You are on page 1of 13

Transport Policy 48 (2016) 92–104

Contents lists available at ScienceDirect

Transport Policy
journal homepage: www.elsevier.com/locate/tranpol

An evaluation of the operational efficiency of turkish airports using


data envelopment analysis and the Malmquist productivity index:
2009–2014 case
H. Hasan Örkcü a,n, Cemal Balıkçı b, Mustafa Isa Dogan a, Aşır Genç b
a
Gazi University, Faculty of Sciences, Department of Statistics, 06500 Teknikokullar, Ankara, Turkey
b
Selcuk University, Graduate School Of Natural and Applıed Scıences Department of Statistics, Konya, Turkey

art ic l e i nf o a b s t r a c t

Article history: Turkey’s airport industry has experienced substantial growth over the recent years, but few studies have
Received 15 June 2015 analysed their operational efficiency. This paper uses Malmquist productivity index (classical and
Received in revised form bootstrapping) to assess the operational performance of 21 Turkey airports during the period of 2009
2 February 2016
through 2014. The findings indicated that the efficiency and productivity of the majority of the Turkish
Accepted 22 February 2016
airports increased during the period under investigation. However, in the period of 2011–2012, a sig-
Available online 4 March 2016
nificant decline was observed in efficiency. The main reason of this stagnation is the significant increase
Keywords: in the physical capacity of the Turkish airports in 2011. The non-reflection of the increasing physical
Turkish airports capacity to passenger and cargo traffic caused a decline in 2012. In spite of declining in the period of
Operational airport efficiency
2011–2012, efficiency values of Turkish airports have increased again since 2013. Moreover, decom-
Data envelopment analysis
position of the Malmquist index showed that most Turkey airports experienced losses in efficiency;
Bootstrapping Malmquist index
however, in terms of technology, they have progressed. Two significant factors (i.e. operating hours and
percentage of international traffic) were identified by the Simar-Wilson double bootstrapping regression
analysis as explaining variations in airport efficiency.
& 2016 Published by Elsevier Ltd.

1. Introduction legal entity, autonomy over its activities, liability limited with its
capital, is associated with the Ministry of Transportation, and its
Airports play an important role in countries' economic devel- services are accepted as privilege with the latest legal regulation.
opment. They provide a base for one of the most critical modes of Air transport required with civil aviation activities, management of
transportation around the world. The common purpose of all the airports, performing ground services at airports, and air traffic
airports is to provide an air navigation and airport operating ser- control services, establishment and operation of air navigation
vices at international standards in the aviation sector, leaning on systems and facilities and other related facilities and systems, and
high quality, safe, human-and environment-sensitive high tech- to maintain them at the level of modern aeronautics are listed
nology infrastructure and systems and qualified labor force. Per- among the missions of this entity.
formance evaluation of the airports and producing performance In the last decade, commercial flights have raised up to 196%,
improving policies for the inefficient airports will make important and number of passengers has leapt up 280% while cargo amount
contribution to national economies. has marked 126% increase. Hence, the air passenger level in total
Turkey has 53 operational airports nationwide, the manage- has realized as 131 million and flight traffic as 1.1 million (DHMI
ment of the Turkey airports and the mission of regulation and annual report, 2014).
control of the Turkish airspace are performed by the General Di- Today the airplane traffic has reached to 1.7 million, and Turkey
rectorate of State Airports Authority (Devlet Hava Meydanları is now one of the largest participants of the European air trans-
Işletmesi-DHMI). DHMI is a state economic enterprise, which has portation network with the daily added 342 flights mainly by
domestic and the Middle East flights in its airspace, which places
n
Turkey among the countries with the highest rate of daily added
Corresponding author.
traffic to the European air transportation network.
E-mail addresses: hhorkcu@gazi.edu.tr (H.H. Örkcü),
cemalbalikci@yahoo.com (C. Balıkçı), midogan@gazi.edu.tr (M.I. Dogan), As a result of the positive momentum experienced in the last
agenc@selcuk.edu.tr (A. Genç). decade, in terms of the number of air passengers, Turkey became

http://dx.doi.org/10.1016/j.tranpol.2016.02.008
0967-070X/& 2016 Published by Elsevier Ltd.
H.H. Örkcü et al. / Transport Policy 48 (2016) 92–104 93

the 5th in Europe and 11th in the world, and in terms of con- Sarkis (2000) utilized DEA for evaluating the performance of air-
nection flight, Turkey became the first in Europe in the last 12 ports; however, the analysis focused on three factors that may
years. Istanbul Ataturk Airport became the 4th largest airport in influence operational performance of airports. He analysed how
Europe in the same time period in terms of the number passengers efficiency evolved at 44 North American airports between 1990
(DHMI web site). Turkey will be the most rapidly growing country and 1994. Martin and Roman (2001) used DEA to evaluate the
in terms of air traffic with a yearly increase rate of 7.2% and the performance of 37 Spanish airports in 1997. Gillen and Lall (2001)
largest contributor to the European air traffic network till 2020 recommended the use of DEA for evaluation of performance be-
(DHMI annual report, 2013). cause of its non-parametric requirements and consideration of a
The purpose of this study was to evaluate the operating effi- number of factors simultaneously. Even though they introduced
ciencies and the productivity changes of Turkish airports, to help DEA as a possible technique for evaluation, they did not address
policy-makers, such as airport authorities, develop transparent many advantageous characteristics of various DEA-based models.
airport management strategies, to improve long-term airport in- They also recommended, in their extensions, that additional fac-
vestment plans, and to identify benchmark airports and their best- tors in evaluations could prove beneficial. In their evaluations,
in-class managerial practices that others can emulate. We used they did not incorporate financial measures. Pels et al. (2001)
both the Data Envelopment Analysis (DEA) BCC model the compared European airports using DEA. Fernandes and Pacheco
Malmquist productivity index (classical and bootstrapping), and (2002) studied 35 Brazilian domestic airports using the basic DEA
the regression analysis (Simar-Wilson bootstrapping approach) to model to evaluate capacity efficiency and passenger utilization
evaluate comparative operating efficiencies, productivity changes with operational resources. Pacheco and Fernandes (2003) used
of the Turkish airports over time, and important factors on the DEA to measure the distance from the efficient frontier and to
efficiency to achieve the aforementioned objectives. enable the avenues for management improvement. As an exten-
DEA is a non-parametric mathematical tool for assessing the sion of the work by Sarkis (2000),and Sarkis and Talluri (2004)
relative efficiency of decision-making units (DMUs). In this study, used DEA and clustering analysis to perform benchmarking of
an airport is viewed as a DMU and its operating efficiency com- airports. Barros and Dieke (2007) empirically addressed financial
ponents will be considered. In addition to the DEA model, the and operational performance of Italian airports using DEA meth-
Malmquist productivity index was used to evaluate the change in odology. Lam et al. (2009) analysed various dimensions of opera-
airport efficiency over time. The Malmquist productivity index tional efficiencies in major Asia Pacific airports through Data En-
evaluates a productivity change of a DMU between two time velopment Analysis (DEA) models that simultaneously account for
periods as the product of “catch-up” and “frontier shift” terms. The external macroeconomics and price factors.
catch-up (or recovery) term reflects the degree that a DMU attains In relation to the measurement of airport productivity changes,
for improving its efficiency, while the frontier shift (or innovation) many studies have utilized the MPI to evaluate airport efficiency.
term reflects the change in the efficient frontier surrounding the Murillo-Melchor (1999) analysed the scale efficiency and techno-
DMU between the two time periods (Tone et al., 2004). logical changes in 33 Spanish airports for the years between 1992
The background of the BCC DEA model, the MPI, bootstrapping and 1994 by implementing an input-oriented DEA, complemented
MPI, and the regression analysis (Simar-Wilson bootstrapping with a Malmquist index. Results of a Malmquist index showed that
approach) is presented first, followed by the three methodologies the total productivity decreased from 1992 to 1994. Gillen and Lall
used. Next, the data set of sampled airports for the BCC model and (2001) explored the Malmquist index of productivity changes in
the MPI as well as the key variables for the Simar-Wilson boot- the terminal services and movements of 22 US airports during the
strapping regression analysis are described. The empirical results period 1989–1993. Abbott and Wu (2002) used the Malmquist
are then reported and a summary of the key findings of this study index to examine productivity changes in 12 major Australian
is presented. airports for the period 1989–2000. In their analysis, regression
The lack of a detailed study about the Turkish airports in the analyses were conducted to explore the impacts of airport char-
literature and the need to close this gap increases the importance acteristics and financial variables on varying efficiency changes,
of this study. There is no detailed information all airports in Turkey technological changes, and total factor productivity growth. Fung
(newly formed airports such as Siirt, Usak, Kastamonu, Agri, and et al. (2008) evaluated the efficiency and productivity changes in
Konya), so this study focuses on major 21 airports of Turkey. regional airports in China during the period 1995–2004. By the
DEA and related models have become a popular method of decomposition of the Malmquist index, they identified the major
investigating airport efficiency. The performance of airports has source of productivity growth to be technical progress, rather than
received significant attention in the literature. The DEA-CCR and an improvement in efficiency. Barros and Weber (2009) used DEA
BCC methodology was developed to measure an airport's effi- to estimate the Malmquist input-based index of total factor pro-
ciency relative to other airports using constant return to scale ductivity for 27 UK airports operating during 2000/01 to 2004/05.
(CRS) or variable return to scale (VRS), respectively, incorporating Productivity change was factored into an index of efficiency
multiple airport input versus output variables. Doganis and Gra- change and an index of technological change. Throughout the
ham (1987) presented the first study devoted totally to indicators period, UK airports experienced average decreases in productivity.
of airport performance. These indicators serve to monitor airport Martin et al. (2009) implemented a parametric approach to esti-
performance over time, as well as among a group of airports, and mate the efficiency and the marginal costs of 37 Spanish airports
thus to determine their strong and weak points, and to direct between 1991 and 1997. The findings show that larger airports are
administrators' attention accordingly. Doganis et al. (1995) made a on average more efficient than smaller counterparts, possibly be-
performance study of 25 European airports. Hooper and Hensher cause of the pressure to cross-subsidize the smaller, non-profitable
(1997) use the total factor productivity (TFP) measures, with in- airports within the system. Chow and Fung (2012) measured the
dices, to complete an analysis of airport performance. Gillen and productivity changes of 30 airports in China during the period of
Lall (1997) applied DEA to a series of North American airports. 2000–2006 by computing their Malmquist productivity indices
Parker (1999) used DEA to study technical efficiency of the British using parametric output distance functions. Tsui et al. (2014a)
Airport Authority before and after privatization, and found that examined the efficiency and productivity changes of New Zeal-
privatization had no noticeable impact on technical efficiency. and's major airports between 2010 and 2012, using slacks-based
94 H.H. Örkcü et al. / Transport Policy 48 (2016) 92–104

measure (SBM) model and the Malmquist productivity index decreasing/output-increasing potentials. Gitto and Mancuso
(MPI). The findings suggested that the majority of New Zealand (2012) used the bootstrapping data envelopment analysis to assess
airports increased efficiency and productivity during the period the operational performance of 28 Italian airports during the
under investigation, but should decrease scale of operations in period of 2000 through 2006. This study found that the Italian
order to operate at their most productive size. Ahn and Min (2014) airport industry experienced a significant technological regress,
utilized both the DEA and the Malmquist productivity index to with few airports achieving an increase in productivity led by
evaluate the comparative performances of international airports improvements in efficiency. Tsui et al. (2014a) also utilized the
across the world during the multiple time periods and then slack-based measure (SBM) model, the Malmquist Productivity
measure all of their technical, scale, and mix efficiencies. Index (MPI), and the Simar-Wilson bootstrapping methods to in-
A two-stage approach, combining the first-stage DEA model vestigate the efficiency and productivity changes of 11 New Zeal-
and the second-stage Tobit regression analysis, has become a and airports for the period of 2010–2012.
popular method for determining the factors that affect airport Barros (2008b) analysed the rate of technical progress at Por-
efficiency and productivity (Barros and Sampaio, 2004; Chi-Lok tuguese airports between 1990 and 2000, using a stochastic cost
and Zhang, 2009). However, it has been reported that in using this frontier method. A translog frontier model was used and the
two-stage approach, the efficiency scores generated by the first- maximum likelihood estimation technique was employed to esti-
stage DEA model are dependent on each other (Xue and Harker, mate the empirical model. The results showed that the average
1999), as a result, the second-stage Tobit estimation results may be score is 88%, denoting a relatively high degree of waste, despite the
misleading. To solve this problem, Simar and Wilson (2000, 2007) fact that technical change contributed to a reduction in costs. To-
introduced the bootstrapping approach to improve the techniques var and Martin-Cejas (2009) used an input oriented stochastic
that are adopted in the literature of airport efficiency and pro- translog distance function to 26 Spanish airports for the years
ductivity studies. Tsui et al. (2014b) evaluated the operational ef- between 1993 and 1999, followed by a second stage regression in
ficiency of 21 Asia-Pacific airports between 2002 and 2011. A two- order to examine the effects of outsourcing and commercial ac-
stage method was used: Data Envelopment Analysis (DEA) to as- tivities on airport efficiency. The main result is that efficiency is
sess airport efficiency, followed by the second-stage regression positively influenced by outsourcing level and share of non-aero-
analysis to identify the key determinants of airport efficiency. The nautical revenues. Martin et al. (2009) used Markov chain Monte
second-stage regression analysis indicated that percentage of in- Carlo simulation to measure an SFA model for the efficiency of
ternational passengers handled by an airport, airport hinterland Spanish airports. Yang (2010) applied DEA and SFA to estimate the
population size, and dominant airline of an airport when entering efficiencies of 12 international airports in the Asia-Pacific region
global airline strategic alliance, and an increase in GDP per capita from 1998 to 2006. Tovar and Martin-Cejas (2010) used a translog
are significant in explaining variations in airport efficiency. input distance function and decomposed changes in productivity
Several studies have used the bootstrapping approach for es- into efficiency and technical changes for the years between 1993
timating efficiency determinants of airports. Barros and Dieke and 1999 for a sample of 26 Spanish airports. The results indicated
(2008) used the truncated bootstrapped regression to estimate the an increase in overall productivity, which was driven by a smooth
efficiency and identify the determinants of 31 Italian airports be- positive technical change. Lozano et al. (2013) presented a Direc-
tween 2001 and 2003. They found that the method to bootstrap tional Distance Function (DDF) network DEA approach that takes
the DEA efficiency scores with a truncated regression analysis can into account undesirable outputs, applying it to assess the effi-
better explain DEA efficiency levels. Similarly, Barros (2008a) ciency of 33 Spanish airports for year 2008. The finding revealed
employed the truncated bootstrapping regression analysis to that the network DEA approach has more discriminatory power
analyse the efficiency of Argentinian airports during the period of than its single-process counterpart, uncovering more inefficiencies
intense economic crisis. Curi et al. (2011) used a bootstrapped DEA and providing more valid results.
procedure to estimate technical efficiency of 18 Italian airports With regard to Turkey airports, the literature on efficiency has
during the period 2000–2004. During the same year, Tsekeris been limited to the application of DEA methodology. Kıyıldı and ve
(2011) used the truncated bootstrapped regression to assess the Karaşahin (2006) utilized an input-oriented CCR DEA with a focus
relative technical efficiency of Greek airports and investigate fac- on the influence of infrastructure at 32 small airports in Turkey for
tors that determine airport efficiency. Assaf (2011) used the the years between 1996 and 2002. They found seven out of 32
Malmquist bootstrapped methodology to assess the extent of airports operating on the efficient frontier. Ulutas and Ulutas
productivity, efficiency, scale and technological changes at the (2009) applied a CCR DEA to data from 31 Turkey airports in 2004
major Australian airports. Results showed that most Australian and 2005. Peker and Baki (2009) used an input oriented DEA and
airports have experienced significant total factor productivity in- also compare the results of CCR and BCC models for 37 Turkey
crease between 2002 and 2007, while few airports have recorded airports in 2007. In a separate analysis, they conducted a t-test to
productivity and efficiency decline over the same period. Merkert investigate the efficiency differences between large and small
and Mangia (2012) also applied the bootstrapping two-stage DEA airports and find that large airports are more efficient than small
model to analyse 46 Norwegian airports' efficiency. Merkert et al. ones. Kirankabes and Arikan (2011) applied CCR and BCC models
(2012) employed the input-oriented DEA model and the Simar- to data for 36 Turkey airports in 2009. Their results showed that
Wilson bootstrapping approach to analyse the efficiency of re- most of the airports are technically efficient but suffer from scale
gional airports worldwide, and suggested that the more sophisti- inefficiencies. Kocak (2011) applied both CCR and BCC models to a
cated two-stage model can deliver powerful insights into the sample of 40 Turkey airports in 2008. Ar (2012) investigated the
performance of regional airports. Wanke (2012) showed the use of dynamic changes over time by constructing a Malmquist Index.
different approaches for assessing efficiency related-issues in 63 The analysis included 31 Turkey airports for the years between
major Brazilian airports. Starting out with the bootstrapping 2007 and 2011, and the average total factor productivity grew by
technique presented in Simar and Wilson (2000), several DEA 13% during the five years, which was mainly driven by the tech-
estimates were generated, allowing the use of confidence intervals nical efficiency change. Voltes-Dorta and Pagliari (2012) and
and bias correction in central estimates to test for significant dif- Martin et al. (2013) included only the eight international airports
ferences in efficiency levels, returns-to-scale, and input- in Turkey in their respective analysis of cost efficiencies of
H.H. Örkcü et al. / Transport Policy 48 (2016) 92–104 95

international airports. Ülkü (2015) utilized a data envelopment the scalar vector associated with each DMU. A DMU is considered
analysis (DEA) to compare the relative efficiency of airports within as a BCC-efficient DMU when θ is equivalent to 1 and has zero
AENA (for Spain) and DHMI (for Turkey) for the years between output and input slacks. Otherwise, the DMU is called a BCC-in-
2009 and 2011. The results indicated higher average efficiency efficient DMU.
levels at Spanish airports, but private involvement enhances effi-
ciency at Turkey airports. The majority of the airports in Spain and
Turkey operate under increasing returns to scale. 2.2. The Malmquist productivity index and smoothing bootstrapping
Malmquist approach

Färe et al. (1994) developed a DEA-based Malmquist Pro-


2. Methodology
ductivity Index (MPI) that measures total productivity changes
over time, which Caves et al. (1982) showed to capture the
2.1. Data envelopment analysis and the BCC model
changes in an airport's total productivity between two adjacent
periods as shown in Eq. (2):
DEA is a linear programming-based approach for measuring
the relative efficiency of decision making units (DMUs). It was first ⎡ Dt xt + 1, yt + 1
) ⎤⎥
1/2

M xt , yt , xt + 1, yt + 1 = ⎢
( ) (
Dt + 1 xt + 1, yt + 1
proposed by Charnes et al. (1978) as the CCR model. In general ( ) ⎢⎣ Dt xt , yt
×
⎥⎦
terms, the efficiency of a particular unit can be defined as a ratio of ( ) (
Dt + 1 x t , y t) (2)
the value of outputs to the value of inputs, where maximum ef-
ficiencies are restricted to 1; thus, the efficiency of a unit must be where y represents the output vector that can be produced using
less than or equal to 1. Similar to the CCR model, the BCC model is the input vector x and Dt (xt , yt ) is defined as the output distance
also one of the most popular models. The CCR model developed by function and M measures the total productivity changes between
Charnes et al. (1978) is based on Constant Returns to Scale (CRS), period t and period tþ1 with reference to the frontier technology
whereas the BCC model by Banker et al. (1984) is based on Vari- at period t It is common to define M as the geometric mean over
able Returns to Scale (VRS). These DEA models, having the input- the study period. It reflects progress or regress in the efficiency of
oriented and output-oriented models, can identify benchmarking the DMU along with progress or regress of the frontier technology.
entities and suggest the degree of improvement for inefficient The total productivity improves if M41, remains unchanged if
DMUs to be efficient. An input-oriented model aims to minimize M¼1, and declines if M o1.
the level of inputs and maintain the current level of outputs, The MPI can be further decomposed into two components: the
whereas the output-oriented model aims to maximize the level of efficiency change component (TECI), which measures the change
outputs given the current level of inputs. Wober (2007) indicated in the relative efficiency of an airport between period t and period
that the output-oriented model required a given level of inputs to tþ1; and the technological change component (TCI), which cap-
achieve the maximum output levels. In this study, output oriented tures the shift in the frontier technology between two periods
BCC model is considered and this means that airports focus on (Fung et al., 2008). The MPI can also be expressed as
maximizing three categories of air traffic outputs (i.e. air passenger
Malmquist productivity
numbers, air cargo volumes, and aircraft movements), holding all
of the airport inputs constant. index = TECI (Catching − up) *TCI (Frontier − shift ).
It is assumed that there are n DMUs to be evaluated in terms of
The TECI is related to the degree of efforts that the DMU at-
m inputs and s outputs. Let xij (i = 1,2, … , m) and yrj (r = 1,2, … , s )
tained for improving its efficiency. It indicates whether an airport
represent the input and output values of DMUj (j = 1,2, … , n),
has moved closer to, or further from, the frontier technology over
respectively.
the study period. The TCI reflects a change in the efficient frontiers
The BCC model has been developed by Banker et al. (1984) as
surrounding the DMUs between the two time periods and mea-
an extension of the CCR model to allow for returns-to-scale to be
sures the change in an airport's total productivity, which is mainly
variable. Thus, BCC model computes efficiency scores corre-
due to improvements in airport operating technology.
sponding to the assumption of variable returns-to-scale (VRS). It is
However, relation (2) does not allow us to determine whether
a more flexible than the CCR model since it allows for constant,
changes in productivity, efficiency or technology are real or merely
increasing, and decreasing returns-to-scale. Output-oriented BCC
artifacts of the fact that we do not know the true production
model is as follows:
frontiers and must estimate them from a finite sample (Simar and
⎛ s m ⎞ Wilson, 1999). Thus, we employ a consistent bootstrap estimation
Maxθ = ϕ + ε ⎜⎜ ∑ sr+ + ∑ si−⎟⎟
⎝ r=1 ⎠ procedure for correcting and obtaining confidence interval for the
i=1
Malmquist index and its components TECI and TCI. The idea un-
s. t. derlying the bootstrap is to approximate the sampling distribution
n
∑ λj xij + sio− = xio of MPI the unknown true values of MPI. We traced a modified
j=1 smooth bootstrap procedure proposed by Simar and Wilson (1999)
n similar to Gitto and Mancuso (2012). The smooth bootstrap pro-
∑ λj yrj +
− sro = ϕyro cedure can be found detailed in Gitto and Mancuso (2012). In the
j=1
n
smooth bootstrap procedure, efficiency change and technological
∑ λj = 1 j = 1, 2, …, n change are similar to Malmquist index. With the obtained con-
j=1 fidence interval for the Malmquist index and its components, it is
possible to determine whether productivity improvement (or de-
λ j , si−, sr+ ≥ 0, i = 1, 2, …; m, r = 1, 2, …, s (1)
cline) is significant at the established confidence level. The smooth
+
where, θ is efficiency index, ε is a small positive number, and sio−
sro bootstrap procedure for the productivity was implemented using
are output and input slacks, respectively, λj is the dual variable or the FEAR package (Wilson, 2008).
96 H.H. Örkcü et al. / Transport Policy 48 (2016) 92–104

2.3. Simar-Wilson bootstrapping regression analysis Table 2


Summary statistics about Turkish airports.
For the second stage analysis, the bootstrapping approach to
Outputs Inputs
the second-stage regression analysis is applied, which suggested
by Simar and Wilson (2000, 2007). The bootstrapping approach is y1 y2 y3 x1 x2 x3
a computer-based method for assigning measures of accuracy to
2009 Mean 32973.8 3653418.0 19896.0 1.4 4534.1 18059.0
statistical estimates (Simar and Wilson, 2000). However, Simar
St.Dev. 61629.9 6917287.9 76889.8 0.6 2146.4 29233.7
and Wilson (2007) criticized the potential bias in efficiency esti- Max 283953.0 29812888.0 371391.0 3.0 9700.0 125634.0
mates given the strong correlation between the resulting effi- Min 1232.0 16461.0 20.0 1.0 2640.0 607.0
ciency scores, that is, calculation of the efficiency of one firm in- 2010 Mean 36802.0 4215509.8 24382.7 1.4 4534.1 31609.9
St.Dev. 65416.6 7872483.8 95717.0 0.7 2160.4 74837.5
corporates observation of all other firms in the same data set.
Max 288246.0 32143819.0 452146.0 3.0 9700.0 345270.0
Hence, direct regression analysis is invalid owing to the de- Min 1281.0 33411.0 1.0 1.0 2640.0 1090.0
pendency of the efficiency scores. In this study, following Simar 2011 Mean 41613.3 4810361.5 26327.5 1.4 4534.1 31609.9
and Wilson (2007), the double bootstrap approach is used to St.Dev. 73345.7 9080578.3 105482.8 0.7 2160.4 74837.5
Max 325209.0 37394694.0 498049.0 3.0 9700.0 345270.0
overcome this problem. Simar and Wilson (2007) detail the this
Min 2017.0 21559.0 1.9 1.0 2640.0 1090.0
algorithm, also described in a step-by-step approach in Barros and 2012 Mean 43229.2 5298079.0 28321.7 1.4 4534.1 56033.8
Assaf (2009) and Barros and Garcia-del-Barrio (2011). We refer the St.Dev. 80696.6 10468124.0 115174.2 0.7 2160.4 75154.5
interested reader to these studies for details. Max 364322.0 45091962.0 543506.9 3.0 9700.0 345270.0
Min 2052.0 26257.0 1.7 1.0 2640.0 3500.0
During the second stage regression analysis, the efficiency
2013 Mean 47825.1 5952898.8 33219.4 1.4 4534.1 58998.4
scores obtained from the first-stage BCC model to be regressed on St.Dev. 89277.9 11768375.7 133574.0 0.7 2160.4 84456.1
the specific operating characteristics and ownerships related to Max 406317.0 51297790.0 630679.0 3.0 9700.0 345270.0
Turkish airports, identifying the significant factors to explain var- Min 2403.0 66929.0 2.1 1.0 2640.0 3500.0
2014 Mean 47825.14 5952898.81 33219.4 1.43 4534.09 58998.43
iations in airport efficiency.
St.Dev. 91492.62 12064816.19 137013.31 0.68 2251.12 87202.15
Max 406317 51297790 630679 3.0 9700 345270.0
Min 2403 66929 2.1 1.0 2640 3500.0
3. Data and selection of variables

There are 53 active airports in Turkey. It has not reached to the


desired input-output variables for the newly formed airports, such variables (Boussofiane et al., 1991; Tsui et al., 2014a).
as Siirt, Usak, Kastamonu, Agri, and Konya. For this reason, this Based on the literature review and data availability, we obtain
study focuses on 21 major airports of Turkey (Table 1). Data were three output variables and three input variables. The inputs con-
collected via the annual reports of the airports and DHMI website. sidered in this study are: x1: number of runway, x2: dimension of
Annual data were collected for the period of 2009–2014 (with a runway units (m), and x3: passenger terminal area (m2). The
total of 126 observations). Airport managers were contacted to outputs included in the present study were y1: annual number of
obtain additional information for this study. flights (planes), y2: annual passenger throughputs (persons), and
A general DEA convention was followed to decide the total y3: annual cargo throughputs (tons), as indicated by Doganis
number of airport observations, ensuring that satisfactory dis- (1992) that an airport's primary function is to provide an interface
criminating power was met; that is, the number of airports ob- between aircraft and its passengers or freight.
served should be at least twice the sum of the number of input and A summary of the descriptive statistics related to the airport
output variables, or the number of airport observations be equal or input and output variables for 21 major Turkey airports is pre-
larger than the product of the number of airport input and output sented in Table 2 for the period of 2009–2014.
The statistics suggested that there are large variations among
the sampled Turkey airports in terms of their input and output
levels. In terms of airport input variables, the passenger terminal
Table 1 area was found to have a large degree of variation. Moreover, there
List of the selected airports.
was no change in the number of runways among the sampled
The Name of the Airports Turkey airports because the construction of new runways often
Adana requires large investments and long-term planning. However, the
Izmir Adnan Menderes passenger terminal area has increased 2–3 times in 2012 for most
Antalya airports. 2011 was investment year for DHMI. In this year, DHMI
Istanbul Atatürk
Bursa Yenisehir
marked some organizational improvements such as MSSR sys-
Diyarbakır tems, PSR systems, ILS systems, DVOR systems, AVIAN Radar sys-
Elazig tems, CAT III operation procedure, and RNAV routes procedures.
Erzurum Most airports were renewed and new terminal buildings were
Ankara Esenboga
completed. Masterplans were implemented in most airports to
Van Ferit Melen
Gaziantep increase the usage ratio of capacity. The SMART Project were
Kayseri planned for safety and capacity to meet the necessary demand in
Malatya this year (DHMI annual report, 2011).
Mugla Milas-Bodrum
In terms of airport output variables, the annual number of
Mugla Dalaman
Nevsehir Kapodakya flights, annual passenger throughputs, and annual cargo throughputs
Isparta Süleyman Demirel increased over the study period, the mean number of flights and
Samsun Carsamba passenger and cargo movements increased by nearly 45%, 63%, and
Sivas Nuri Demirag 67%, respectively, during the study period.
Tekirdag Corlu
Trabzon
An excessive number of inputs and/or outputs with respect to
the number of observations causes in a large number of efficient
H.H. Örkcü et al. / Transport Policy 48 (2016) 92–104 97

Table 3 4. Empirical results


Factors, inputs, outputs and inertia.
The output-oriented BCC model and the MPI (classical and
Factors Original variables Inertia
Outputs bootstrapping approaches) were applied to investigate the op-
erational efficiency and the total productivity changes of 21 major
o1 number of flights (y1) and annual passenger throughputs (y2) 0.926 Turkey airports between 2009 and 2014. The results are as follows.
o2 annual cargo throughputs (y3)
Inputs
i1 number of runway (x1) and dimension of runway units (x2) 0.935
4.1. BCC results
i2 passenger terminal area (x3)
Table 4 presents the DEA efficiency scores of 21 airports during
the multi-year period of 2009 through 2014 using the output-or-
iented BCC model. An examination of Table 4 revealed that the BCC
efficiency scores of half of the Turkish airports are relatively high
units (Simar and Wilson, 2008). Therefore, we first analyse the
during the study period.
relationship among inputs (outputs), and then we reduce the Adana, Antalya, and Istanbul Atatürk airports were the most
number of variables by employing the methodology proposed by efficient airports for the six-year period. On the other hand, Nev-
Daraio and Simar (2007). Scatter plots among output and input sehir Kapodakya, Bursa Yenisehir, Sivas Nuri Demirag, and Erzur-
variables with the correlations, respectively, are given in Appen- um airports were the least efficient airports during the entire
dix. There is a clear linear dependence among variables and this period. As mentioned above, there are large differences in size and
dependence allows us to reduce the number of variables by ag- operating capacities between Turkey airports. Despite the poor
gregating them in factors with minimum loss of information. The performance of Nevsehir Kapodakya and Bursa Yenisehir airports,
factor input (output) is obtained as the weighted sum of the ori- their efficiency values have increased during the period. Moreover,
ginal variables with weights represented by the values of the first Isparta Süleyman Demirel, Nevsehir Kapodakya and Bursa Yeni-
eigenvalue of the input (output) matrix. Mathematically, a factor, sehir airports, except for the decrease in 2012 for Isparta Süleyman
Demirel and slightly decrease in 2011 for Nevsehir Kapodakya,
F, is given by: F ¼Xa. Where, X is the matrix of the input (output)
outstand with the ever-increasing level of efficiency. Also, this
variables and a is the first eigen vector of the matrix XX0 . The
result also clearly appears from these findings Malmquist index.
capacity of a factor to summarize the information contained in the
For some airports, such as Izmir Adnan Menderes, Diyarbakır,
original variables is expressed by the inertia, which is computed Ankara Esenboga, Van Ferit Melen, and Tekirdag Corlu, passenger
by dividing the first eigen value by the sum of all eigen values of and cargo traffics did not increase in parallel to the increasing
the matrix XX0 . A value close to 1 indicates an accurate re- physical capacity, which was reflected as a decline in our study,
presentation. The output and input factors and their relative in- although these values remained same in reality. As a matter of fact
ertias are shown in Table 3. that, Turkish Air Lines was the top airline in Europe in 2012.
The percentage of inertia explained by the two factors is about
93%. Therefore, it is appropriate to summarize the information of 4.2. Malmquist productivity index results
the full data matrix by the two factors.
All analyses were performed with the new output and input In addition to the BCC model, the Malmquist productivity index
factors (variables). (classical and bootstrapping approaches) was used to evaluate the
change in airport efficiency over time. It is important to evaluate
changes in the total productivity of Turkish airports to understand
whether the total productivity of individual airports is improving
Table 4 or deteriorating over time. This section summarizes the MPI re-
The efficiency scores of the selected airports during the period of 2009–2014. sults obtained from the analysis of the multi-year airport perfor-
mance data. The Malmquist productivity index can be decomposed
Airport 2009 2010 2011 2012 2013 2014
into two components: technical change and technical efficiency
Adana 1.000 1.000 1.000 1.000 1.000 1.000 change. These components were separately calculated and then
Izmir Adnan Menderes 0.615 1.000 1.000 0.525 0.555 0.566 analysed. The classical and bootstrapping TFPI, TECI, and TCI scores
Antalya 1.000 1.000 1.000 0.859 0.962 0.929
for Turkish airports during 2009–2014 are presented in Table 5.
Istanbul Atatürk 1.000 1.000 1.000 1.000 1.000 1.000
Bursa Yenisehir 0.095 0.145 0.152 0.169 0.169 0.167 The TFPI, TECI, and TCI scores greater than 1 indicated total pro-
Diyarbakır 0.839 0.993 1.000 0.563 0.681 0.625 ductivity growth gains efficiency and technical progress, respec-
Elazig 1.000 1.000 1.000 0.195 0.203 0.193 tively (and vice versa). The classical and bootstrapping MPI, TECI,
Erzurum 0.255 0.296 0.251 0.213 0.209 0.202 and TCI scores were calculated using the software of “DEAP Ver-
Ankara Esenboga 0.501 0.971 0.975 0.425 0.452 0.395
Van Ferit Melen 1.000 1.000 1.000 1.000 1.000 1.000
sion 2.1” and the FEAR package (Wilson, 2008), respectively.
Gaziantep 0.339 0.361 0.435 0.375 0.428 0.464 The following sub-sections summarized the classical and
Kayseri 0.515 0.322 0.391 0.392 0.371 0.434 bootstrapping TFPI, as well as the patterns of TECI and TCI during
Malatya 0.902 0.425 0.402 0.245 0.239 0.207 the six-year span and then interpreted the results of index figures.
Mugla Milas-Bodrum 1.000 1.000 1.000 0.901 0.851 0.865
Mugla Dalaman 1.000 1.000 1.000 1.000 0.941 1.000
Nevsehir Kapodakya 0.141 0.145 0.153 1.000 1.000 1.000 4.2.1. The outline of the airport productivity change
Isparta Süleyman Demirel 0.202 0.696 0.901 0.455 0.788 1.000 For the bootstrapping MPI, the geometric mean was 1.039,
Samsun Carsamba 0.363 0.436 0.449 0.345 0.338 0.345 which means that, overall, the 21 Turkey airports' total pro-
Sivas Nuri Demirag 0.355 0.078 0.085 0.069 0.079 0.091 ductivity in 2014 increased by 4% over the period of the study.
Tekirdag Corlu 1.000 1.000 1.000 1.000 1.000 0.987
During 2014, fourteen (more than half) airports experienced
Trabzon 1.000 1.000 1.000 1.000 1.000 1.000
Mean 0.672 0.708 0.724 0.606 0.632 0.641 higher total productivity levels than in 2009. For the international
airports, Bursa Yenisehir, İstanbul Atatürk, Izmir Adnan Menderes,
98 H.H. Örkcü et al. / Transport Policy 48 (2016) 92–104

Table 5
Airport average productivity changes during 2009–2014.

Airport Productivity change TFPI 2009–2014 Efficiency change TECI 2009–2014 Technological change TCI 2009–2014

MPI MPI bootstrap TECI TECI bootstrap TCI TCI bootsrap

Adana 1.089 1.038 1.000 0.912 1.089 1.107


Izmir Adnan Menderes 1.186 1.217 1.135 1.125 1.045 1.041
Antalya 1.086 1.077 1.041 1.040 1.043 1.029
Istanbul Atatürk 1.291 1.343 1.000 1.000 1.291 1.323
Bursa Yenisehir 1.168 1.196 1.241 1.248 0.941 0.934
Diyarbakır 0.992 1.071 0.962 0.969 1.031 1.072
Elazig 0.797 0.815 0.802 0.801 0.994 0.978
Erzurum 0.908 0.910 0.927 0.925 0.979 0.971
Ankara Esenboga 1.213 1.227 1.112 1.109 1.091 1.084
Van Ferit Melen 0.834 0.857 0.859 0.845 0.971 0.989
Gaziantep 1.059 1.065 1.029 1.021 1.029 1.021
Kayseri 0.883 0.900 0.891 0.877 0.991 0.987
Malatya 0.670 0.702 0.714 0.705 0.939 0.945
Mugla Milas Bodrum 1.139 1.145 1.054 1.044 1.081 1.080
Mugla Dalaman 1.110 1.131 1.058 1.051 1.049 1.066
Nevsehir Kapodakya 1.048 1.052 1.116 1.111 0.939 0.922
Isparta Süleyman Demirel 1.203 1.254 1.321 1.315 0.911 0.919
Samsun Carsamba 1.009 1.084 0.979 0.971 1.031 1.086
Sivas Nuri Demirag 0.826 0.798 0.793 0.767 1.042 1.023
Tekirdag Corlu 0.989 0.999 0.991 0.990 0.998 0.989
Trabzon 1.190 1.242 1.101 1.105 1.081 1.087
Mean 1.020 1.039 0.996 0.986 1.024 1.028

Ankara Esenboga, Mugla-Milas Bodrum and Trabzon showed a and Turkey was also affected by this crisis. Despite this, the
great improvement in its total productivity during the study per- Turkish Airlines was the top company among European countries
iod, whereas the performance of Erzurum decreased by 9%. Of the regarding passenger and freight traffic numbers (DHMI, annual
provincial airports, Isparta Süleyman Demirel was subject to the report 2012).
largest increase in total productivity (almost 25%) during the Results of the productivity change, efficiency change, and
period of the study. On the other hand, Elazig, Van Ferit Melen, technological change as presented in Table 5 explain that the
Kayseri, Malatya and Sivas Nuri Demirag experienced declines in productivity of airports is influenced by a technological change
total productivity by at most 18.5%, 14.3%, 10.0%, 29.8%, and 21.20%, more heavily than a technical efficiency change. In 2009–2014, the
respectively between 2009 and 2014. Turkish Airlines found itself in a phase of transition between a
Fig. 1 shows that for both classical and bootstrapping ap- highly regulated civil aviation and free competition. They were,
proaches the Technological Change Index (TCI), Technical Effi- therefore, forced to implement radical programs for institutional
ciency Change Index (TECI), and the Total Factor Productivity Index restructuring. Airport efficiency seems to be affected by airports'
(TFPI) declined between 2011 and 2012, whereas they peaked in own ability to utilize their resources more significantly than the
2011. As displayed in Fig. 1, TECI, TCI, and TFPI exhibit similar external business environments. By beginning from 2003, gov-
patterns for the six-year span. In this Figure, B indicates the ernment has taken new steps to improve infrastructure of the
bootstrapping approach. airports, development of fleet and quality of the service. In this
Similar to the investment situation in 2011, the figure shows context, airports were equipped with modern equipment, new and
that, in the period of 2011–2012, there was a significant decline in modern airports were put into service and quality of the service
airports efficiency. The Boeing World Air Cargo Forecast (WACF) has improved. As a result of this improvements, there have been
2014/2015 also reported this decline. The forecast summarizes the significant increases in the number of flights, passengers and
world's major air trade markets, identifies major trends, and pre- amount of cargo.
sents forecasts for the future performance and development of
markets as well as for the world freighter airplane fleet. According 4.2.2. Total Factor Productivity Index (TFPI)
to the report, after rebounding more than 19% in 2010 over the Table 6 shows the changes of TFPI over time for both classical
depressed levels of 2009, the world air cargo traffic stagnated from and bootstrapping approaches, which increased from 2009 to
mid-2011 to early 2013. 2011. For bootstrapping approach, the geometric mean is 1.298
This prolonged period of weak growth can be attributed to two which means that, the 21 Turkish airports' total productivity in
factors: a weak world economy and slack trade growth. Since the 2011 significantly increased (by 30%). All airports included in this
onset of global economic crisis in 2008, the world air cargo traffic study showed improvement in its total productivity until 2011.
has averaged only 1.7% growth per year through 2013. On a posi- Especially, except Tekirdag Corlu airport, TFPI values of all airports
tive note, the world air cargo traffic began to grow again in the increased at least 10% in 2011. By increasing its capacity, the
second quarter of 2013. By July 2014, traffic grew 4.4% compared Turkish Airlines has continued to rise with a growing momentum
with the first seven months of 2013. Forecasts for even better within the last 10 years. It reached a significant volume pro-
economic and trade growth should lead to sustained air cargo spectively in 2011 (DHMI, annual report 2011). DHMI has executed
traffic growth in 2015 and 2016. airports and terminals modernization projects that it enables to
In the period 2011–2012, there was a global crisis in the world manage government and civilian enterprises company together.
H.H. Örkcü et al. / Transport Policy 48 (2016) 92–104 99

1.40 Technical change is often triggered by external factors, such as


shifts in government policies, advances in technology, and changes
1.20
in economic environments. It was observed in the present study
1.00 that the productivity of the airports was influenced by technical
change, as discussed previously. This result coincides with the fact
0.80
that Turkey made important contributions to the aviation sector in
0.60 the last ten years and made significant investment and new pro-
jects in this area.
0.40
2009-2010 2010-2011 2011-2012 2012-2013 2013-2014
4.2.3. Technical Efficiency Change Index (TECI)
MPI MPI (B) As mentioned in Section 2, the MPI can be decomposed into
two components: the technical efficiency change (TECI) and the
(a)
technological change (TCI). Therefore, the growth in total pro-
1.20 ductivity of Turkey airports could be the result of either a gain in
1.10 efficiency, or the progress in technological efficiency, or both.
The technical efficiency change (TECI) represents the change in
1.00 efficiency for the individual Turkey airports and for all of the
Turkey airports between 2009 and 2014. Often, the change in an
0.90
airport’s efficiency scores could be attributed to proper execution
0.80 of the technical experience accumulated over the years, better
management of airport activity, and sound investment planning
0.70 (Barros et al., 2010). For bootstrapping approach, the geometric
2009-2010 2010-2011 2011-2012 2012-2013 2013-2014
mean of the TECI scores was 0.985, which means that the average
TECI TECI (B) efficiency level of Turkey airports decrease very slightly (1.5%) over
(b) the study period. Ten airports increased their efficiency: Izmir
1.250 Adnan Menderes, Antalya, Bursa Yenisehir, Ankara Esenboga, Ga-
ziantep, Mugla-Milas Bodrum, Mugla Dalaman, Nevsehir Kapoda-
1.150 kya, Isparta Süleyman Demirel, and Trabzon. Of these, Isparta
Süleyman Demirel experienced the largest improvement in its
1.050 efficiency, increasing by around 32% between 2009 and 2014. Is-
0.950 tanbul Atatürk neither increased nor decreased its efficiency
(TECI¼1), signifying that they were located on the efficiency
0.850 frontier. Finally, Adana, Diyarbakır, Elazig, Erzurum, Van Ferit
Melen, Kayseri, Malatya, Samsun Carsamba, Sivas Nuri Demirag,
0.750 and Tekirdag Corlu showed declines in efficiency over the study
0.650
period, ranging from 3% to 30%. The TECI results means that the
2009-2010 2010-2011 2011-2012 2012-2013 2013-2014 majority of Turkey airports experienced efficiency gains above the
average.
TCI TCI (B) Table 7 summarizes the changes in TECI over time. Similar to
(c) the TFPI results, significant increases occurred in 2011. Most of the
airports showed improvements in their TECI in this year (espe-
Fig. 1. The productivity trend of the airports during the period of 2009–2014.
cially Bursa Yenisehir, Isparta Süleyman Demirel, and Sivas Nuri
(a) Classical MPI and bootstrapping MPI (b) Classical TECI and bootstrapping TECI
(c) Classical TCI and bootstrapping TCI. Demirag airports). After the capacity expansion in 2011, a recovery
occurred and the index value rose above 1 again in 2013 and 2014
(TECI values were 1.109 and 1.017 for 2012–2013 and 2013–2014,
respectively). According to Table 7; Istanbul Atatürk airports' TECI
This strategy gave efficiency to DHMI management. Total 25 scores are equal or greater than one over the period, which means
terminals buildings constructed and modernized that 17 of them that it did not change their productivity and Bursa Yenisehir and
from main sources of DHMI and 8 of them sharing corporation and Nevsehir Kapodakya airports are in good state in terms of TECI.
27 airports runway, taxi road and aprons had renewed or re- The efficiency change levels of the most of the airports are all
constructed until 2011 year. After the capacity expansion in the above the average during the entire period.
passenger terminal area (about 3 times for most airports) in 2011,
despite the increase in the amount of passengers, flights and cargo 4.2.4. Technological Change Index (TCI)
compared to the previous years, the airports' performances Technological change (TCI) is regarded as the consequences of
showed a downward momentum. However, due to the inability to R&D and innovation (Barros et al., 2010; Fung et al., 2008). It
meet the new increased capacity, their performance was perceived captures the effect of a shift in the productivity change frontier of
to have fallen. Because of the increase in both passenger and the individual Turkey airports, which helps clarify the effect of
freight traffic, Izmir Adnan Menderes, Antalya, and Istanbul Ata- technical change on productivity change using production func-
türk airports were slightly affected from the crisis and continued tions (Tsui et al., 2014b). From Table 5, for bootstrapping approach,
to its efficiency, whereas the other 18 airports suffered from a the geometric mean of the TCI scores was 1.028; this indicated that
decreased productivity. However, in this six-year period, despite a on average, Turkey airports experienced technical progress
temporary decline, the general recovery in activity and efficiency (around 3%). Five international airports (Bursa Yenisehir, Erzurum,
of Turkish airports is noteworthy. Kayseri, Malatya, and Nevsehir Kapodakya) showed technical re-
100 H.H. Örkcü et al. / Transport Policy 48 (2016) 92–104

Table 6
Changes of TFPI over time.

Airport 2009–2010 2010–2011 2011–2012 2012–2013 2013–2014

MPI MPI bootstrap MPI MPI bootstrap MPI MPI bootstrap MPI MPI bootstrap MPI MPI bootstrap

Adana 1.135 1.117 1.191 1.214 0.815 0.784 1.119 1.105 1.129 1.169
Izmir Adnan Menderes 1.191 1.249 1.242 1.241 1.232 1.192 1.093 1.106 1.121 1.159
Antalya 0.895 0.888 1.221 1.269 1.072 1.058 1.079 1.039 1.090 1.111
Istanbul Atatürk 1.284 1.346 1.176 1.393 1.231 1.252 1.159 1.167 1.212 1.240
Bursa Yenisehir 1.737 1.736 1.455 1.436 0.775 0.728 1.037 1.043 1.328 1.343
Diyarbakır 1.367 1.408 1.418 1.429 0.364 0.348 1.544 1.525 1.036 1.060
Elazig 0.814 0.816 1.259 1.258 0.264 0.249 1.252 1.262 1.057 1.064
Erzurum 1.339 1.363 1.096 1.105 0.474 0.462 1.104 1.130 1.153 1.151
Ankara Esenboga 1.663 1.641 1.184 1.170 0.968 0.981 1.333 1.341 1.012 0.992
Van Ferit Melen 1.208 1.251 1.188 1.220 0.281 0.279 0.784 0.792 1.067 1.056
Gaziantep 1.265 1.298 1.363 1.363 0.519 0.471 1.268 1.260 1.142 1.204
Kayseri 0.779 0.826 1.191 1.186 0.470 0.474 1.379 1.380 1.066 1.133
Malatya 0.505 0.500 1.141 1.136 0.285 0.263 1.032 1.048 0.943 0.988
Mugla Milas Bodrum 2.008 2.080 1.166 1.196 0.469 0.419 1.469 1.506 1.147 1.128
Mugla Dalaman 1.266 1.308 1.339 1.388 0.764 0.777 1.070 1.060 1.101 1.201
Nevsehir Kapodakya 1.054 1.091 1.173 1.222 0.606 0.589 1.179 1.184 1.267 1.238
Isparta Süleyman Demirel 2.081 2.029 1.620 1.665 0.289 0.319 2.113 2.093 1.348 1.389
Samsun Carsamba 1.273 1.288 1.214 1.240 0.525 0.524 1.067 1.077 1.152 1.143
Sivas Nuri Demirag 0.131 0.170 1.905 1.935 0.644 0.602 1.416 1.452 1.295 1.282
Tekirdag Corlu 1.198 1.262 1.059 1.068 0.430 0.475 1.415 1.415 1.029 1.053
Trabzon 1.403 1.435 1.349 1.396 0.848 0.838 1.151 1.130 1.092 1.156
Mean 1.091 1.125 1.271 1.298 0.568 0.556 1.217 1.219 1.128 1.151

Table 7
Changes of TECI over time.

Airport 2009–2010 2010–2011 2011–2012 2012–2013 2013–2014

TECI TECI bootstrap TECI TECI bootstrap TECI TECI bootstrap TECI TECI bootstrap TECI TECI bootstrap

Adana 1 0.987 1 0.992 1 0.974 1 1 1 1.009


Izmir Adnan Menderes 1.409 1.422 0.988 0.981 1.215 1.185 1.091 1.096 0.989 0.986
Antalya 1 0.991 1 1.003 1.082 1.055 1.089 1.083 0.984 0.989
Istanbul Atatürk 1 1.023 1 1.143 1 1 1 1.005 1 1.023
Bursa Yenisehir 1.477 1.456 1.255 1.235 1.444 1.422 0.959 0.951 1.231 1.213
Diyarbakır 1.195 1.234 1.029 1.034 0.569 0.571 1.332 1.322 0.951 0.941
Elazig 0.814 0.811 1.051 1.023 0.396 0.389 1.119 1.122 0.972 0.966
Erzurum 1.145 1.141 0.924 0.921 0.888 0.891 0.964 0.988 1.055 1.051
Ankara Esenboga 1.892 1.877 0.994 0.967 0.829 0.841 1.199 1.212 0.929 0.909
Van Ferit Melen 1.099 1.091 1 1.003 0.537 0.522 0.685 0.681 0.987 0.981
Gaziantep 1.101 1.116 1.186 1.191 0.691 0.634 1.173 1.179 1.045 1.055
Kayseri 0.695 0.699 1.169 1.161 0.641 0.622 1.266 1.269 0.971 0.979
Malatya 0.441 0.440 0.991 0.989 0.597 0.546 0.951 0.958 0.875 0.858
Mugla Milas Bodrum 1.863 1.861 1.019 1.011 0.488 0.423 1.333 1.387 1.061 1.034
Mugla Dalaman 1.699 1.705 1.134 1.139 0.695 0.701 0.939 0.922 0.979 0.978
Nevsehir Kapodakya 0.895 0.888 1.001 1.011 1.294 1.255 1.083 1.099 1.194 1.145
Isparta Süleyman Demirel 1.925 1.921 1.422 1.456 0.609 0.601 1.949 1.911 1.233 1.266
Samsun Carsamba 1.081 1.067 1.041 1.055 0.807 0.823 0.934 0.931 1.058 1.034
Sivas Nuri Demirag 0.117 0.132 1.641 1.621 0.855 0.824 1.312 1.334 1.187 1.123
Tekirdag Corlu 1 1.000 1 0.987 0.829 0.809 1.292 1.291 0.929 0.909
Trabzon 1.416 1.405 1.129 1.145 0.899 0.871 1.055 1.034 0.985 0.991
Mean 1.022 1.028 1.083 1.088 0.785 0.764 1.107 1.109 1.025 1.017

gress of between 1% and 8%, while the average technical efficiency technically efficient airports (twelve airports) showed above
of four provincial airports (Elazig, Van Ferit Melen, Isparta Süley- average technical progress.
man Demirel, and Tekirdag Corlu) decreased between 1% and 9%, Table 8 shows changes in both TCI over time, which increased
over the study period. On the other hand, the provincial airports by 2.8% from 2009 to 2014. Similar to TFPI and TECI results, sig-
showed technical progress: Diyarbakır (7.2%) and Sivas Nuri De- nificant increases occurred in 2011. All of the airports showed
mirag (2.3%). This implied that infrastructure investment and im- improvements in TCI values in this year (especially Izmir Adnan
provement of the management techniques (such as AMAN/DMAN) Menderes, Antalya, and Diyarbakır airports). After the capacity
may have a significant impact on their operations through in- expansion in 2011, a recovery occurred and the index value rose
vestment into new airport operating techniques and production above 1 again in 2014 (TCI values are 1.092 and 1.102 for 2012-
technology methods. TCI results suggested that the majority of 2013 and 2013-2014, respectively). As clearly indicated, the total
H.H. Örkcü et al. / Transport Policy 48 (2016) 92–104 101

Table 8
Changes in TCI over time.

Airport 2009–2010 2010–2011 2011–2012 2012–2013 2013–2014

TCI TCI bootstrap TCI TCI bootstrap TCI TCI bootstrap TCI TCI bootstrap TCI TCI bootstrap

Adana 1.135 1.124 1.191 1.178 0.815 0.802 1.119 1.103 1.129 1.145
Izmir Adnan Menderes 0.845 0.849 1.257 1.251 1.004 1.006 1.002 1.003 1.133 1.139
Antalya 0.895 0.891 1.221 1.229 0.991 0.993 0.991 0.956 1.108 1.102
Istanbul Atatürk 1.284 1.301 1.176 1.198 1.231 1.234 1.159 1.159 1.212 1.211
Bursa Yenisehir 1.176 1.171 1.159 1.151 0.537 0.511 1.081 1.094 1.079 1.085
Diyarbakır 1.144 1.140 1.378 1.345 0.639 0.601 1.159 1.151 1.089 1.081
Elazig 1 0.987 1.198 1.205 0.666 0.631 1.119 1.122 1.087 1.087
Erzurum 1.169 1.161 1.186 1.191 0.534 0.512 1.145 1.135 1.093 1.077
Ankara Esenboga 0.879 0.865 1.191 1.189 1.168 1.165 1.112 1.106 1.089 1.044
Van Ferit Melen 1.099 1.123 1.188 1.189 0.524 0.529 1.145 1.149 1.081 1.033
Gaziantep 1.149 1.157 1.149 1.132 0.751 0.721 1.081 1.056 1.093 1.102
Kayseri 1.121 1.119 1.019 1.008 0.733 0.731 1.089 1.081 1.098 1.123
Malatya 1.144 1.132 1.151 1.145 0.477 0.471 1.085 1.081 1.078 1.099
Mugla Milas-Bodrum 1.078 1.098 1.144 1.153 0.961 0.981 1.102 1.077 1.081 1.056
Mugla Dalaman 0.745 0.741 1.181 1.189 1.099 1.091 1.139 1.141 1.125 1.178
Nevsehir Kapodakya 1.178 1.177 1.172 1.179 0.468 0.467 1.089 1.066 1.061 1.051
Isparta Süleyman Demirel 1.081 1.056 1.139 1.122 0.474 0.499 1.084 1.088 1.093 1.091
Samsun Carsamba 1.178 1.184 1.166 1.134 0.651 0.634 1.142 1.144 1.089 1.101
Sivas Nuri Demirag 1.116 1.145 1.161 1.168 0.753 0.711 1.079 1.078 1.091 1.115
Tekirdag Corlu 1.198 1.222 1.059 1.051 0.519 0.587 1.095 1.087 1.108 1.111
Trabzon 0.991 0.989 1.195 1.181 0.943 0.947 1.091 1.085 1.109 1.119
Mean 1.067 1.068 1.173 1.169 0.723 0.717 1.100 1.092 1.101 1.102

Table 9 (2007). After obtaining the DEA BCC efficiency scores in the
Estimation results for Simar-Wilson double bootstrapping regression analysis. first-stage, we calculated the regression coefficients through the
bootstrapped procedure in the second stage (with BCC efficiency
Dependent variable: BCC efficiency scores
scores bounded at both ends of the 0–1 distribution). The esti-
Explanatory Variables Coefficients t-value mated results for the Simar-Wilson double bootstrapping re-
gression analysis are provided in Table 9. Taking the literature
**
Constant 0.488 9.098 and data availability into account, five explanatory variables
Population around the airport 0.038 0.945
Airport hub status 0.011 1.615
were developed for the second-stage regression analysis, which
Airport operating hours 0.135** 4.318 represents Turkey airports' operating characteristics, manage-
Joint military-civil airport 0.017 0.241 ment/ownerships, and regional locations (see Table 9). Since a
Percentage of international traffic  0.033*  2.519 higher score reflects a higher efficiency level for a DMU, a po-
R-Square:0.41
sitive coefficient for the independent variables from the second
Number of observations: 126
stage regression indicates a positive effect of the corresponding
Remarks: Total number of iterations ¼2000. variable on the level of efficiency. Data related to the selected
*
Indicate that the explanatory variable is significant at the 0.05 significance explanatory variables was obtained from DHMI annual reports,
level. Turkish Statistical Institute, Eurostat and airports' websites. In-
**
Indicate that the explanatory variable is significant at the 0.01 significance
dividual airports were also contacted to obtain additional
level.
information.
Used explanatory variables are as follows: population around
the airport, airport hub status (the airport hub status dummy
TCI was observed to be positive during the entire period and the takes the value of 1 when an airport is an international airport,
technical change levels of the majority of the airports were all and 0 otherwise), airport operating hours, joint military-civil
above the average during the entire period. airport (this dummy variable takes the value of 1 when an air-
Overall, the breakdown of total productivity changes of Turkey port is an open to military operations, and 0 otherwise), and
airports into the change in efficiency gain (TECI) and technical percentage of international traffic. The coefficients of the ex-
progress (TCI) showed mixed results. More than half of 21 major planatory variables were tested. Two explanatory variables were
found to be the significant factors that explain variations in
Turkey airports showed total productivity growth (MPI 41) and
airport efficiency among the sampled Turkish airports, including
efficiency gains (TECI o1) over the study period, and the technical
the operating hours and percentage of international traffic. The
progress (TCI 41) were experienced by at least half of the Turkey
sign of the explanatory variables' coefficients indicated the di-
airports in the group.
rection of the effect.
For the variable of ‘airport operating hours’, the positive coef-
4.3. The Simar-Wilson bootstrapping regression analysis results ficient (i.e. the positive impact) implied that longer airport oper-
ating hours could trigger higher efficiency for Turkey airports;
To evaluate the determinants of efficiency of Turkish airports, every hour that increased from an airport's daily operating hours
we adopted the double bootstrap approach of Simar and Wilson could result in an increase of the airport's efficiency by 0.135 units,
102 H.H. Örkcü et al. / Transport Policy 48 (2016) 92–104

with more airport traffic volumes to handle. To some degree, was influenced by a technological change index (TCI) rather than a
longer airport operating hours may also allow an airport to gen- technical efficiency change index (TECI). The rationale being that a
erate more revenues from handling increasing flight movements technological change is greatly affected by external factors such as
and more air travellers passing through the airport; this will po- shifts in government policies and technological advances, while
sitively affect an airport's efficiency. the technical efficiency change is influenced by an operational
For ‘percentage of international traffic’ the coefficient was ne-
change of the airport itself.
gative; for every percentage increase in international passengers
From the BCC and MPI results, Turkey airports experienced a
handled by an airport, its efficiency reduced by 0.033 units. Im-
significant decline in the 2011–2012. According to the Boeing
portantly, this finding appears to be consistent with Pathomsiri
et al. (2006), Tsui et al. (2014b), Ülku (2015), who claimed that the World Air Cargo Forecast (WACF) 2014/2015 report, after re-
handling of international passenger traffic has a negative impact bounding more than 19% in 2010 over the depressed levels of
on an airport’s efficiency as larger airport infrastructure and fa- 2009, world air cargo traffic stagnated from mid-2011 to early
cilities (e.g. check-in counters and baggage handling areas) need to 2013. A weak world economy and slack trade growth also trig-
be built to serve international travellers comparing with domestic gered this regress. World air cargo traffic began to grow again in
passengers. This adverse effect also can be explained by the fact the second quarter of 2013. The second reason for the decline
that international traffic requires more sophisticated infra- was the increase in the capacity of the Turkey airports in 2011-
structure and raises the operational complexity and costs. 2012. Turkey airports' infrastructure was strengthened and their
The remaining variables were not statistically significant and capacity increased two- or three-fold in this period. In the
carrying an implausible sign. This might be due to having a small
meantime, due to the mismatch of the passenger and cargo
sample size, data limitation, and conceptual difficulties exist with
traffic increase with the amount of capacity increase, the per-
the construction of these variables. Although the ‘population
formance values declined. In reality, this should not be inter-
around the airport’ variable was not statistically significant, it was
reported as having the positive coefficient (0.038), which sug- preted as a decline because of the huge increase in capacity.
gested that if an airport in Turkey serves a larger hinterland po- After 2012, the passenger and cargo traffic improved rapidly and
pulation, its efficiency would be only minimally increased. It also met airports capacity. After government's decision about air-
implied that an airport that serves a larger population size is more ports modernization project, an improvement has been started
efficient than an airport that serves a smaller population size in as of 2011 and it was determined that increase in positive di-
Turkey. rection has been catched since 2013. It is understand that air-
port performance was more affected from out effects like air-
ports modernization projects, economic changing, increasing
5. Concluding remarks social welfare, which are related to changing in government
politics. (DHMI, annual report 2012).
Evaluation of the operational performance of airports is im-
Two important factors were found to account for the identified
portant for ensuring efficiency in the wider aviation industry.
variations in airport efficiency among Turkey airports: (i) longer
Recently, many airports around the world have been commercia-
airport operating hours triggered higher airport efficiency and (ii)
lized and/or privatized in order to airports are operated more like
a business. However, the impact of commercialization or privati- a higher share of international traffic has an adverse effect on the
zation on productivity is still an open question in the related lit- performance.
erature. In the present study, we evaluated the productivity evo- Three potential limitations of this study were observed: (i) only
lution, from 2009 through 2014, of the Turkey airports by applying 21 major Turkey airports (a total of 53 airports currently operating
the Malmquist index (classical and bootstrapping) and its two in Turkey) were included in the study, which may have limited the
components: efficiency change (TECI) and technological change discriminating power; (ii) three airport inputs and three airport
(TCI) to a sample of 21 airports. outputs were selected when estimating the efficiency and total
Turkey airports experienced total productivity growth, which productivity changes of Turkey airports. Unfortunately, the com-
was shown by a geometric mean MPI score of 1.039 for boot- monly used airport efficiency measurements (such as baggage
strapping approach, suggesting that the average total productivity claim belts, work load units, boarding gates, and delays) could not
growth of Turkey airports increased by 3.9% over the period
be included because data were not available for most Turkey
(shown Table 5). When we examined the detail of the MPI, for
airports.
bootstrapping approach the geometric means of the TECI and TCI
As an extension of this study, it may be meaningful to include
scores were 0.986 and 1.028. When we examined the Table 5, it is
more data (when available) that may have a significant impact
seen that efficiency change was 0.986 in contrast to technological
change which was 1.028. It means that the total productivity im- on Turkey airports' efficiency and productivity such as full time
provements of Turkey airports were largely a result of technolo- labor employment, delay flights, and work load units. Further
gical progress. Furthermore, the analysis of technological change assessment of the efficiency and the total productivity changes
indicates that most of the future policy intervention should be of Turkey airports will be necessary to confirm the findings of
directed to enhancing the technological level of the Turkey airport this study.
industry. By increasing passenger and load capacity with a good
management and planning, the current infrastructure capacity
should increase its rate of use. Technologically, there is an ad-
vancement. This status must be maintained so that MPI preserve Appendix
its increase in positive direction. Also, in terms of the Malmquist
productivity index, we found that the productivity of an airport See appendix Fig. A1.
H.H. Örkcü et al. / Transport Policy 48 (2016) 92–104 103

Output scatterplot (r correlation coefficient.)

Input scatterplot (r correlation coefficient.)


Fig. A1. xxx.

References Barros, C.P., Sampaio, A., 2004. Technical and allocative efficiency in airports. Int. J.
Transp. Econ. 31 (3), 355–377.
Barros, C.P., Dieke, P.U.C., 2007. Performance evaluation of Italian airports: a data
Abbott, M., Wu, S., 2002. Austr. Econ. Rev. 35 (3), 244–260. envelopment analysis. J. Air Transp. Manag. 13, 184–191.
Ahn, Y.-H., Min, H., 2014. Evaluating the multi-period operating efficiency of in- Barros, C.P., Dieke, P.U.C., 2008. Measuring the economic efficiency of airports: A
ternational airports using data envelopment analysis and the Malmquist pro- Simar–Wilson methodology analysis. Transp. Res. Part E 44, 1039–1051.
ductivity index. J. Air Transp. Manag. 39, 12–22. Barros, C.P., 2008a. Airports in Argentina: Technical efficiency in the context of an
Ar, I.M., 2012. Türkiye’deki Havalimanlarının Etkinliklerindeki Değişimin İn- economic crisis. Journal. Air Transp. Manag. 14, 315–319.
celenmesi: 2007-2011 Dönemi İçin Malmquist-TFV Endeksi Uygulaması. Iktis. Barros, C.P., 2008b. Technical change and productivity growth in airports: A case
ve Idari Bilim. Derg., Atatürk Üniversitesi 26 (3–4), 143–160 (in Turkish with study. Transp. Res. Part A 42, 818–832.
English Summary). Barros, C.P., Weber, W.L., 2009. Productivity growth and biased technological
Assaf, A., 2011. Bootstrapped Malmquist indices of Australian airports. Serv. Ind. J. change in UK airports. Transp. Res. Part E 45 (4), 642–653.
31 (5), 829–846. Barros, C.P., Assaf, A., 2009. Bootstrapped efficiency measures of oil blocks in An-
Banker, R.D., Charnes, A., Cooper, W.W., 1984. Models for the estimation of tech- gola. Energy Policy 37 (10), 4098–4103.
nical and scale efficiencies in data envelopment analysis. Manag. Sci. 30, Barros, C.P., Mangi, S., Yoshida, Y., 2010. Productivity growth and biased technolo-
1078–1092. gical change in Japanese airports. Transp. Policy 17 (4), 259–265.
104 H.H. Örkcü et al. / Transport Policy 48 (2016) 92–104

Barros, C.P., Garcia-del-Barrio, P., 2011. Productivity drivers and market dynamics in Murillo-Melchor, C., 1999. An analysis of technical efficiency and productive change
the Spanish first division football league. J. Product. Anal. 35 (1), 5–13. in Spanish airports using the Malmquist index. Int. J. Transp. Econ. 26 (2),
Boussofiane, A., Dyson, R.G., Thanassoulis, E., 1991. Applied data envelopment 271–292.
analysis. Eur. J. Oper. Res. 52, 1–15. Pacheco, R.R., Fernandes, E., 2003. Managerial efficiency of Brazilian airports.
Caves, D., Christensen, L., Diewert, W.E., 1982. The economic theory of index Transp. Res. Part A: Policy Pract. 37 (8), 667–680.
numbers and the measurement of input, output, and productivity. Econome- Parker, D., 1999. The performance of BAA before and after privatization. J. Transp.
trica 50 (6), 1393–1414. Econ. Policy 33 (2), 133–145.
Charnes, A., Cooper, W.W., Rhodes, E., 1978. Measuring efficiency of decision Pathomsiri, S., Haghani, A., Dresner, M., Windle, R.J., 2006. Measuring and de-
making unit. Eur. J. Oper. Res. 2, 429–444. termination of airport productivity in competitive markets. Paper presented at
Chi-Lok, A.Y., Zhang, A., 2009. Effects of competition and policy changes on Chinese the 85th Annual meeting of the Transport Research Board, Washington DC,
airport productivity: An empirical investigation. J. Air Transp. Manag. 15, United States.
166–174. Pels, E., Nijkamp, P., Rietvle, P., 2001. Relative efficiency of European airports.
Chow, C.K.W., Fung, M.K.Y., 2012. Estimating indices of airport productivity in Transp. Policy 8, 183–192.
Greater China. J. Air Transp. Manag. 24, 12–17. Peker, İ., ve Baki, B., 2009. Veri Zarflama Analizi ile Türkiye Havalimanlarında Bir
Curi, C., Gitto, S., Mancuso, P., 2011. New evidence on the efficiency of Italian air- Etkinlik Ölçümü Uygulaması. Çukurova Üniversitesi Sos. Bilim. Enstitüsü Derg.
ports: a bootstrapped DEA analysis. Socio-Econ. Plan. Sci. 45, 84–93. 18 (2), 72–88.
Daraio, C., Simar, L., 2007. Advanced Robust and Nonparametric Methods in Effi- Sarkis, J., 2000. An analysis of the operational efficiency of major airports in the
ciency Analysis. Springer, US. United States. J. Oper. Manag. 18 (3), 335–351.
Doganis, R., 1992. The Airport Business. Routledge, London. Sarkis, J., Talluri, S., 2004. Performance-based clustering for benchmarking of US
Doganis, R., Graham, A., 1987. Airport Management: The Role of Performance In- airports. Transp. Res. Part A Policy Pract. 38, 329–346.
dicators. Transport Studies Group, Polytechnic of Central London. Simar, L., Wilson, P.W., 1999. Estimating and bootstrapping Malmquist indices. Eur.
Doganis, R., Graham, A., Lobbenberg, A., 1995. The Economic Performance of Eur- J. Oper. Res. 115, 459–471.
opean Airports. 3. Cranfield University, Cranfield, UK. Simar, L., Wilson, P.W., 2000. Statistical inference in nonparametric frontier models:
DHMI website, 〈www.dhmi.gov.tr〉, (last visit: September, 2015). the state of the art. J. Product. Anal. 13, 49–78.
DHMI, İstatistik Yıllıkları (General Directorate of State Airports Authority Statistical Simar, L., Wilson, P.W., 2007. Estimation and inference in two-stage, semi-para-
Reports, 2009–2014). metric models of production processes. J. Econ.. 136, 31–64.
Färe, R., Grosskopf, S., Norris, M., Zhang, Z., 1994. Productivity growth, technical Simar, L., Wilson, P.W., 2008. Statistical inference in non-parametric frontier
progress, and efficiency change in industrialized countries. Am. Econ. Rev. 84 models: recent developments and perspectives. In: Fried, H.O., Lovell, C.A.K.,
(1), 66–83. Schmidt, S.S. (Eds.), The Measurement of Productivity Efficiency and Pro-
Fernandes, E., Pacheco, R.R., 2002. Efficient use of airport capacity. Transp. Res. Part ductivity Change. Oxford University Press, New York, pp. 421–522.
A Policy Pract. 36 (3), 225–230. Tone, K., 2004. Malmquist productivity index: efficiency change over time. In:
Fung, M.K.Y., Wan, K.K.H., Hui, Y.V., Law, J.S., 2008. Productivity changes in Chinese Cooper, W.W., Seiford, L.M., Zhu, J. (Eds.), The Handbook of Data Envelopment
airports 1995–2004. Transp. Res. Part E 44 (3), 521–542. Analysis. Kluwer Academic Publisher, Norwell, MA, pp. 203–227.
Gillen, D., Lall, A., 1997. Developing measures of airport productivity and perfor- Tovar, B., Martin-Cejas, R.R., 2009. Are outsourcing and non-aeronautical revenues
mance: an application of data envelopment analysis. Transp. Res. Part E Logist. important drivers in the efficiency of Spanish airports? J. Air Transp. Manag. 15,
Transp. Rev. 33 (4), 261–273. 217–220.
Gillen, D., Lall, A., 2001. Non-parametric measures of efficiency of US airports. Int. J. Tovar, B., Martin-Cejas, R.R., 2010. Technical efficiency and productivity changes in
Transp. Econ. 28 (3), 283–306. Spanish airports: a parametric distance functions approach. Transp. Res. Part E
Gitto, S., Mancuso, P., 2012. Bootstrapping the Malmquist indexes for Italian air- 46, 249–260.
ports. Int. J. Prod. Econ. 135, 403–411. Tsekeris, T., 2011. Greek airports: efficiency measurement and analysis of de-
Hooper, P.G., Hensher, D.A., 1997. Measuring total factor productivity of airports- an terminants. J. Air Transp. Manag. 17, 140–142.
index number approach. Transp. Res. Part E Logist. Transp. Rev. 33 (4), 249–259. Tsui, W.H.K., Gilbey, A., Balli, H.O., 2014a. Estimating airport efficiency of New
Kıyıldı, R.K., ve Karaşahin, M., 2006. Türkiye’deki Hava Alanlarının Veri Zarflama Zealand airports. J. Air Transp. Manag. 35, 78–86.
Analizi ile Altyapı Performansının Değerlendirilmesi. Süleyman Demirel Üni- Tsui, W.H.K., Balli, H.O., Gilbey, A., Gow, H., 2014b. Operational efficiency of Asia-
versitesi Fen. Bilim. Enstitüsü Derg. 10 (3), 391–397. Pacific airports. J. Air Transp. Manag. 40, 16–24.
Kirankabes, A., Arikan, F., 2011. Devlet Hava Meydanlarinin Etkinliginin Veri Zar- Ulutaş, B. ve Ulutaş, B., 2009. An analytic network process combined data envel-
flama Analizi (VZA) ile Ölcülmesi. In: Proceedings of the 12th International opment analysis methodology to evaluate the performance of airports in
Symposium on Econometrics Statistics and Operations Research, Pamukkale Turkey. In: Proceedings of the 10th International Symposium on the Analytic
University, 39–50. (in Turkish with English Summary). Hierarchy/Network Process, July 29–August 1, University of Pittsburgh Pitts-
Kocak, H., 2011. Efficiency examination of Turkish airports with DEA approach. Int. burgh, Pennsylvania, USA.
Bus. Res. 4 (2), 204–212. Ülkü, T., 2015. A comparative efficiency analysis of Spanish and Turkish airports. J.
Lam, S.W., Low, J.M.W., Tang, L.C., 2009. Operational efficiencies of Asia Pacific Air Transp. Manag. 46, 56–68.
airports. Transp. Res. Part E 45 (4), 654–665. Voltes-Dorta, A., Pagliari, R., 2012. The impact of recession on airports' cost effi-
Lozano, S., Gutiérrez, E., Moreno, P., 2013. Network DEA approach to airports per- ciency. Transp. Policy 24, 211–222.
formance assessment considering undesirable outputs. Appl. Math. Model. 37, Wanke, P.F., 2012. Capacity shortfall and efficiency determinants in Brazilian air-
1665–1676. ports: evidence from bootstrapped DEA estimates. Socio-Econ. Plan. Sci. 46,
Martin, J.C., Roman, C., 2001. An application of DEA to measure the efficiency of 216–229.
Spanish airports prior to privatization. J. Air Transp. Manag. 7 (3), 149–157. Wilson, P.W., 2008. FEAR: a software package for frontier efficiency analysis with R.
Martín, J.C., Roman, C., Voltes-Dorta, A., 2009. A stochastic frontier analysis to es- Socio-Econ. Plan. Sci. 42 (4), 247–254.
timate the relative efficiency of Spanish airports. J. Prod. Anal. 31 (3), 163–176. Wober, K.W., 2007. Data envelopment analysis. J. Travel Tour. Market. 21 (4),
Martín, J.C., Rodríguez-Déniz, H., Voltes-Dorta, A., 2013. Determinants of airport 91–108.
cost flexibility in a context of economic recession. Transp. Res. Part E: Logist. World Air Cargo Forecast, 2014–2015, http://www.boeing.com/resources/boeing
Transp. Rev. 57, 70–84. dotcom/commercial/about-our-market/cargo-market-detail-wacf/download-re
Merkert, R., Mangia, L., 2012. Management of airports in extreme winter condi- port/assets/pdfs/wacf.pdf.
tions-some lessons from analysing the efficiency of Norwegian airports. Res. Yang, H.H., 2010. Measuring the efficiencies of Asia-Pacific international airports –
Transp. Bus. Manag. 4, 53–60. Parametric and non-parametric evidence. Comput. Ind. Eng. 59 (4), 697–702.
Merkert, R., Odeck, J., Brathen, S., Pagliari, R., 2012. A review of different bench- Xue, M., Harker, P.T., 1999. Overcoming the inherent dependency of DEA efficiency
marking methods in the context of regional airports. Transp. Rev. 32 (3), scores: a bootstrap approach. Wharton Financial Institutions Center, University
379–395. of Pennsylvania, Unpublished Working Paper.

You might also like