., Kessides, I.N., 1993. Localized market power in the US airline industry. Rev.
Belotti, F., Daidone, S., Ilardi, G., Atella, V., 2012. Stochastic Frontier Analysis Using Econ. Stat. 66–75.
Stata. Fu, X., Homsombat, W., Oum, T.H., 2011. Airport–airline vertical relationships, their
Borenstein, S., 1990. Airline mergers, airport dominance, and market power. Am. Econ. effects and regulatory policy implications. J. Air Transport. Manag. 17 (6), 347–353.
Rev. 80 (2), 400–404. Gaggero, A.A., Piga, C.A., 2011. Airline market power and intertemporal price dispersion.
Borenstein, S., Rose, N.L., 1994. Competition and price dispersion in the US airline in- J. Ind. Econ. 59 (4), 552–577.
dustry. J. Polit. Econ. 102 (4), 653–683. Gerardi, K.S., Shapiro, A.H., 2009. Does competition reduce price dispersion? New evi-
Bottasso, A., Conti, M., 2012. The cost structure of the UK airport industry. J. Transport dence from the airline industry. J. Polit. Econ. 117 (1), 1–37.
Econ. Pol. 46 (3), 313–332. Giaume, S., Guillou, S., 2004. Price discrimination and concentration in European airline
Brueckner, J.K., Lee, D., Singer, E.S., 2013. Airline competition and domestic US airfares: markets. J. Air Transport. Manag. 10, 305–310 2004.
a comprehensive reappraisal. Economics of Transportation 2 (1), 1–17. Graham, D.R., Kaplan, D.P., Sibley, D.S., 1983. Efficiency and competition in the airline
Cachon, G.P., Lariviere, M.A., 2005. Supply chain coordination with revenue-sharing industry. Bell J. Econ. 118–138.
contracts: strengths and limitations. Manag. Sci. 51 (1), 30–44. Greene, W., 2005. Fixed and random effects in stochastic frontier models. J. Prod. Anal.
Cao, K.H., Krier, B., Liu, C.M., McNamara, B., Sharpe, J., 2017. The nonlinear effects of 23 (1), 7–32.
market structure on service quality: evidence from the US airline industry. Rev. Ind. Greenfield, D., 2014. Competition and service quality: new evidence from the airline

65
S. Hidalgo-Gallego and I. Mateo-Mantecón Journal of Air Transport Management 76 (2019) 56–66

industry. Economics of Transportation 3 (1), 80–89. Pacheco, R.R., Fernandes, E., 2003. Managerial efficiency of Brazilian airports. Transport.
Ha, H.K., Wan, Y., Yoshida, Y., Zhang, A., 2013. Airline market structure and airport Res. Pol. Pract. 37 (8), 667–680.
efficiency: evidence from major Northeast Asian airports. J. Air Transport. Manag. Pavlyuk, D., 2009. Spatial competition pressure as a factor of European airports' effi-
33, 32–42. ciency. Transport and Telecommunication 10 (4), 8–17.
Hidalgo-Gallego, S., Martínez-San Román, V., Núñez-Sánchez, R., 2017. Estimation of Pavlyuk, D., 2010. Multi-tier spatial stochastic frontier model for competition and co-
allocative efficiency in airports for a pre-privatization period. In: The Economics of operation of European airports. Transport and Telecommunication 11 (3), 57–66.
Airport Operations. Emerald Publishing Limited, pp. 69–95. Salazar de la Cruz, F., 2013. In: Círculo, Rojo (Ed.), Industria Aeroportuaria, (Seville,
Jorgenson, D.W., Griliches, Z., 1967. The explanation of productivity change. Rev. Econ. Spain).
Stud. 34 (3), 249–283. Sarkis, J., 2000. An analysis of the operational efficiency of major airports in the United
Jondrow, J., Lovell, C.K., Materov, I.S., Schmidt, P., 1982. On the estimation of technical States. J. Oper. Manag. 18 (3), 335–351.
inefficiency in the stochastic frontier production function model. J. Econom. 19 Scotti, D., Malighetti, P., Martini, G., Volta, N., 2012. The impact of airport competition
(2–3), 233–238. on technical efficiency: a stochastic frontier analysis applied to Italian airport. J. Air
Mazzeo, M.J., 2003. Competition and service quality in the US airline industry. Rev. Ind. Transport. Manag. 22, 9–15.
Organ. 22 (4), 275–296. Sivrikaya, O., Tunç, E., 2013. Demand forecasting for domestic air transportation in
Merkert, R., Mangia, L., 2014. Efficiency of Italian and Norwegian airports: a matter of Turkey. Open Transport. J. 7 (1).
management or of the level of competition in remote regions? Transport. Res. Pol. Stavins, J., 2001. Price discrimination in the airline market: the effect of market con-
Pract. 62, 30–38. centration. Rev. Econ. Stat. 83 (1), 200–202.
Merkert, R., Assaf, A.G., 2015. Using DEA models to jointly estimate service quality Tovar, B., Martin-Cejas, R.R., 2009. Are outsourcing and non-aeronautical revenues im-
perception and profitability–Evidence from international airports. Transport. Res. portant drivers in the efficiency of Spanish airports? J. Air Transport. Manag. 15 (5),
Pol. Pract. 75, 42–50. 217–220.
Njegovan, N., 2006. Elasticities of demand for leisure air travel: a system modelling ap- U.S. Department of Justice and the Federal Trade Commission, 2010. Horizontal mergers
proach. J. Air Transport. Manag. 12 (1), 33–39. guideline. Available in: https://www.ftc.gov/sites/default/files/attachments/
Núñez-Sánchez, R., 2015. Regional public support to airlines and airports: an unsolved merger-review/100819hmg.pdf.
puzzle. Transport. Res. E Logist. Transport. Rev. 76, 93–107. Wang, H.J., 2002. Heteroscedasticity and non-monotonic efficiency effects of a stochastic
OECD, 2001. Measuring Capital OECD Manual: Measurement of Capital Stocks, frontier model. J. Prod. Anal. 18 (3), 241–253.
Consumption of Fixed Capital and Capital Services. Organization for Economic Wang, H.J., Schmidt, P., 2002. One-step and two-step estimation of the effects of exo-
Cooperation and Development, Paris. genous variables on technical efficiency levels. J. Prod. Anal. 18 (2), 129–144.
Oum, T.H., Yan, J., Yu, C., 2008. Ownership forms matter for airport efficiency: a sto-
chastic frontier investigation of worldwide airports. J. Urban Econ. 64 (2), 422–435.

66