A Study On The Relationship Between Airport Privatisation and Airport Efficiency

A STUDY ON THE RELATIONSHIP BETWEEN
AIRPORT PRIVATISATION AND AIRPORT EFFICIENCY
AN APPLICATION OF USING AHP/DEA METHODS
A thesis submitted in fulfillment of the requirements for the degree of
DOCTOR OF PHILOSOPHY
in
CARDIFF UNIVERSITY
by
PO-LIN LAI
Logistics and Operations Management Section

Cardiff Business School
Cardiff University
April 2013
b DECLARATION AND STATEMENT a
DECLARATION
This work has not previously been accepted in substance for any degree and is not
concurrently submitted in candidature for any degree.
Signed Po-Lin Lai
Date 10 April 2013
STATEMENT 1
This thesis is being submitted in partial fulfilment of the requirements for

the degree of PhD
Signed Po-Lin Lai
Date 10 April 2013
STATEMENT 2
This thesis is the result of my own independent work/investigation, except where

otherwise stated.
Other sources are acknowledged by footnotes giving explicit references.
Signed Po-Lin Lai
Date 10 April 2013
STATEMENT 3
I hereby give consent for my thesis, if accepted, to be available for photocopying and
for inter-library loan, and for the title and summary to be made available to outside
organisations.
Signed Po-Lin Lai
Date 10 April 2013
i
b ABSTRACT a
In order to deal with the competitive environment surrounding the air transport industry,
civil aviation authorities have undertaken several approaches to improve airport
efficiency, such as investing in the infrastructure and privatising airport ownership or
governance. Among these methods, airport privatisation policy has been implemented
for around 25 years in the U.K., closely followed by other European countries. By
contrast, decision makers elsewhere, such as in the Asia-Pacific region, are now
interested in privatisation and in doing so evaluate the impact of this process elsewhere.
Therefore, the primary aim of this research is to examine the relationship between
airport privatisation and efficiency, through an Airport Efficiency Evaluation System
(AEES). The study covers Europe and the Asia-Pacific region, reflecting different
attitudes towards the role of government within airport management.
Focussing on the most popular method for assessing airport efficiency, with Data
Envelopment Analysis (DEA) a unit can appear efficient simply because of its pattern
of inputs and outputs rather than any inherent efficiency. But only using DEA may not
provide useful results about the efficiency of airports as different decision makers may
weight the relative importance of inputs and outputs differently (for example, airport
managers, and airline companies). In this research, another aim is to develop and
demonstrate the applicability of different analysis techniques within the AEES. For this
reason, Analytic Hierarchy Process (AHP) analysis is adopted to calculate the
importance of each variable. These results are then integrated into both DEA and DEA,
Assurance Region (AR) models, to reflect the different importance of the metrics. In the
context of air transportation, an integrated AHP/DEA and AHP/DEA-AR model are
applied for the first time to evaluate airport efficiency. A sensitivity analysis with
different variable sets is carried out.
In conclusion, an AEES is established and the result shows that the approach by
adopting AHP/DEA-AR model in particular can provide more accurate values of
relative efficiency than using the traditional DEA approach. There are also different
priorities between stakeholder groups and these can affect the efficiency scores of
airports. However, the results for each of the different analysis techniques show that
there is no statistically significant relationship between airport ownership and efficiency.
ii
b ACKNOWLEDGEMENTS a
First and foremost, I would like to express my greatest appreciation to my supervisor,

Dr. Andrew Potter, who has always been patient, generous and encouraging to me
throughout this long journey. I will be forever in his debt. I would like to express my
sincere gratitude to Professor Malcolm Beynon and Professor Anthony Beresford for
their invaluable comments and suggestions on my research. My special thanks also to
Professor John Doyle and Professor Kevin Cullinane for their suggestions and
comments in the viva.
It is a pleasure to thank the participants of the questionnaire survey of this research.

Without their time and patience, this thesis would not been completed. I wish to thank
Professor Chin-Shan Lu and Dr. Chung-Shan Yang for their kindly support not only on
academia but also on my life in Cardiff. Also my thanks also go to all the supporting
staff, especially: Penny, Elaine, Elsie, Lainey and Sara.
I wish to thank my friends, Poti, Champ, Wan, Seayon, Hyunmi, Jane, Virginia, Suhan,
Suneerat, Qinyun, Ko-Yang and many others for providing fuel for my brain and for
their warm friendship in Cardiff.
From the bottom of my heart, I am grateful to my parents, my elder brother, and uncle
John, whose unconditional support and love has provided me with this lifetime
opportunity. Without their spiritual support, this research would not have gone this far.
Last but not least, thanks to lovely wife Janice for her continuous support and endless
sacrifices. Without her love and understanding, I think I cannot finish this research.
iii
bTABLE OF CONTENTS a
Preface Declaration and statements i

Abstract ii
Acknowledgements iii
Table of contents iv
List of figures viii
List of tables ix
Chapter 1 INTRODUCTION 1
1.1 Introduction and motivation 2
1.2 Research questions 5
1.3 The scope and structure of the research 7
1.4 Research contribution 11
Chapter 2 ANALYSIS OF AIRPORT PERFORMANCE RESEARCH 12

2.1 Motivation 13
2.2 Methodological approaches 13
2.2.1 Partial measures 14
2.2.2 Multi-Criteria Decision Making method (MCDM) 14
2.2.3 Frontier analysis 15
2.2.3.1 Stochastic Frontier Analysis (SFA) 16
2.2.3.2 Total Factor Productivity (TFP) 16
2.2.3.3 Data Envelopment Analysis (DEA) 17
2.3 Literature analytical disciplines 18
2.3.1 Distribution by year of publication and methodology 19
2.3.2 Distribution of articles by journal 22
2.3.3 Distribution of articles by geography 24
2.3.4 Analysis of input and output variables 24
2.3.4.1 Input variables analysis 25
2.3.4.2 Output variables analysis 27
2.4 Analysis of research sample 29
2.4.1 Single country research 29
2.4.2 Cross country research 30
2.5 Summary 33
Chapter 3 DEVOLUTION OF AIRPORT OWNERSHIP AND GOVERNANCE 34

3.1 Airport ownership and governance 35
3.1.1 The period of governmental control 35
3.1.2 The period of corporatisation and commercialisation 36
3.1.3 The period of privatisation 42
3.2 Airport ownership evolution in different areas 43
3.2.1 North America: The United State of America 44
3.2.2 Europe: The United Kingdom 46
3.2.3 Asia: China 54
3.3 Summary 60
iv
Chapter 4 Research Design and Methodology 62
4.1 Research design 63
4.2 Research process approach 64
4.2.1 Research philosophy 64
4.2.2 Research approach 69
4.2.3 Research strategy 73
4.2.4 Time horizon 74
4.3 Data analysis methods 75
4.4 Analytical Hierarchy Process (AHP) 77
4.4.1 The rationale of AHP 77
4.4.2 The AHP process 78
4.4.3 Alternative scales within AHP 82
4.4.4 The features of alternative scales 85
4.5 Data Envelopment Analysis (DEA) 86
4.6 Literature of integrated AHP/DEA model 89
4.7 Data collection method 92
4.7.1 Semi-structure interview 92
4.7.2 Structured interview/survey 94
4.8 Research structure and survey design 97
4.8.1 Stage I: Variables collection 97
4.8.2 Stage II: Weights calculation 99
4.8.3 Stage III: Efficiency evaluation 100
4.9 Summary 100
Chapter 5 EMPIRICAL ANALYSIS: AHP ANALYSIS 102

5.1 Preliminary evaluation variables 103
5.2 The pilot study process 103
5.2.1 Questions in the pilot expert interview 105
5.2.2 Interview result analysis 105
5.2.2.1 Input perspective 106
5.2.2.2 Output perspective 107
5.3 Relative importance evaluation of variables 109
5.3.1 The weights analysis of Variable Set I(VS I): main criteria 110
5.3.2 The weights analysis of VS I: sub-criteria 111
5.3.3 Weights analysis of Variable Set II (VS II): main criteria 114
5.4 Overall weights of airport performance evaluation 116
5.4.1 Overall weights of VS I 116
5.4.2 Overall weights of VS II 117
5.5 Overall weights based on alternative scale 118
5.5.1 Weights of the variables in VS I 118
5.5.2 Weights of the variables in VS II 122
v
5.5.3 Alternative scales in different group: VS I 123
5.5.3.1 Main-criteria 124
5.5.3.2 Sub-criteria 125
5.5.4 Alternative scales in different group: VS II 127
5.6 Cross discussion and analysis 129
5.7 Summary 130
Chapter 6 Airport Efficiency Analysis I: Basic DEA models and 131

Integrated AHP/DEA model
6.1 Concepts of airport efficiency analysis 132
6.1.1 The DEA models that are used in this research 132
6.1.2 The orientation of the DEA model 132
6.1.3 The DEA analysis process 133
6.2 Airport efficiency analysis: Variables set I 134
6.2.1 Relations between variables 134
6.2.2 Relative efficiency analysis 135
6.2.3 The clustering analysis referral 140
6.2.4 The slack variable analysis 140
6.2.5 Hypothesis testing 146
6.3 Airport efficiency analysis: VS II 146
6.3.1 Relations between variables 147
6.3.3 The referral of clustering analysis 150
6.3.4 The slack variable analysis 151
6.4 Airport efficiency analysis: An integrated AHP/DEA model 156
6.4.1 The process description of airport efficiency evaluation 156
6.4.2 Relative efficiency analysis: VS I 161
6.4.3 Relative efficiency analysis: VS II 164
6.4.4 Relative efficiency analysis in different group: VS I 167
6.4.5 Relative efficiency analysis in different group: VS II 167
6.4.6 Hypothesis testing: VS I 171
6.4.7 Hypothesis testing: VS II 171
6.5.1 Different variables set 172
6.5.2 Different group 174
6.5.3 Alternative scales in the AHP/DEA model 176
6.6 Summary 177
Chapter 7 Airport Efficiency Analysis II: AHP/DEA-AR model 178

7.1 AHP/DEA-AR model 179
7.1.1 Concepts of DEA-AR model 179
7.1.2 Literature of DEA-AR model and AHP/DEA-AR model 180
7.1.3 AHP/DEA-AR model in this research 181
7.2 Airport efficiency analysis: VS I 182
7.2.1 Bounds calculation 182
7.2.3 Relative efficiency analysis in different group 190
vi
7.3 Airport efficiency analysis: VS II 192
7.3.1 Bounds calculation 192
7.3.3 Relative efficiency analysis in different group 199
7.5 Other thoughts of airport efficiency analysis 205
7.5.1 Straightforward AHP approach 205
7.5.2 Referral cluster approach 209
7.5.3 The comparison between models 211
7.5.4 Another hypothesis testing method 214
7.6 Summary 215
Chapter 8 RESEARCH CONCLUSIONS 217

8.1 Research conclusions 218
8.2 Research finding 219
8.3 Contribution and implication 221
8.3.1 Theoretical contributions 222
8.3.2 Methodological contributions 222
8.3.3 Managerial implications and practical contributions 222
8.4 Limitations and recommendations for further research 224
BIBLIOGRAPHY 225
APPENDICES 241
I. Research of Airport Performance Evaluation by DEA model 241
II. Analytical Hierarchy Process (AHP) questionnaire 245
III. Relative weighted input and output variables scores 250
IV. The weight distribution of variables in AHP/DEA model 255
V. Upper and lower bounds of variables weight ratios: VS I 257
VI. Upper and lower bounds of variables weight ratios: VS II 260
vii
b LIST OF FIGURES a
Figure
1.1 World annual air traffic 2

1.2 Passenger development 3
1.3 Annual traffic growth 3
1.4 Structure of the research 10
2.1 Distribution of sample region classified by country and continent 24
3.1 Traditional airport organisation 38

3.2 Commercial oriented airport organisation 38
3.3 The ownership structure of UK airports in 1997 50
3.4 The ownership structure of UK airports in 2010 51
3.5 Evolution of airport ownership structure 60
4.1 The research onion 64

4.2 The difference between a deductive and an inductive approach 71
4.3 The operational process in AHP 79
4.4 A simple structure of AHP 80
4.5 Methods of questionnaire administration 95
4.6 Framework of research procedures 97
5.1 AHP hierarchies of Variable Set I (ten variables) 108

5.2 AHP hierarchies of Variable Set II (six variables) 108
5.3 Weights of main-criteria 119
5.4 Weights of sub-criteria 119
5.5 Weights of main-criteria 121
5.6 Weights of sub-criteria 121
5.7 Weights of criteria in input aspect 123
5.8 Weights of criteria in output aspect 123
viii
b LIST OF TABLES a
Table
2.1 Analysis methods used in airport performance evaluation 20

2.2 Distribution of airport performance articles by journal 23
2.3 The most popular input variables 26
2.4 The most popular output variables 28
2.5 Single country research 31
2.6 Cross country research 32
3.1 Activities of non-aeronautical revenue and aeronautical revenue 39

3.2 Average revenue and cost structures at European airports 40
3.3 Level of privatisation for major commercial US airports 44
3.4 The changing of UK airport ownership 53
3.5 Major airports in China 59
4.1 Comparisons of philosophical research paradigms 66

4.2 Critical realist approach of airport efficiency evaluation research 70
4.3 Differences between a deductive and an inductive approach 71
4.4 Research strategy and data collection method 74
4.5 The fundamental scale of absolute numbers 80
4.6 Average Random Index (RI) for corresponding matrix size 82
4.7 Definition of alternative scales 85
4.8 The characteristics of semi-structured interviews 93
4.9 Sample airports in this research 99
5.1 The preliminary indictors of airport efficiency evaluation system 104

5.2 Effectiveness of questionnaires 110
5.3 Pair-wise comparison matrix and weights for level 2 111
5.4 Weights in level two in different groups 111
5.6 Local weights in level 3 in different groups 113
5.8 Local weights in level 3 in different groups 114
5.9 Pair-wise comparison matrix and weights in input aspect 115
5.10 Weights in different groups 115
5.11 Weights of airport efficiency evaluation system (VS I) 117
5.12 Weights of airport efficiency evaluation system (VS II) 118
5.13 Weights of main-criteria variables in input perspective 119
5.14 Weights of sub-criteria variables in input perspective 120
5.15 Weights of main-criteria variables in output perspective 121
5.16 Weights of sub-criteria variables in output perspective 122
5.17 Weights of the variables in VS II 123
5.18 Weights of main criteria of different scales by groups: 124
input perspective
ix
5.19 Weights of main criteria of different scales by groups: 125
output perspective
5.20 Weights of sub criteria of different scales by groups: 126
input perspective
5.21 Weights of sub criteria of different scales by groups: 127
output perspective
5.22 Weights of different scales by groups: input perspective 128
5.23 Weights of different scales by groups: output perspective 129
6.1 Correlation coefficients among inputs and outputs variables: VS I 135

6.2 Panel data of sample airports 136
6.3 Efficiency scores obtained by basic DEA model: VS I 138
6.4 Airport efficiency in different ownership: VS I 139
6.5 The referral of clustering analysis: VS I 140
6.6 Long-term projection of individual airports (1): VS I 142
6.7 Long-term projection of individual airports (2): VS I 143
6.8 Short-term projection of individual airports (1): VS I 144
6.9 Short-term projection of individual airports (2): VS I 145
6.10 Mann-Whitney test of differences in efficiency: VS I 146
6.11 Correlation coefficients among inputs and outputs variables: VS II 147
6.12 Efficiency scores obtained by basic DEA model: VS II 149
6.13 Airport efficiency in different ownership: VS II 150
6.14 The referral of clustering analysis: VS II 151
6.15 Long-term projection of individual airports(1): VS II 152
6.16 Long-term projection of individual airports (2): VSII 153
6.17 Short-term projection of individual airports(1): VS II 154
6.18 Short-term projection of individual airports(2): VS II 155
6.19 Mann-Whitney test of differences in efficiency: VS II 156
6.20 Weights of variables in alternative scales 158
6.21 Relative input and output scores 159
6.22 Relative weighted input and output scores (scale 1-9) 160
6.23 Efficiency scores obtained by AHP/DEA model: VS I 162
6.25 Efficiency scores obtained by AHP/DEA model: VS II 165
6.27 Relative efficiency scores obtained by AHP/DEA model by groups: VS I 169
6.28 Relative efficiency scores obtained by AHP/DEA model by groups: VS II 170
6.30 Mann-Whitney test of differences in efficiency: VS II 172
6.31 A comparison of efficiency scores between the two variable sets 173
6.32 The weight distribution of variables in VS I and VS II 175
7.1 AHP weights of input and output variables of respondents (1-9 scale) 183
7.2 Upper and lower bounds of variables weight ratios (1-9 scale) 184
7.3 Efficiency scores obtained by AHP/DEA-AR model: VS I 187
7.5 The weight distribution of output variables: VS I 189
x
7.6 Relative efficiency scores obtained by AHP/DEA-AR model 191
by groups: VS I
7.8 AHP weights of input and output variables of respondents (1-9 scale) 193
7.9 Upper and lower bounds of variables weight ratios (scale 1-9) 193
7.10 Efficiency scores obtained by AHP/DEA-AR model: VS II 196
7.12 The weight distribution of output variables: VS II 198
7.13 Relative efficiency scores obtained by AHP/DEA-AR model 200
by groups: VS II
7.14 Mann-Whitney test of differences in efficiency: VSII 201
7.15 Relative efficiency scores obtained by AHP/DEA model and 202
AHP/DEA-AR model: VS I
7.16 Relative efficiency scores obtained by AHP/DEA model and 204
AHP/DEA-AR model: VS II
7.17 Progress on AHP approach: VS I(1) 206
7.18 Progress on AHP approach: VS I(2) 207
7.19 Efficiency scores by different approaches 208
7.20 The referral of clustering analysis and efficiency scores 210
7.21 The efficiency scores by different models and approaches (1) 212
7.22 The efficiency scores by different models and approaches (2) 213
7.23 The result of regression analysis (1) 214
7.24 The result of regression analysis (2) 215
8.1 Summary of efficiency scores and Mann-Whitney test 222
xi
Ch1
Introduction
CHAPTER 1
INTRODUCTION
An overview of the research is presented in this chapter. It outlines the challenges of

the research intends to solve, explains why these issues are interesting to explore, and
describes how these issues are addressed. The first section presents the introduction and
motivation. The second section presents the research objectives and research questions.
The third and fourth sections illustrate the structure and contributions.
1
Ch1
Introduction
1.1 INTRODUCTION AND MOTIVATION

Air transport is tightly connected with a country’s economic development; this has been
especially true since the globalisation of most modern economies. Figure 1.1 illustrates
the increase in global air travel since the 1970s. It can be seen that air transport has
faced several external exogenous events (such as the Gulf crisis in 1991, the Asian
crisis in 1998, the War on Terrorism in 2001, the SARS international health crisis in
2003, and the financial crises of 2008), as well as more medium to long term challenges
(oil price surges, airport congestion, and competition with high speed train networks).
Although the performance of the air transport industry was affected by these events, air
transport still grew at a yearly average of 4.2% from 1990 to 2010. It should be noted
that from 2004 and 2010, the Revenue Passenger Kilometers (RPKs) rebounded
quickly after the above events, increasing 14% in 2004 and 7% in 2010 (Airbus 2012).
Figure 1.1: World annual air traffic
Source: Airbus (2012).
According to a recent forecast report produced by Boeing in 2012, air transport will
double in the next 20 years, and the centre of the world air transport flow is expected to
move towards the Asia-Pacific region. More than one third of the value of new
airplanes delivered will be accounted for by this region, compared with about a quarter
for North America and a quarter for Europe. Figures 1.2 and 1.3 illustrate the projected
growth of passengers and traffic in different world regions over the next 20 years.
Figure 1.2 shows that the Asia-Pacific region, specifically China, will become the
principal air transport market. Although the growth rate of North America, Europe, and
2
Ch1
Introduction
the CIS1 are lower than that of the Asia-Pacific, these areas will still maintain their
growth rate because of the increased airline competition brought about by deregulation
and liberalisation, which has heightened this recognition and placed airports in a much
more competitive environment (Barros and Dieke 2007). Further, Figure 1.3 shows that
traffic flow has experienced a trend similar to that of passenger flow.
Figure 1.2: Passenger development Figure1.3: Annual traffic growth
Source: Boeing (2012). Source: Boeing (2012).
Among all participants in the air transport industry, such as carriers (airline companies
and logistics or rail companies) and loading points (such as airports, warehouses,
distribution centres, and seaports), airports are the core of the air transport industry due
to their provision of the logistics of both passenger and cargo services. The most
rapidly developing economies of the world have enormous requirements for advanced
air transport facilities at their airports in order to accommodate the increasing volumes
of air transport in the areas of cargo and passenger services and in order to sustain their
operational efficiency in places such as the Asia-Pacific region. In order to deal with
increasing levels of competition, recently, performance measurement has become an
important means by which civil aviation authorities can determine weaknesses,
especially in those regions facing increasing volumes of air transport in both cargo and
passenger services (Oum et al. 2008).
Under intense global market competition, many countries have explored different long-
term options to maximise efficiency or productivity in regard to operation and resource
1
The CIS includes Russia and the Commonwealth of Independent States.
3
Ch1
Introduction
utilisation, including such things as reforming existing airports (e.g. Taipei Taoyuan
Airport), building new airport terminals (e.g. Beijing, London Heathrow, and Istanbul
airports), or privatising airport management (e.g. London airports). Among these
options, privatising airport management is one of the ways to help governments to
reduce budget barriers, which will in turn contribute to profits. Therefore, in Europe,
airport privatisation policies have been used to improve airport efficiency and resource
utilisation for almost 25 years including even full airport privatisation.
Recently, there has been less progress with privatisation of airports in Asia compared to
other regions of the world because from their viewpoint it would be more cost effective
to restructure public sector enterprises and attempt to turn them around before
instituting privatisation (Joshi 2000). Even though some airports in the Asia-Pacific
region are listed on the stock exchanges (i.e. six Chinese airports: Shenzhen, Shanghai,
Xiamen, Hainan, Beijing, and Guangzhou airports), full airport privatisation is still not
the first option for many authorities in the Asia-Pacific (Zhang and Yuen 2008).
Partially privatised airports may be restructuring for privatisation, but the government
still controls the majority shares of those airports. In addition, the lack of a consistent
privatisation policy also leads to failure to consider different ways for the private sector
to participate as well as a consideration of the relative effectiveness of such alternatives
in regard to achieving a given set of objects (Vickers and Yarrow 1991). This is the
most significant difference existing in the implementation of airport privatisation
policies in these two regions (i.e. Asia-Pacific and Europe).
However, Zhang and Yuen (2008) pointed out that public listing does not significantly
improve airport productivity in China. After reviewing relevant studies, Gong et al.
(2012) also revealed that airport industries did not provide clear patterns of superior
performance associated with particular forms of ownership or organisation. This is
quite different from the common opinions of privatisation. Therefore, one of the aims
of this research is to determine if an airport privatisation policy can really help airport
authorities improve airport efficiency. To achieve this aim, establishing a proper
Airport Efficiency Evaluation System (AEES) is the first task of this research.
In this research, Data Envelopment Analysis (DEA) and an Analytic Hierarchy Process
(AHP) are applied to establish the AEES. DEA can help to recognise relative efficient
4
Ch1
Introduction
airports. However, only using DEA may not provide robust results about the efficiency
of airports because stakeholders may weight the relative importance of input and output
variables differently. For example, airport managers may focus on financial
performance, but civil aviation authorities (i.e. the public sector) may place emphasis
on the number of passengers or aircraft movements. Therefore, another aim in this
research is to develop and demonstrate the applicability of an integrated DEA and AHP
evaluation model for addressing this concern. AHP can help researchers outline the
preferences of different stake holders (i.e. airport managers or airport analysts). In
addition, two means are adopted in this research by addressing sensitivity analysis.
Firstly, basic DEA models, an integrated AHP/DEA model, and an AHP/DEA with
Assurance Region (AR) method are used to evaluate airport efficiency. Secondly,
adjusting the number of variables in the DEA analysis is also adopted as the sensitivity
analysis in this research. The sample airports in this research are selected from Europe
and the Asia-Pacific region as a result of reflection on their different attitudes towards
the role of the government in airport management and also because these two regions
are currently the most competitive areas with regard to air transport. According to the
author’s best ability, in the context of air transportation, an integrated AHP/DEA model
and an AHP/DEA-AR model are firstly applied to evaluate the efficiency of the airports
under consideration.
1.2 RESEARCH QUESTIONS

This research can make some theoretical and methodological contributions. Therefore,
there are two main types of research questions that can be addressed in this research,
including theoretical research questions and methodological research questions.
Theoretical research questions are listed in the first section.
One of the aims of this research is to establish an AEES. Interest in this topic has
prompted a substantial body of research utilising both qualitative and quantitative
approaches. Many of the quantitative approaches calculate efficiency frontiers with an
assumption that all the input and output variables are assumed as having the same
weight (such as Gillen and Lall 1997; Murillo-Melchor 1999; Bazargan and Vasigh
2003; Wang et al. 2004). However, it has been shown through qualitative research that
different stakeholders may place greater emphasis on particular variables (Humphreys
5
Ch1
Introduction
and Francis 2002b). Such an emphasis can be captured through the AHP method and
incorporated into efficiency evaluations. In this research, airport efficiency is evaluated
using three methods (i.e. basic DEA models, an integrated AHP/DEA model, and an
AHP/DEA-AR model). Therefore, the first research question should be addressed as
follows:
Research Question 1:
Does the result of airport efficiency vary as a result of conducting different
evaluation methods?
Another aim of this research is to determine if airport privatisation influences airport

efficiency. Privatisation of their airports is one of the popular means adopted by many
civil aviation authorities to improve efficiency. This strategy was first implemented in
the United Kingdom about 25 years ago (Ison et al. 2011) and has since been adopted
by other western industrialised countries, such as Germany and the Netherlands.
However, most of the major airports in Asia are still operated and owned by the
government or quasi-public enterprises. Therefore, another research question to be
addressed in this research is:
Would an airport privatisation policy (airport ownership) influence the performance
of an airport’s operational efficiency?
The research questions in the following section can be classified as methodological

research questions in this research. In the AHP method, a series of pairwise
comparisons and the unit scale used in its procedure play a fundamental role in
quantifying a decision maker’s preference judgements. To date, the Saaty 1-9 scale is
the 9-unit scale which has been used widely in AHP research. However, the AHP
literature has addressed the question of which of the available alternative scales are
more appropriate for the process of pairwise comparisons (French 1980; Freeling 1983;
Jensen 1984; Legrady et al. 1984; Belton 1986; Harker and Vargas 1987; Schoner and
Wedley 1989; Dyer 1990; Salo and Hämäläinen 1993; Pöyhönen et al. 1997; Beynon
2002). Therefore, the third research question of this research is:
Does the influence of alternative scales on the results of the AHP analysis cause a
different weight for each variable?
6
Ch1
Introduction
The DEA model was first proposed by Charnes et al. (1978) and was subsequently
extended by Banker et al. (1984). It is now widely applied for measuring the efficiency
of many entities, such as schools, public agencies, and banks, among others. (Giokas
1991; Anderson et al. 1998; Oum and Yu 1994). During recent years, the issue of both
the sensitivity and stability of DEA models has been extensively studied. By updating
the inverse of an optimal basis matrix, Charnes et al. (1985) discussed the sensitivity of
the original DEA model. Also, Charnes and Neralic (1990) investigated the sensitivity
of the DEA-additive model and proposed different models to find the stability radius
for an efficient DMU. This research intends to introduce DEA sensitivity analysis by
adjusting a given number of input and output variables and by adopting different DEA
models to undertake and evaluation of airport efficiency. This analysis can help
determine which number of variables or which DEA model can provide the most robust
results. Therefore, the fourth research question is:
Does the number of input and output variables affect the results of airport efficiency
evaluation?
1.3 THE SCOPE AND STRUCTURE OF THE RESEARCH

This research primarily examines the impact of airport privatisation policy on airport
efficiency via sample airports selected from the Asia-Pacific region and Europe (the
reasons are explained later) and also is an attempt to provide the answers to the above
mentioned research questions. It is structured into nine chapters, as follows:
Chapter 1 introduces the subject area of the research, including a basic background,
motivation, objectives, scope, research questions, and the structure of the research. The
chapter also concludes with possible research contributions of the research for
academics, practitioners and policy makers.
Chapter 2 presents a structured literature review that encompasses airport performance

studies. In addition, theories and disciplines involved in airport performance research
are investigated. In addition, changes in research trends are examined in order to
understand the positioning of this research in the periodic trends of airport research.
7
Ch1
Introduction
Potential analysis techniques that can be used in this research are described in detail in
this chapter as well.
Chapter 3 reviews the changing nature of airports in order to understand the categories
of airport ownership and governance, illustrates the evolution of airport ownership and
governance, and justifies the differences between diverse countries. Airport
privatisation policies are described that are related to Research Question 2.
Chapter 4 examines the technical considerations of the current research development.

This chapter connects the previous chapters to the following chapters that develop the
research philosophies applicable to this study. This research adopts the positivist
paradigm for the purpose of understanding airport performance by using different
analysis methods. Choice of methodological approaches, such as research philosophy,
research strategies, survey tool selection, data collection methods, and research design
are examined. This chapter also describes the data analysis methods that are employed
in this research. In this chapter, the characteristics and processes for the AHP, DEA,
and the integrated AHP/DEA model are illustrated. The definitions of alternative scales
for AHP analysis are listed, which can be used to answer Research Question 3.
Chapter 5 addresses the results of the pilot AHP questionnaire and the complete AEES
used in this research is confirmed. Along with variables sets, the AHP method is used
to acquire the weights of variables in 1-9 scales and other alternative scales. An
overview of the relative weights of variables and classification of different groups are
also listed. The chapter ends with a brief summary of the descriptive analysis for this
research.
Chapter 6 shows the empirical results of the model through implementation of the
variable weights discussed in Chapter 6. Prior to the analysis of the measurement
models, the collected data are examined and prepared. After the data preparation
processes, each DEA model is validated and purified through a series of analytical
processes. Finally, the results are examined and a discussion of the proposed
hypotheses is provided. After hypothesis testing, the results can be applied to confirm
the impact of airport privatisation policies. In this chapter, airport efficiency analysis is
computed using a basic DEA model and an integrated AHP/DEA model. One of the
8
Ch1
Introduction
sensitivity analyses is also conducted in this chapter (i.e. different numbers of input and
output variables).
Chapter 7 conducts another sensitivity analysis by applying the AHP/DEA-AR model

to assess airport efficiency. The concepts and analysis process for this model are
described, and airport efficiency is also computed based on different numbers of
variables. In this chapter, the results are examined and a discussion of the proposed
hypotheses is provided. After hypothesis testing, the influence of airport privatisation
policy is confirmed, and the results are also compared with those of Chapter 6. The
results acquired from Chapters 6 and 7 can help the author to provide proper answers to
the research questions. At the end of this chapter, different thoughts on efficiency
evaluation approaches and hypothesis testing methods are also presented.
Chapter 8 concludes the research with an overall summary and a discussion of the key
findings. Finally, the thesis presents a description of the theoretical, methodological
contributions, and the managerial implications of this research, along with some
limitations and recommendations for future research. Figure 1.4 illustrates the research
structure and highlights the scope of each chapter, its context, and the links between the
chapters.
9
Ch1
Introduction
Figure 1.4: Structure of the research
Chapter Scope Research

Questions
Research Motivation Addressed
·∙ Exploration of airport efficiency studies

Chapter 1 ·∙ Statement of research questions
Introduction ·∙ Structure of this study
Literature Review
Chapter 2 ·∙ Presentation of methodological approaches in
Classification and airport performance research RQ 1
analysis of airport ·∙ Classification of results and analysis
performance research ·∙ Analysis of research target
Airport Privatisation and

Commercialisation
Chapter 3
Devolution of Airport ·∙ Discussion of airport ownership and governance RQ 2
ownership and ·∙ Review of airport ownership evolution on different
Government continents
Research Methodology and

Data analysis methods
·∙ Discussions of research philosophy
·∙ Methodologies and approaches of current study
Chapter 4 RQ 3
·∙ Description of AHP
Research Design and
·∙ Description of DEA
Methodology
·∙ Literature review of integrated AHP/DEA model
·∙ Establishment of research design
Descriptive Analysis and

Airport efficiency Analysis
Chapter 5 ·∙ Analysis of AHP survey results
Empirical Analysis: ·∙ Justification of the alternative judgment scales
AHP analysis ·∙ Overview of respondents’ preference
RQ 1
·∙ Preparation of weight of each variable
RQ 2
Chapter 6 ·∙ Analysis of survey variable weights with DEA model RQ 3
·∙ Presentation of relative efficiency of each sample RQ 4
Airport Efficiency
airport
Analysis I
·∙ Comparison of results for different regions
Chapter 7 ·∙ Validity of evaluation models
Airport Efficiency ·∙ Examination of proposed AHP/DEA-AR model
Analysis II ·∙ Results of the hypothesis
·∙ Influence of airport privatisation policy
Research Conclusions and

Suggestions
·∙ Discussion of research findings
Chapter 8
·∙ Theoretical contributions and managerial
Research Conclusions
implications
and Suggestions
·∙ Limitations and future research
10
Ch1
Introduction
1.4 RESEARCH CONTRIBUTIONS

Broadly, this research aims to provide a better understanding of airport privatisation
policy and also to establish a robust AEES, with a particular focus on Europe and the
Asia-Pacific region, as well as to identify the importance of different variables in AEES.
The core of the research is based on perspectives of Europe and Asia-Pacific experts. In
detail, using a combination of theoretical and practical perspectives, this research would
like to:
Ø Build a theoretical system of airport efficiency evaluation.
Ø Employ different evaluation methods to recognise a holistic and robust airport
efficiency evaluation model.
Ø Adopt the AHP/DEA model and AHP/DEA-AR model in airport efficiency
evaluation in the first place.
Ø Understand the privatisation of airport governance and ownership.
It is hoped that this research will provide significant contributions to academics and
practitioners in the airport management sector.
Ø For academics, an improved AEES understanding of links between academia and
practice and a new evaluation model established from different viewpoints should
emerge.
Ø For practitioners, the identification of the variables affecting airport operations,
explored in this study, may stimulate more considered transport decision-making by
providing a more accurate and precise framework for airport planning.
11
Ch2
Analysis of Airport performance Research
CHAPTER 2
ANALYSIS OF
AIRPORT PERFORMANCE RESEARCH
This chapter aims to review existing literature that is related to this research in order to
first provide a context for the undertaken research and to show where this research fits
into the existing body of knowledge; secondly, to illustrate what kind of topics have
been studied previously; thirdly, to outline differences between existing studies, and
finally, to justify the existing studies on this topic.
12
Ch2
2.1 MOTIVATION
Airport performance measures are important to business and operations management,
regulatory bodies, governments, and other stakeholders (Humphreys and Francis
2002b). Airport managers and governments evaluate airport performance for a number
of reasons, including the assessment of financial and operational efficiency, the
evaluation of alternative investment strategies, the monitoring of airport activities from
a safety perspective, and for the purpose of monitoring environmental impact (Doganis
1992). In the mid-1990s, the literature on efficiency evaluation, which had already been
applied to numerous industries (for example, electricity, water, banking, health, and
agriculture) (Giokas 1991; Bureau et al. 1995; Ozcan and McCue 1996; Zang and
Bartels 1998), was introduced to the airport sector.
To follow this trend, a number of relevant studies have been published in the past 20
years although the level of interest in aviation has still been relatively modest as
compared to other industries, with the range of approaches applied reflecting a lack of
consensus in determining the methods that best define the complex reality of the airport
industry. The primary objective of this chapter is to examine how airport performance
evaluation research has been conducted. This chapter takes a methodological
perspective and, as such, considers the broad range of performance variables present in
the literature. This includes aspects such as finance, operations, service quality and the
environment. To achieve this, a structured review of published airport performance
evaluation literature for the last two decades (1990-2012) was undertaken.
2.2 METHODOLOGICAL APPROACHES

The research on airport performance can be classified into two main types: efficiency
evaluations and productivity evaluations. The main difference between efficiency and
productivity evaluations lies in the concept of maximum attainable outputs (Oum and
Yu 2004). Efficiency takes the maximum potential output that can be produced, and it
takes the available inputs into account, while productivity considers the actual outputs.
Therefore, efficiency often relies on comparisons with other Decision Making Units
(DMUs). However, the terms efficiency and productivity are often used
interchangeably even though the underlying meanings of these two terms are not
identical. The fact that changes in productivity are due to changes in efficiency, among
13
Ch2
other factors, may have had an influence in considering both terms as equivalent (Zhu
2009). In order to evaluate efficiency and productivity, previous studies usually have
adopted quantitative methods, relying on numerical and secondary data. They also have
formulated production functions using econometric techniques and advanced efficiency
analysis tools, such as those applied in studies by Sarkis (2000) and Martin et al. (2009).
A theoretical overview of the main approaches is provided as follows:
2.2.1 PARTIAL MEASURES

This method uses partial ratio data to carry out performance comparisons of a target
sample in a single dimension, such as financial or cost performance. It deals with the
ratio of one output to the ratio of one input in order to assess efficiency or productivity
with respect to a specific dimension. It does not give any conclusions on the overall
efficiency (Forsyth 2000). Since the partial measures method only focuses on certain
fields of airport performance, it is relatively easy to calculate and interpret. However,
the evaluation results from this method are not able to provide a more comprehensive
evaluation of an airport’s performance unless they is a part of a broad performance
measurement system. This approach also has other weakness. As discussed by Forsyth
(2000), partial measurement should only be used if no data for overall measures are
available.
2.2.2 MULTI-CRITERIA DECISION MAKING (MCDM)

The Multi-Criteria Decision Making (MCDM) method establishes preferences between
options by reference to an explicit set of objectives that have been identified by the
decision making body and for which it has established measurable criteria intended to
assess the extent to which the objectives have been achieved. In simple circumstances,
the process of identifying objectives and criteria may provide enough information alone
for decision makers. Historically, employing the MCDM method can be divided into
two main steps: acquiring the relative weights for each criterion and ranking the options.
The first stage often uses expert questionnaires or interviews to evaluate the selected
variables, providing the weights to choose an optimal solution (Roy 2005).
The most widely used method in the MCDM is AHP, which was developed from a
linear additive model but, in its standard format, uses procedures for deriving the
weights and the scores achieved by alternatives which are based, respectively, on pair-
14
Ch2
wise comparisons between criteria and between alternatives. Thus, for example, in
assessing weights, the decision makers are asked a series of questions, each of which
asks how important one particular alternative is relative to another for the object being
addressed. The strengths and weaknesses of AHP have been the subject of substantial
debate among specialists in the MCDM (for example, French 1980; Freeling 1983;
Jensen 1984; Legrady et al. 1984; Belton 1986; Harker and Vargas 1987; Schoner and
Wedley 1989; Dyer 1990; Saaty 1990; Salo and Hämäläinen 1993; Pöyhönen et al
1997). It is, however, clear that users generally find the pair-wise comparison form of
data input straightforward and convenient.
The other methods which are used in the MCDM include:

Ø PROMETHEE (Preference Ranking Organisation Method for Enrichment
Evaluations), which was proposed by Brans and Mareschal (2005) and which
defines preference functions based on the differences between criteria among
different schemes (Hu et al. 2010).
Ø TOPSIS (Technique for Order Preference by Similarity to Ideal Solution), which
was proposed by Hwang and Yoon (1981). This uses ideal and anti-ideal solutions to
find the best alternative. It assumes that each indicator takes a monotonic function
(increasing or decreasing) utility (Wang et al. 2004 and Hu et al. 2010).
Ø ELECTRE (Elimination and Choice Expressing Reality), which was developed by
Benayoun et al. (1966) and improved by Roy (1971; 1991). The ELECTRE
methodology has evolved through a number of different versions (I through IV).
ELECTRE I is designed for choosing a single action, while ELECTRE II, III and IV
deal with ranking. The subset containing all the most satisfying alternatives is
obtained by eliminating the greatest number of alternatives (Bojkovic et al. 2010).
2.2.3 FRONTIER ANALYSIS

Three main methods which have been adopted by scholars to analyse efficiency are:
Stochastic Frontier Analysis (SFA), Total Factor Productivity (TFP), and Data
Envelopment Analysis (DEA). Although these methods are similar in that they
determine a frontier and an inefficiency based upon that frontier, which is the efficient
frontier, there is a significant difference between them. Although SFA estimates
inefficiency, it can also be used as an explanation for inefficiency (Pels et al. 2003).
15
Ch2
Furthermore, the DEA approach provides a measurement of inefficiency. A brief

description of each method is given below:
2.2.3.1 STOCHASTIC FRONTIER ANALYSIS (SFA)

SFA models were first proposed by Aigner et al. (1977); they have since been used
extensively in scientific literature. In recent years, several alternatives have been
proposed in the literature to relax the restrictive assumption that all DMUs share the
same technological parameters. For example, Kalirajan and Obwona (1994), Tsionas
(2002), Huang (2004) and Greene (2005) developed different versions of random
coefficient models which are based on SFA, in which differences between DMUs (i.e.
heterogeneity) are modelled in the form of continuous parameter variations. More
recently, Kumbhakar et al. (2007) developed a non-parametric stochastic frontier using
a local maximum likelihood approach.
In 2005, Greene proposed two alternative panel data estimators that the author labelled
as true random effects and true fixed effects. The main feature of Greene’s (2005)
model is that a time-invariant DMU effect co-exists with inefficiency in order to avoid
the inefficiency term picking up a DMU’s heterogeneity. In addition, the true random
effects model (which assumes that there is a DMU-specific random term to capture
DMU heterogeneity) has been used widely. On the other hand, the true fixed effects
model assumes that the DMU-specific term is a fixed parameter and is allowed to be
correlated with the included variables. Although these parametric approaches take into
account the effect error, they still face challenges with regard to separating random
error from efficiency.
2.2.3.2 TOTAL FACTOR PRODUCTIVITY (TFP)

Total Factor Productivity (TFP) is determined by how efficiently and intensely the
inputs are utilised in production. In other words, it is the portion of output not explained
used in production (Comin, 2006). TFP requires an aggregation of all outputs into a
weighted output index. It also requires that all inputs be placed into a weighted input
index using pre-defined weights, which can be biased. The key drawback of the TFP
technique is that it does not allow for random error in the data, assuming that
measurement error and luck are factors that affect outcome. This implies that the
measured inefficiency is likely to be overstated (Berger and Humphrey 1997).
16
Ch2
Moreover, it constructs a frontier based on the actual data in the sample, and the
relative efficiency of each DMU in the population is calculated in relation to this
frontier. Therefore, the result may be very sensitive to the chosen sample and the
outliers.
2.2.3.3 DATA ENVELOPMENT ANALYSIS (DEA)

DEA requires no assumptions about the functional form and calculates a maximal
performance measure for each DMU relative to all other DMUs. DEA was originally
developed by Charnes et al. (1978) by applying a linear programming technique that
converted multiple inputs and multiple outputs into a scalar measure of relative
productive efficiency to construct a frontier based on the sample. DMUs on the frontier
are efficient, while DMUs inside the frontier are inefficient. With the assumption of no
random errors, all deviations from the estimated frontier are attributed to inefficiencies.
Both non-parametric methods (DEA and TFP) compare a weighted output variable
relative to a weighted input variable. However, the key advantage of DEA is that the
input and output weights result from a linear programming procedure rather than being
pre-determined (Graham 2005). DEA is often a more attractive technique when
compared with the other two methods because of its less demanding data requirements.
In general, the main motivation for choosing DEA is often its flexibility with regard to
accounting for multiple input/output variables in the estimation of efficiency (Banker
1984). This method can also account for external factors that are related to the
environment in which a particular DMU operates. One of the method’s major
weaknesses is that it has no statistical properties and, hence, does not account for
measurement error in the estimation of efficiency (Charnes et al. 1985). The use of
DEA can become even more problematic in the presence of outliers, which can simply
distort the derived efficiency results (Russell 1985).
Recently, the bootstrapping method, which is one of the DEA approaches, is a

computer-based method for assigning measures of accuracy to sample estimates (Efron
and Tibshirani 1994) has became popular. This technique allows estimation of the
sample distribution of almost any statistic using only very simple methods (Varian
2005). Generally, it falls in the broader class of resampling methods. In addition,
17
Ch2
Network Data Envelopment Analysis (NDEA) is currently emerging as a way of

evaluating the efficiency of sub-processes within the overall system (Kao 2009).
2.3 LITERATURE ANALYTICAL DISCIPLINES

This section describes the results of a structured literature review about the air transport
field which has been carried out to confirm the existent research gaps, which is a new
approach to adopt in airport performance evaluation research. A structured literature
review is a means of identifying, evaluating and interpreting all available research
which is relevant to a particular research question, topic area, or phenomenon of
interest. Individual studies contributing to a structured review are called primary studies;
a structured review is a form of secondary study (Armitage and Keeble-Allen, 2008). In
this research, only papers published in academic journals are included for content
analysis in this chapter. Other reports published by research institutions (such as the
annual Global Airport Benchmarking Report published by the Air Transport Research
Society or other industry reports), airport authorities (individual airport performance
report) or the government sector (annual reports) were not included. Moreover,
conference papers, news reports, book reviews, viewpoints, master and doctoral
dissertations, textbooks, and unpublished working papers were also excluded. The
reason why this research only includes academic papers is because most academic
papers are published under a serious trial process. Therefore, their results tend to be
more reliable than those found in other reports. In addition, a structured literature
review was undertaken in this research. It is very difficult to construct a structured
procedure to review industry reports due to the large number of items. Besides, no
specific data base can include most industry reports, and it is difficult to narrow down
these kinds of reports. An example can help to show how tough it is to track non-
academic publications. If one uses the search term “airport efficiency” in a Google
search engine, there are more than 200,000 items that can be found. Therefore, only
academic papers were included in this literature review.
The main search terms in the survey were ‘airport’ with ‘efficiency’, ‘performance’, or
‘productivity’, and the end of June 2012 was selected as the cut-off date. Various online
journal databases were selected and searched to provide a comprehensive bibliography
on airport performance evaluation literature. The literature contributions included
18
Ch2
articles from the following research databases: Emerald, Science Direct (Elsevier),
ProQuest Global, Google scholar, and SCOPUS. These databases provide online
delivery systems with full text access to thousands of high quality articles and journals
that cover a wide range of social and applied science titles, including business and
management disciplines. However, some journals may be beyond the scope of these
databases, and therefore, their contributions may not be included in the results. The
search yielded 66 airport performance evaluation articles from 23 journals. Details of
these can be found in Appendix I and in the bibliography. Each article was carefully
reviewed, and the data was organised to produce a classification from several
perspectives. Consequently, this research serves as a comprehensive base for an
understanding of airport performance evaluation research.
The classification framework is based on the literature review, the nature of airport
performance evaluation research and the work of Gonzalez and Trujillo (2009) and
Pallis et al. (2010) (who have conducted studies in a similar field related to (sea) port
efficiency). The articles were reviewed, analysed, and classified based on four
perspectives, as follows:
(1) Distribution by year of publication and methodology;
(2) Distribution of articles by journal;
(3) Geographical distribution of airport performance research; and,
(4) Analysis of input and output variables.
This framework provides guidelines for pursuing rigorous airport performance
evaluation research by explaining the chronological growth of the benchmarking
technique, challenging themes of airport performance evaluation research and
application areas in airport performance evaluation.
2.3.1 DISTRIBUTION BY YEAR OF PUBLICATION AND METHODOLOGY

Table 2.1 shows the distribution of 66 articles published between 1990 and June of
2012. It reveals that only limited papers were published on this topic before the year
2000. The first recorded paper was published in 1997. However, Graham (2005)
highlighted a small number of non-journal publications that preceded this date. In the
past ten years, the number of journal articles has increased significantly. This growth
trend is quite similar to the early trends in (sea) port performance evaluation although
(sea) ports have been studied over a longer period, leading to more than one hundred
19
Ch2
articles evaluating their performance (Gonzalez and Trujillo 2009; Woo et al. 2010).
This suggests that there is a potentially rich continuous flow of research in the air
transport field for many years to come especially in this year (2012).
Table 2.1: Analysis methods used in airport performance evaluation
Total
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
Partial
- - - 1 - 1 - - - - - - - - - 2
Measures
MCDM - - - - - - - 1 - - - - - - - 1
Frontier Analysis
Parametric Approach
SFA - - - - - - - - - - - 2 1 1 1 1 6
Non-Parametric Approach
DEA 1 - 2 1 2 1 2 3 - 1 1 3 2 2 2 3 26
TFP 1 - - - - - 1 1 - 1 - 1 - - 1 1 8
Combination* - - - - 1 1 1 1 - 1 - 1 1 3 1 4 15
Other
Research - - - - - 1 - 1 - 1 - - 2 2 1 1 9
Methods
Total 2 0 2 2 3 4 4 7 0 4 1 7 6 8 6 10 66
* Combinations include: DEA and TFP; DEA and SFA; and SFA and TFP.
Source: Organised by author.
Only two papers adopted the partial measures method to evaluate airport efficiency
between 1990 and 2012. Francis et al. (2002) revealed that most traditional airport
financial performance measures were based around a Work Load Unit (WLU), defined
as one passenger processed or 100kg of freight handled. In the 1980s, this measure was
taken from the airlines and adopted by airports to provide a single measure of output for
both the passenger and freight business. Typical measures used included total cost per
WLU, operating cost per WLU, and labour cost per WLU. A deeper discussion of the
measures (including details on a number of publications which lie beyond the scope of
the method adopted in this thesis) can be found in Graham (2005). A recent example of
the application of this approach can be found in the Competition Commission’s
investigation into the BAA in the UK (Competition Commission 2008).
MCDM approaches have also seen very limited applications in the context of airport
efficiency. The only example is Wang et al. (2004), who used TOPSIS to evaluate the
operational performance of Taiwan’s major airports. PROMETHEE have seen no
applications that has been confirmed in Behzadian et al. 2009. AHP applications within
aviation generally have focused on airport logistics (Tsai and Su 2002) and hub airport
allocation (Berrittella et al. 2009), but not on airport efficiency evaluation (the use of
20
Ch2
online journal databases in this study has been previously described). Several key
papers are listed as follows: Tsai and Su (2002) used AHP to compute the relative
weights and to identify the critical political risk factors that influenced the development
of an air logistics hub in Taiwan. Yoo and Choi (2006) conducted an AHP analysis of
surveyed data about the relative importance of the factors and elements concerned with
the improvement of passenger screening. Berrittella et al. (2009) developed an
application of AHP to rank the operating cost components of full service and low cost
airlines; however, in this particular study, AHP was not used in the context of
efficiency/performance evaluation, and it was not combined with DEA. Castelli and
Pellegrini (2011) used AHP to assess the opportunity of implementing this concept by
considering the views of experts. These findings indicate that there are some net
benefits for airlines and air navigation service providers who use AHP, but not for
airports. From the above brief review, according to the author’s knowledge, currently
there has been no paper published that has attempted to combine AHP with DEA in the
area of air transport.
Furthermore, DEA is the most popular method that is used when evaluating airport
performance, producing a steady flow of research throughout the time period under
examination. Twenty-six papers were found that used DEA to evaluate airport
performance from the point of view of the airport authorities. Although Adler and
Berechman (2001) also used DEA, they chose the airlines’ viewpoint to analyse airport
quality and performance. TFP has also been used in occasional publications, such as the
regular Global Airport Benchmarking report. Meanwhile, Hooper and Hensher (1997)
were the first researchers to use TFP, examining the performance of six Australian
airports over a four year period. More recently, between 2008 and 2012, the SFA has
been used as an individual method in six papers. Oum et al. (2008) and Barros (2008b)
were the first two papers that adopted SFA to evaluate airport efficiency. Oum et al.
(2008) evaluated the effects of ownership form on airport cost efficiency by applying
SFA on the world’s major airports. Barros (2008b) used a random stochastic frontier
model to estimate the technical efficiency of UK airports. Moreover, the bootstrap
approach in the DEA context has been widely adopted in the past few years (Assaf
2010; Curi et al. 2010; 2011). In 2011, Assaf also used the Malmquist bootstrapped
combined methodology to assess the extent of productivity, efficiency, scale and
technological changes at the major Australian airports. In 2012, there was a new
21
Ch2
method applied to evaluate airport efficiency called the Bayesian dynamic frontier
model. Assaf et al. (2012) applied this model to assess UK airport efficiency. Assaf and
Gillen (2012) adopted this model to combine with SFA to compare the efficiency of 73
international airports. Finally, combinations of methods and other approaches have also
seen a small but regular flow of publications (e.g. Pels et al. 2001; Martin and Roman
2006 and Yang 2010). However, these combinations have focused on bringing together
different (objective) frontier analysis techniques, rather than bringing in the subjectivity
of the MCDM (Wang et al. 2004). As outlined earlier, the motivation behind this
research is to overcome the relative weaknesses of individual methods. However, after
reviewing all of the papers on this topic, it was found that most of the previous studies
on airport performance measurement have failed to consider other important variables
that can influence an airport’s performance evaluation, such as the characteristics of
airport authorities and airport users (e.g. airline companies or passengers).
2.3.2 DISTRIBUTION OF ARTICLES BY JOURNAL

A total of 23 different journals from various subjects (including urban planning,
operations management, economics, and transportation) published airport performance
articles in the target period. Table 2.2 illustrates that the vast majority of articles (i.e. 37
articles, or 56% of the total) were published in just two journals, which are: Journal of
Air Transport Management (JATM) and Transportation Research Part E (TRE). The
list of the articles published shows that publications in the TRE represent the first
publications of particular analytical approaches (or combinations of them) in the
context of airport efficiency, such as Gillen and Lall (1997) for DEA, and Hooper and
Hensher (1997) for TFP. By contrast, the publications within the JATM tend to contain
papers that focus more on the applications of these techniques. All of the analysis
methods in Table 2.1 have been applied across the 24 papers in the JATM. This
concentration of publications within a limited number of sector-specific publications is
similar to that occurring with regard to port efficiency (see Gonzalez and Trujillo 2009;
Woo et al. 2010). This concentration of publications in a limited number of journals
does have some advantages since it means that there is a solid body of research which
is relatively easy to locate, while these journals clearly appear to be natural homes for
airport efficiency publications. However, there is a danger that this research may
become quite insular and lacking in impact, or that it draws inspiration from other
research outside of the air transport sector. Looking forward, the findings here suggest
22
Ch2
that researchers should consider a broader range of journals for publications to ensure
that opportunities from other disciplines can be exploited.
Table 2.2: Distribution of airport performance articles by journal

Journal Title Author Name and Published Date
The Australian Economic Review Abbott and Wu (2002). 1
Computers and Industrial Engineering Yang (2010). 1
European journal of operational research Assaf and Gillen (2012). 1
International Journal of production economic Gitto and Mancuso (2012b). 1
Murillo-Melchor (1999); Gillen and Lall (2001);

International Journal of Transport Economics 3
Barros and Sampaio (2004).
International Journal of
Humphreys and Francis (2002b). 1
Transport Management
Journal of Air Transportation Vashigh and Gorjidooz (2006). 1
Hamzaee and Vasigh (2000);

Martin and Roman (2001); Francis et al. (2002);
Martin-Cejas (2002); Bazargan and Vasigh (2003);
Oum et al. (2003); Wang et al. (2004); Yu (2004);
Oum et al. (2006); Lin and Hong (2006);
Journal of Air Transport Management Barros and Dieke (2007); Barros (2008a); Barros (2008b); 24
Assaf (2009); Chi-Lok and Zhang (2009);
Manataki and Zografos (2009); Curi et al. (2010);
Ablanedo-Rosas and Gemoets (2010); Tsekeris (2011);
Chow anf Fung (2012); Gitto and Mancuso (2012a);
Scotti et al. (2012); Wanke (2012a); Zhang et al. (2012).
Journal of Operations Management Sarkis (2000). 1
Journal of Productivity Analysis Martin et al. (2009). 1
Journal of Transport Economics and Policy Parker (1999). 1
Journal of Urban Economics Oum et al. (2008). 1
Networks and Spatial Economics Martin and Roman (2006); Lozano and Gutierrez (2011). 2
Omega Yu (2010). 1
Pacific Economic Review Barros et al. (2010); Fung and Chow (2011). 2
The Service Industries Journal Assaf (2011). 1
Socio-Economic Planning Sciences Curi et al. (2011); Wanke (2012). 2
Transport Policy Pels et al. (2001); Barros (2011). 2

Fernandes and Pacheco (2002);
Transportation Research Part A 3
Pacheco and Fernandes (2003); Sarkis and Talluri (2004).
Gillen and Lall (1997); Hooper and Hensher (1997);
Pels et al. (2003); Oum and Yu (2004); Yoshida (2004); Yoshida
and Fujimoto (2004); Barros and Dieke (2008);
Transportation Research Part E 13
Fung et al. (2008); Pathomsiri et al. (2008);
Yu et al. (2008); Lam et al. (2009);
Tovar and Martin-Cejas (2010); Assaf et al. (2012).
Transport Review Barros (2009). 1

Tourism Management Assaf (2010). 1
Utilities policy Perelman and Serebrisky (2012). 1
Total Number of Journals: 23 Total Number of Papers: 66
23
Ch2
2.3.3 DISTRIBUTION OF ARTICLES BY GEOGRAPHY

Figure 2.1 below illustrates the geographical trend in airport performance research. Of
these 66 papers, 12% (eight papers)* studied airports in the United States (US), while
10% (seven papers)* studied Spanish airports. The focus on these countries reflects both
the affiliations of those undertaking the research plus the availability of data for
analysis. By grouping these papers into continents, it can be seen that 36% (24 papers)
attempted to evaluate airport performance in Europe. This survey reveals an interesting
point in that only five papers investigated airport performance in South America
(Fernandes and Pacheco 2002; Barros 2008a; Perelma and Serebrisky 2012; Wanke
2012a; 2012b). To date, only one paper has investigated African airport performance
(Barros 2011). Two reasons behind this are the availability of suitable data for analysis
and the relative importance of these areas in the global air transport industry. For future
research, there are a considerable number of opportunities for efficiency evaluations in
other countries and regions. Further opportunities exist in conducting comparative
studies across different regions and, as time progresses, conducting increasingly
longitudinal studies in specific countries to evaluate policy changes.
Figure 2.1: Distribution of sample region classified by country and continent

9
Africa
8 1%
7 World
6 Asia-‐ 11%
Paciﬁc
5 29%
America
4 23%
3
2 EU
36%
1
0
2.3.4 ANALYSIS OF INPUT AND OUTPUT VARIABLES

Tables 2.3 and 2.4 highlight the degree of diversity in the input and output variables in
regard to the research conducted, reflecting previous comments by Yoshida (2004).
*
These papers are: Gillen and Lall (1997); Sarkis (2000a); Hamzaee and Vasigh (2000); Gillen and Lall (2001);
Bazargan and Vasigh (2003); Sarkis and Talluri (2004); Vashigh and Gorjidooz (2006); Pathomsiri et al. (2008).
*
These papers are: Murillo-Melchor (1999); Martin and Roman (2001); Martin-Cejas (2002); Martin and Roman (2006);
Martin et al. (2009); Tovar and Martin-Cejas (2010); Lozano and Gutierrez (2011).
24
Ch2
This section considers these variables in more detail. Ülkü (2009) provided a
classification that was based on broad categories of input and output variables.
However, this research considers each of these individually, and it also looks at the
specific variables used. The following analysis can provide the concepts by which to
construct variables in the preliminary AEES.
2.3.4.1 INPUT VARIABLES ANALYSIS

The input variables that have been previously used in airport performance research
(Table 2.3) can be divided into two categories: airport service variables (such as the
number of employees, gates or length of runway), which were related to the services
that were provided by the airport; and financial variables (such as operational cost and
capital cost), which were related to financial performance. In the first category of
airport service variables, the number of employees (which is an essential variable of
production in airport activities) was adopted in almost half of the airport performance
papers as an input variable. In addition, the size of the terminal and the number of gates
were also widely adopted for use in a large number of studies. Among financial
variables, operational cost was the most widely introduced. Generally, most of the
studies used between three and eight input variables. As noted by Ülkü (2009), the
majority drew their inputs from either airport service or financial variables. Airport
service variables were generally more popular, which may be due to data availability,
the lack of published financial accounts for airports (especially publicly owned airports),
difficulties in reconciling different accounting practices, or currency fluctuations
affecting international comparisons. There were two types of approach that were used
in studies of service variables. Most authors adopted a counting approach to airport
facilities, using values such as the number of runways or the number of gates (Gillen
and Lall 2001; Lin and Hong 2006). A few studies considered area instead, with inputs
such as area of departure lounge or the area of the apron (Pacheco and Fernandes 2003;
Yu 2004). The former approach considers the efficient use of assets while the latter
evaluates the efficient use of space. Only a few studies took variables from both service
and finance categories (Bazargan and Vasigh 2003). This represents a good opportunity
for future research because it is possible to provide a more balanced approach.
25
Ch2
Table 2.3: The most popular input variables

Input Variables Papers
Airport Service Variables
Gillen and Lall (1997; 2001); Murillo-Melchor (1999); Parker (1999);

Sarkis (2000); Abbott and Wu (2002); Oum et al. (2003; 2008);
Barros and Sampaio (2004); Oum and Yu (2004); Sarkis and Talluri (2004);
Yoshida and Fujimoto (2004); Lin and Hong (2006); Barros (2008a);
Number of employees 27
Yu et al. (2008); Lam et al. (2009); Martin et al. (2009); Assaf (2010);
Barros et al. (2010); Tovar and Martin-Cejas (2010); Yang (2010);
Yu (2010); Assaf (2011); Curi et al. (2011); Assaf and Gillen (2012);
Perelman and Serebrisky (2012); Scotti et al. (2012).
Gillen and Lall (1997); Pels et al. (2001; 2003);

Fernandes and Pacheco (2002); Pacheco and Fernandes (2003);
Oum et al. (2003; 2008); Yoshida (2004); Yoshida and Fujimoto (2004);
Yu (2004); Lin and Hong (2006); Barros (2008a); Fung et al. (2008);
Size of terminal area 25
Chi-Lok and Zhang (2009); Barros et al. (2010); Yu (2010); Assaf (2011);
Fung and Chow (2011); Lozano and Gutiterrez (2011); Tsekeris (2011);
Assaf and Gillen (2012); Chow and Fung (2012);
Perelman and Serebrisky (2012); Scotti et al. (2012); Wanke (2012b).
Gillen and Lall (1997; 2001); Sarkis (2000); Bazargan and Vasigh (2003);
Pels et al. (2003); Oum et al. (2003; 2008); Sarkis and Talluri (2004);
Number of runways 16
Lin and Hong (2006); Barros (2008a); Pathomsiri et al. (2008);
Assaf (2010); Yang (2010); Curi et al. (2011); Tsekeris (2011); Wanke (2012b).
Gillen and Lall (1997; 2001); Sarkis (2000); Bazargan and Vasigh (2003);
Number of gates Oum et al. (2003); Sarkis and Talluri (2004); Lin and Hong (2006); 9
Tovar and Martin-Cejas (2010); Lozano and Gutiterrez (2011).
Fernandes and Pacheco (2002); Pacheco and Fernandes (2003);

Size of apron Yu (2004); Barros (2008a); Yu (2010); Curi et al. (2011); 9
Lozano and Gutiterrez (2011); Tsekeris (2011); Wanke (2012b).
Pels et al. (2001; 2003); Fernandes and Pacheco (2002);

Number of check-in desks Pacheco and Fernandes (2003); Lin and Hong (2006); 7
Lozano and Gutiterrez (2011); Scotti et al. (2012).
Gillen and Lall (1997); Abbott and Wu (2002);
Length of runway Yoshida and Fujimoto (2004); Fung et al. (2008); Chi-Lok and Zhang (2009); Fung, 7
Chow (2011); Chow and Fung (2012); Wanke (2012b).
Gillen and Lall (1997; 2001); Fernandes and Pacheco (2002);

Number of parking spots 6
Pacheco and Fernandes (2003); Lin and Hong (2006); Wanke (2012b).
Gillen and Lall (1997; 2001); Pels et al. (2001);

Number of collection belts 6
Lin and Hong (2006); Lozano and Gutiterrez (2011); Scotti et al. (2012).
Pels et al. (2001; 2003); Lin and Hong (2006); Scotti et al. (2012);
Number of aprons 5
Wanke (2012b).
Financial Variables
Sarkis (2000); Bazargan and Vasigh (2003); Sarkis and Talluri (2004);
Vashigh and Gorjidooz (2006); Barros and Dieke (2007; 2008);
Operational cost Barros (2008b); Barros (2009); Curi et al. (2010); Yang (2010); 14
Assaf (2011); Barros (2011); Assaf and Gillen (2012);
Gitto and Mancuso (2012b).
Hooper and Hensher (1997); Parker (1999);
Martin and Roman (2001; 2006); Barros and Sampaio (2004);
Capital cost 10
Barros and Dieke (2007; 2008); Martin et al. (2009); Curi et al. (2010);
Assaf et al. (2012).
Hooper and Hensher (1997); Martin and Roman (2001; 2006);

Labour cost Barros and Dieke (2007; 2008); Oum et al. (2008); Curi et al. (2010); 9
Assaf et al. (2012); Gitto and Mancuso (2012b).
Murillo-Melchor (1999); Parker (1999);
Amount of capital stock 4
Abbott and Wu (2002); Yu et al. (2008).
26
Ch2
2.3.4.2 OUTPUT VARIABLES ANALYSIS

The output variables (see Table 2.4) used in airport performance research can again be
divided into airport service variables (such as number of passengers, amount of cargo,
and number of aircraft movement) and financial variables (such as operational revenue,
non-operational revenue, aeronautical revenue, and non-aeronautical revenue).
However, despite the importance of such variables for airport managers, the number of
studies using financial outputs is relatively limited, even where financial inputs are used
(Humphreys and Francis 2002b).
Among output variables, the number of passengers was the most broadly adopted
variable to evaluate airport efficiency. In addition, the amount of cargo, mail and
aircraft movement were also considered by many studies to be essential in airport
activities. Among financial variables, non-aeronautical revenue was the most broadly
used financial variables. Oum et al. (2003) revealed that, in addition to passenger traffic,
cargo traffic and aircraft movements, airports also derive revenues from concessions,
car parking, and numerous other services. These ‘other’ services are not directly related
to aeronautical activities in a traditional sense, but they are becoming increasingly more
important for airports around the world. Consequently, when considering output
variables, that of revenues from commercial or non-aeronautical services should be
included.
Most studies only applied three output variables, a notable exception being the study of
Barros and Dieke (2007), which used both service and financial variables. Considering
other outputs of airports, variables such as punctuality were only included in one paper
(i.e. Bazargan and Vasigh 2003) even though there were many external influences on
this variable. More interestingly, with topics related to sustainability now becoming
increasingly important, environmental outputs only have been studied in a few papers,
such as research by Yu (2004), who considered aircraft noise as an output. In addition,
Graham (2004) suggested the use of several variables when evaluating airport
environmental performance (such as waste per passenger and water consumption per
passenger). These environmental variables offer significant potential scope for future
research.
27
Ch2
Table 2.4: The most popular output variables

Input Variables Papers
Airport Service Variables
Gillen and Lall (1997; 2001); Murillo-Melchor (1999); Parker (1999);

Sarkis (2000); Martin and Roman (2001); Pels et al. (2001; 2003);
Abbott and Wu (2002); Fernandes and Pacheco (2002);
Pacheco and Fernandes (2003); Oum et al. (2003; 2006; 2008);
Barros and Sampaio (2004); Oum and Yu (2004); Sarkis and Talluri (2004);
Yoshida (2004); Yoshida and Fujimoto (2004); Yu (2004);
Lin and Hong (2006); Martin and Roman (2006);
Vashigh and Gorjidooz (2006); Barros and Dieke (2007; 2008);
Number of passengers 50
Barros (2008a; 2008b; 2009); Fung et al. (2008); Pathomsiri et al. (2008);
Yu et al. (2008); Chi-Lok and Zhang (2009); Lam et al. (2009); Assaf (2010);
Ablanedo- Rosas and Gemoets (2010); Curi et al. (2010); Yu (2010);
Assaf (2011); Barros (2011); Curi et al. (2011); Fung, Chow (2011);
Lozano and Gutiterrez (2011); Tsekeris (2011); Assaf and Gillen (2012);
Chow and Fung (2012); Gitto and Mancuso (2012b);
Perelman and Serebrisky (2012); Scotti et al. (2012); Wanke (2012a);
Wanke (2012b); Zhang et al. (2012).
Sarkis (2000); Martin and Roman (2001); Pels et al. (2001; 2003);
Oum et al. (2003; 2006; 2008); Oum and Yu (2004);
Barros and Sampaio (2004); Sarkis and Talluri (2004); Yoshida (2004);
Yoshida and Fujimoto (2004); Yu (2004); Lin and Hong (2006);
Martin and Roman (2006); Vashigh and Gorjidooz (2006);
Barros (2008a; 2008b; 2009); Barros and Dieke (2008); Fung et al. (2008);
Aircraft movements Chi-Lok and Zhang (2009); Lam et al. (2009); Martin et al. (2009); 43
Assaf (2010); Ablanedo- Rosas and Gemoets (2010); Curi et al. (2010);
Yu (2010); Assaf (2011); Barros (2011); Fung, Chow (2011); Curi et al. (2011);
Lozano and Gutiterrez (2011); Tsekeris (2011); Assaf and Gillen (2012);
Perelman and Serebrisky (2012); Scotti et al. (2012); Wanke (2012b);
Zhang et al. (2012).
Gillen and Lall (1997; 2001); Parker (1999); Sarkis (2000);

Martin and Roman (2001); Abbott and Wu (2002); Oum et al. (2003; 2008);
Barros and Sampaio (2004); Sarkis and Talluri (2004); Wang et al. (2004);
Yoshida (2004); Yoshida and Fujimoto (2004); Lin and Hong (2006);
Martin and Roman (2006); Barros and Dieke (2007, 2008); Barros (2008a);
Fung et al. (2008); Pathomsiri et al. (2008); Chi-Lok and Zhang (2009);
Amount of cargo and mail 39
Lam et al. (2009); Ablanedo- Rosas and Gemoets (2010);
Assaf (2010); Curi et al. (2010); Yu (2010); Assaf (2011); Curi et al. (2011); Fung,
Chow (2011); Lozano and Gutiterrez (2011); Tsekeris (2011);
Perelman and Serebrisky (2012); Scotti et al. (2012); Wanke (2012a);
Wanke (2012b); Zhang et al. (2012).
Financial Variables
Hooper and Hensher (1997); Bazargan and Vasigh (2003);

Oum et al. (2003; 2006; 2008); Oum and Yu (2004);
Amount of Barros and Sampaio (2004); Martin and Roman (2006);
16
non-aeronautical revenue Barros and Dieke (2007; 2008); Curi et al. (2010);
Tovar and Martin-Cejas (2010); Assaf and Gillen (2012); Assaf et al. (2012); Gitto and
Mancuso (2012a); Gitto and Mancuso (2012b).
Hooper and Hensher (1997); Bazargan and Vasigh (2003);

Amount of Barros and Sampaio (2004); Martin and Roman (2006);
10
aeronautical revenue Barros and Dieke (2007; 2008) Curi et al. (2010); Assaf et al. (2012);
Gitto and Mancuso (2012a); Gitto and Mancuso (2012b).
Amount of Sarkis (2000); Sarkis and Talluri (2004);

4
operational revenue Vashigh and Gorjidooz (2006); Yang (2010).
Amount of
Vashigh and Gorjidooz (2006). 1
non-operational revenue
28
Ch2
2.4 ANALYSIS OF RESEARCH SAMPLE

Regarding comparing the results arising from the research, it should be noted that
efficiency is a relative concept: the efficiency of a DMU is measured in relation to the
frontier that, in turn, is defined by a group of DMUs. This means that any change in the
group of DMUs analysed, such as the inclusion or exclusion of an airport, will change
the calculated efficiency indexes. In general terms, people indicate that the performance
of airports has improved over time since most studies have found evidence of
improvements in efficiency, productivity, or in regard to the introduction of new
technology. A review of the previous studies have shown that many different variables
affect airport efficiency, including airport characteristics, hub status and traffic
structure (Gillen and Lall 1997; Oum et al. 2006; Oum et al. 2006; Tovar and Martin-
Cejas 2010). In terms of managerial control, commercialisation, privatisation, and
outsourcing policy also influence airport performance (Oum et al. 2003; Barros and
Dieke 2008; Vasigh and Gorjidooz 2006). However, airports in different countries are
affected by different variables. Therefore, the following sections discuss papers that
have a single country research focus and those which cover multiple countries
(reflecting the data in Figure 2.1).
2.4.1 SINGLE COUNTRY RESEARCH

This section focuses upon those countries where there have been at least four different
studies into airport efficiency, which include the USA, UK, Spain, and Taiwan. Details
of these studies can be found in Table 2.5. In general terms, most of these studies
showed an improvement in airport efficiency over time, with only Gillen and Lall
(2001) claiming that there was a 0.1% decline in the terminal side and movement side.
A key feature of many of these single country studies was the comparison of airports by
size. The typical classification that is used is small, medium and large. In the US, early
studies were conducted by Gillen and Lall (1997) and Sarkis (2000), who both used
data covering a period from 1989 to 1994. They both found that larger airports are more
efficient than small airports. More recent studies covering the period from 1996 to 2003,
showed that smaller US airports have became more efficient than large airports
(Bazargan and Vasigh 2003; Pathomsiri et al. 2008). This finding suggests that there
are changing dynamics within the airport system in the US.
29
Ch2
Similar trends can be seen in studies of airports in the UK and Taiwan. In the UK, large
airports have gone from being less efficient (Barros 2008b; 2009) to being more
efficient (Assaf 2009). Meanwhile, in Taiwan, small airports were originally found to
be more efficient (Yu 2004), but by 2001, the efficiency levels were similar for all
airport sizes (Wang et al. 2004). Only Spain has exhibited a consistent trend. Spain’s
larger airports continued to be the most efficient airport size throughout the period
studied. However, all of these studies used data sets from similar years.
2.4.2 CROSS COUNTRY RESEARCH

This section addresses those studies in airport efficiency that were conducted in
multiple countries. Details of these studies can be found in Table 2.6. One of the key
features in these studies of airport efficiency was ownership structure. Like other
studies on the empirical evidence of ownership, the question of whether privatisation
increases the efficiency of airports was still inconclusive in these studies. Pels et al.
(2001) separated airport operation into landside and airside operations and developed
separate DEA models to evaluate their productive efficiency. Their results indicated
that most private airports achieve higher levels of efficiency as compared to their non-
privatised counterparts. More recently, Vasigh and Gorjidooz (2006) measured the
effects of ownership on airport TFP for a sample of 24 airports from the UK (seven of
which were private), other European countries (seven public-private) and the US (eight
public). The results revealed that there was no significant relationship between financial
and operational efficiencies. The same result was achieved by Lin and Hong (2006),
who used DEA to assess airport operating efficiency.
The various studies published by Oum, T. H. also portrayed a varied response to the
question of ownership, suggesting a potential evolution over time. For example, the
results of Oum et al. (2003) and Oum and Yu (2004), using data from 1999 to 2001,
concluded that ownership had no statistically significant impact on productivity.
However, in studies that were published in 2006 and 2008, the results showed that
majority ownership by the private sector brings about higher efficiency as compared to
public sector ownership (Oum et al. 2006; 2008).
30
Ch2
Table 2.5: Single country research

Studies Sample Findings
US
Terminal efficiency in 11 airports improved while movement efficiency in 12

Gillen and Lall 21 US airports
airports improved.
(1997) (1989~1993)
Large airports are more efficient than small and medium airports.
Sarkis 44 US airports The sample saw a 5.5% improvement in average efficiency.

(2000) (1990~1994) Large airports are more efficient than small and medium airports.
Gillen and Lall 22 US airports TFP growth of -0.1% per year in the terminal side.
(2001) (1989~1993) TFP growth of -0.1% per year in the movement side.
Bazargan and
45 US airports Efficiency for large and medium hub airports is not statistically different. Overall,
Vasigh
(1996~2000) small airports are more efficient than large airports.
(2003)
Sarkis and Talluri 44 US airports
Among 44 airports, average efficiency increased from 0.681 to 0.737.
(2004) (1990~1994)
Pathomsiri et al. 56 US airports The number of efficient airports increased from 4 to 28. Overall, small airports
(2008) (2000~2003) are more efficient than medium and large airports.
Spain
Murillo-Melchor 33 Spanish airports Average efficiency was 0.6141, with little difference over the time period
(1999) (1992~1994) analysed.
With constant returns to scale, average efficiency was 0.6, and 8 airports are
Martin and
37 Spanish airports relatively efficient. With variable returns to scale, average efficiency was 0.7, and
Roman
(1997) 13 airports are relatively efficient. Large airports such as Madrid and Barcelona
(2001)
are more efficient than others.
Martin-Cejas 40 Spanish airports
Small and large airports achieve higher efficiency.
(2002) (1996~1997)
Martin and
34 Spanish airports
Roman Large airports achieve higher efficiency.
(1997)
(2006)
Martin et al. 37 Spanish airports
The larger airports are more efficient.
(2009) (1991~1997)
Tovar et al. 26 Spanish airports
TFP growth at 0.9% per year. Offshore airports above average efficiency.
(2010) (1993~1999)
Lozano and
41 Spanish airports
Gutierrez Half of the airports were found to be technically efficient.
(2006)
(2011)
UK
Parker 32 UK airports
Efficiency before privatisation was 0.988, after privatisation was 0.931.
(1999) (1979/80~1995/96)
Barros 27 UK airports The most efficient airport is Luton and the largest airports
(2008b) (2000/01~2004/05) (i.e. Heathrow, Gatwick, and Manchester) are the weakest.
Barros 27 UK airports
Luton airport is the most efficient airport, and Heathrow is the least efficient.
(2009) (2000~2006)
Assaf 27 UK airports
Large airports are more efficient than small airports.
(2009) (2007)
Taiwan
Wang et al. 10 Taiwan Airports Among three different size airports, the efficiency is almost the same. (Large:
(2004) (2001) 0.461, Medium: 0.457, and Small: 0.461).
Yu 14 Taiwan Airports Small airports can achieve a higher level of efficiency as compared to large
(2004) (1994~2000) airports.
Yu et al. 4 Taiwan airports
Airport efficiency is increasing yearly.
(2008) (1995~1999)
Yu 15 Taiwan airports
Offshore airports are more efficient than mainland airports.
(2010) (2006)
Note: The study by Hamazee and Vasigh (2000) is not included for the US because they established a revenue and cost model for
current US airport authorities but did not apply it to empirical data.
31
Ch2
Table 2.6: Cross country research

Studies Sample Findings
Pels et al. 34 European airports Most private airports achieve higher efficiency
(2001) (1995~1997) than public airports.
Larger airports achieve higher gross TFP

because of the economies of scale in airport
50 airports around
Oum et al. operations. An airport’s ownership structure
the world
(2003) does not appear to have any statistically
(1999)
significant effect on its productivity
performance.
Larger airports are more efficient due to

76 airports around
Oum and Yu economies of scale. There is no statistical
the world
(2004) significance in the difference between different
(2001-2002)
ownership structures.
The form of ownership and the size of an airport

are not apparently correlated with operational
performance of airports. In contrast, the
20 airports around
Lin and Hong existence of a hub airport, the location of the
the world
(2006) airport, and the economic growth rate of the
(2003)
country in which the airport is located are all
related to the operational performance of
airports.
Airports with government majority ownership

116 airports around
Oum et al. and those owned by multi-levels of government
the world
(2006) are significantly less efficient than airports with
(2003~2005)
a private majority ownership.
In every year, the performance of airports in the

Vashigh and 22 US and European
US was better than airports in the UK and EU,
Gorjidooz airports
but there was no obvious difference in terms of
(2006) (2000~2004)
ownership.
100% privately owned airports perform better

109 airports around
Oum et al. than public airports. Privatisation of one or more
the world
(2008) airports in cities with multiple airports would
(2007)
improve the efficiency of all airports.
Technical, scale and mix efficiencies are high

Lam et al. 11 airports in Asia Pacific among the major Asia Pacific airports. Between
(2009) (2001-2005) these eleven airports, Brisbane achieves relative
efficiency in every year under consideration.
Airports should focus more on investment than

on human resources. In addition, the inefficiency
Yang 12 airports in Asia Pacific
effects associated with the production functions
(2010) (1998~2006)
of airports increased over the investigated
period.
32
Ch2
2.5 SUMMARY
This chapter examined the current studies on airport efficiency. After reviewing the
majority of published airport benchmarking studies, this literature survey showed that
the analysis of performance evaluation in the airport sector has enjoyed significant
contributions in recent years. For the period under consideration, airport research has
shown a growth in terms of the number of research areas, and a number of analysis
techniques from other disciplines have been used to meet the research demand, derived
from the complex phenomena taking place in the airport industry. It is also
characterised by a number of dominating features, including a focus on publication in
two journals and a focus on airports in the most important regions for air transport.
The potential analysis techniques which can be used in the research are also widely
described. A few important methodological points emerge suggesting that the DEA
approach has been the methodology that has been traditionally used to reflect the multi-
production nature of the airport sector, although SFA and TFP have also seen limited
use.
A wide variety of input variables have been used with regard to the measurement of
labour, capital and the inclusion of material and outsourcing. Meanwhile, a number of
studies have only used the output variable of aeronautical activities although the
commercial side has recently received more attention. Other studies have combined the
output variables of passengers and cargo to movement as a single measure, effectively
treating them equally. From this analysis, a principle of variables selection for this
research has been confirmed.
A structure of research themes used in airport research has been constructed in this
literature review, and theoretical bases and disciplinary characteristics have been
identified. This chapter has defined the positioning of this thesis in airport research with
regard to these features of airport research. The further concepts of this thesis are
described in the next chapter.
33
Ch 3
Devolution of airport ownership and governance
CHAPTER 3
DEVOLUTION OF
AIRPORT OWNERSHIP AND GOVERNANCE
The aims of reviewing the changing nature of airports in this chapter are to understand
the nature of airport ownership and governance, to illustrate the evolution of airport
ownership and governance, and to justify the differences existing between diverse areas.
To achieve these aims, first how an airport is constructed is described. Then an
investigation is conducted into the periodic changes that have occurred in airport
ownership in order to enable an understanding of the current positioning of this period
in the periodic timeline.
34
Ch 3
3.1 AIRPORT OWNERSHIP AND GOVERNANCE

From previous experience with the UK and US, the periodicity of airport ownership
and governance evolution can be illustrated in three periods: governmental control,
corporatisation and commercialisation, and privatisation.
3.1.1 THE PERIOD OF GOVERNMENTAL CONTROL

In this period, airport ownership patterns evolved unevenly over time and space for a
variety of reasons. A similar evolution is revealed when reviewing the formation of
airports in the US and UK. During World War II, the governments of these two
countries spent significant resources to repair and construct airports for military service.
In 1947, the UK government transferred the ownership and management of 44 airports
to the Ministry of Civil Aviation (Humphreys 1999). In addition, in the US, more than
500 airports were declared to be military surplus and were subsequently handed over to
the authorities of cities, counties, and states for strictly civilian aviation use (Wells and
Young 2004). During this time, all airports were typically owned by the public sector
and used by political bodies as a fundamental tool in the practice of establishing and
reinforcing their citizens’ consensus (Jarach 2005). Graham (2008) found that during
this period, airport ownership could be categorised into three main patterns.
(1) Owned and operated by a national government
These airports served major cities (such as Tokyo, Bangkok, Paris, London, and
Singapore) and were usually operated by national governments.
(2) Owned and operated by local governments
These were regional airports that were usually operated by local governments (at a
regional or municipal level). Most of the airports in the US and some regional
airports in the UK followed this pattern.
(3) Owned and operated by a multi-level public sector
A number of airports were controlled by both the local and national government.
For example, Munich airport, which was founded in 1949, was owned by three
shareholders: the Free State of Bavaria (51%), the Federal Republic of Germany
(26%) and the City of Munich (23%) (Wragg 2009).
35
Ch 3
3.1.2 THE PERIOD OF CORPORATISATION AND COMMERCIALISATION

Between the 1970s and 1980s, the first steps towards airline privatisation and
deregulation took place as the air transport industry grew and matured, and at the same
time the philosophies of airport management began to change (Graham 2008). In this
period, a trend labelled corporatisation and commercialisation was developed and
introduced by several acts of the US Congress that related to the air transport industry
at this time1. After those Acts were implemented, the passage of the American Air
Cargo Deregulation Act of 1976 and the Airline Deregulation Act of 1978 implied that
the US air transport industry was growing and maturing. However, U.S. domestic
deregulation spilled over into international air transport (Doganis 2002). In Europe,
there were similar winds of change, such as in the UK, where the civil aviation
authority became increasingly liberal in its licensing decisions from 1975 onwards
(Doganis 2002). Consequently, many airports gradually started to be considered much
more as commercial enterprises, and a more businesslike management philosophy was
adopted. Therefore, commercialisation of the airport industry had started to take place.
However, most countries at this time had some sort of restrictions on public-sector
engagement in commercial activities. Therefore, before moving towards
commercialisation, airport corporatisation first took place.
There are many different interpretations of what exactly is meant by the terms
corporatisation and commercialisation. Generally, corporatisation is defined as an
attempt to introduce the rigours and philosophies of private industry while the
government retains control and ownership (Graham 2011). Meanwhile, airport
corporatisation may be defined as the establishment of a legal and independent airport
company wholly owned and controlled by the government or local authorities
(Shearman 1992; Graham 2008). Airport commercialisation can be defined as the
transformation of an airport from a public utility to a commercial enterprise with the
adoption of more business-like management philosophies, values, and approaches while
the airport remains publicly owned (Humphreys 2002a). The drivers of
commercialisation are determinant on the need for investment expertise and resources
that are not available or accessible to the public sector. Moves toward
1
The Department of Transportation (DOT) was established in 1967; in 1970, the Airport and Airway
Development Act and the Airport and Airway Revenue Act were signed; in 1976, the Airport and
Airway Development Act Amendments was implemented
36
Ch 3
commercialisation can be recognised through a number of different inter-related

developments. Firstly, a number of airports achieved a degree of autonomy by
establishing more independent airport authorities with public sector shareholding.
Secondly, many airports moved towards becoming profit-orientated self-financing
entities by addressing non-aeronautical activities, such as retail and concession
management. Commercialisation indicates a greater emphasis on financial rather than
operational issues. Thirdly, airport commercialisation is typified by an improvement in
customer focus with improvements to airline facilities, commercial areas, and landside
amenities and infrastructure (Doganis 1992; Graham 2008; Starkie 2008). The most
significant changes of how airports were operated in this period are described in the
following points:
(1) Management organisation structure

Traditionally, airport management structures have reflected a functional approach in
regard to dividing up responsibilities and lines of authority (Jarach 2005). They also
provide a framework within which management functions can be carried out. Usually,
these include operations, administration, engineering, finance and personnel or safety.
An example of this type of organisation drawn from an Asian airport is shown in Figure
3.1, which shows the formal airport authority relationship between superiors and
subordinates at various levels, as well as the formal channels of communication within
the organisation in the past.
After corporatisation, in order to increase non-aeronautical revenue, an airport must be

organised in such a way that commercial activities are given the importance they
deserve, which includes the need to be a clearly identified focus of responsibility for
commercial activities. The airport must be organised into different business areas with a
senior manager responsible for each, rather than being organised along functional
divisions. An airport following the commercial airport model is a business as well as a
provider of services. Each of the businesses then either retains the functional skills
which they require within their own department or draws upon skills as necessary from
separate service departments or from outside the airport organisation. An example of
modern airport organisation is shown in Figure 3.2.
37
Ch 3
Figure 3.1: Traditional airport organisation
Director
Deputy Director Deputy Director
Maintenance section
Civil service ethics
Accounting office
Planning section
Business service
Personnel office
Flight operation
Central control
General affairs
section
section
centre
office
Office car Fire

service brigade
centre
Source: Wensveen (2007).
Figure 3.2: Commercially oriented airport organisation

Chairman
Audit office Secretariat

Board of directors
Chief Executive Officer (CEO)
Executive vice Senior vice Senior vice Chief

president president president Engineer
Operation safety Department
Flight operation Department
General affairs Department

Maintenance Department
Accounting Department
Engineering Department
Planning and Marketing
Public Affairs Division
Refuelling Department
Legal affairs division

Business Department
Finance Department
Procurement Centre
Ethics Department
Cargo Department
Human resources
IT Department
Department
Department
Source: Wensveen (2007).
Airport organisations have become more and more specific, and airports are now
operated in a dynamic environment and as entities that continually adapt to changing
conditions. Consequently, some old positions might no longer be required, or they may
be merged with other departments while some new positions, such as the public affairs
division, might have to be created in order that new objectives can be reached (Wragg
2009). After corporatisation, airport organisations structures can be revised and updated
periodically more easily and are more flexible with regard to reflecting these changing
conditions. In some airports, the typical functional organisational structure with
38
Ch 3
different departments for finance, operations, administration, and so on, was replaced
with departments or business units that were more focused on their customers’ needs
(such as airline or passenger services) because the commercial functions of the airports
were gradually recognised as being equally important. Therefore, the resources and
staff numbers employed in these areas were expanded. In addition, the benchmarking of
financial performance and quality management techniques also began to be accepted by
a growing number of airports as essential management tools in this period (Humphreys
1999).
(2) Revenue generation

Airport revenue is usually classified into two main categories: aeronautical and non-
aeronautical revenues. Aeronautical revenues are those sources of income which arise
directly from the operation of aircraft and the processing of passengers and freight.
Non-aeronautical revenues are those generated by activities that are not directly related
to the operation of aircraft, notably income from commercial activities within the
terminal and rents for terminal space and airport land (Graham 2008). The International
Civil Aviation Organisation (ICAO) has a long list of activities that can be classified as
non-aeronautical revenue or aeronautical revenue (see Table 3.1).
Table 3.1: Activities of non-aeronautical revenue and aeronautical revenue

Aeronautical Revenues Non-Aeronautical Revenues
Ø Landing charges (including lighting and Ø Aviation fuel and oil concessions (including
approaching, and aerodrome control charges) throughput charges)
Ø Passenger service charges Ø Restaurants, bars, cafeterias and catering
Ø Cargo charges services
Ø Parking and hangar charges Ø Duty-free shops
Ø Security charges Ø Automobile parking
Ø Noise-related charges Ø Other concession and commercial activities
Ø Other charges on air traffic operations operated by the airport.
Ø Ground-handling charges2 Ø Rentals
Ø Other revenues from non- Aeronautical
activities
Source: ICAO (2006).
The evidence to show that airport authorities put emphasis on revenue generation can
be found in Chapter 2. From Table 2.4, for the purpose of evaluating airport efficiency,
2
Some airports have already outsourced ground-handling service, and at the majority of airports, this
service is largely carried out by one or more airlines or special ground-handling enterprises. In some
cases, the airport will impose concession and/or rental fees which are recorded as revenues from non-
aeronautical activities.
39
Ch 3
the amount of both aeronautical revenue and non-aeronautical revenue were adopted by
26 different papers, and most of them were published after the year 2000. Another
significant characteristic of the trend towards commercialisation and an increased focus
on treating airports as businesses was greater reliance being placed on non-aeronautical
or commercial revenues (Graham 2006). Aeronautical revenues, such as landing and
passenger fees from the airlines, had traditionally been the most important source. In
1970s, aircraft landing fees represented by far the most significant part of aeronautical
revenues from passenger-related charges (Doganis 1992). For a number of airports,
especially in Europe, non-aeronautical sources overtook aeronautical sources as being
the most important source of revenue.
A breakdown of revenues for a sample of European airports is shown in Table 3.2 for
the period between 1983 and 2009. The main watershed in this table happened in
1998/99 with the rise in the importance of aeronautical revenues concurrent with a
subsequent increase in reliance on non-aeronautical sources. This reflects not only
pressure from airlines and regulatory bodies to keep airport charge increases to a
minimum, but it also reflects the increased focus being placed on commercial activities.
This development was primarily the result of greater space being allocated to retail and
other non-aeronautical facilities, the quality being improved and the range of
commercial activities being expanded. Not only in the Europe, according to the
statistics which were published by Airport Council International (ACI) in 2011, in 2010,
non-aeronautical revenue also became the majority of total revenue in some other
regions such as Africa/the Middle East ( 55%), Asia/Pacific (52%), and North America
(55%).
Table 3.2: Average revenue and cost structures at European airports

1983 1988 1993 1998 2003 2006 2007 2008
/84 /89 /94 /99 /04 /07 /08 /09
Revenues shares (%)
Aeronautical 59 56 54 50 51 52 49 48
Non-
41 44 46 50 49 48 51 52
Aeronautical
Total 100 100 100 100 100 100 100 100
Source: Graham (2008); ATRS annual report (2010).
40
Ch 3
(3) Airport marketing

The airport industry historically had played a passive role towards marketing, and it
responded to customer needs only when necessary (Jarach 2005). Because of
commercialisation and the deregulation of the airline industry, the competition between
airports has gradually increased over the last three decades. Airport competition can be
considered on two different levels: competition between airport groups and competition
within airport groups. In some major urban areas or cities, there are a number of
situations when more than one airport serves the population. Notable examples are the
European cities of London and Paris, and the Asian cities of Tokyo and Shanghai. In
many cases when there are overlapping areas, one airport tends to become the dominant
player in a preferred location with the other airports playing a more secondary role.
However, when airports are operated as a system, such as Tokyo Narita and Haneda
airport, or as a group, such as London Heathrow and Gatwick airport, which were also
supported by Luton and Standsted before 2009, there is an important issue as to
whether this inhibits competition (Forsyth 2006).
However, in such a turbulent environment, the development and management of a

supplier and customer relationship is of primary strategic importance. A more
businesslike approach to airport management should be coupled with a more
commercially driven and competitive airline industry. This encourages airports to take
more active and positive roles. Modern airports have to undertake this kind of
recruitment because the modern airline industry, which has been transformed in many
places from a regulated and public sector controlled activity into a liberalised and
commercially orientated business, has played a major role in this changing airport
situation (Graham 2008). In addition, some airline developments (such as the formation
of global alliances) have been particularly important, as has been the development of
the low-cost sector, with regard to creating new views on airport competition. Airport
competition is a complex area to examine because there are different aspects which can
be considered (Graham 2006). In order to cope with this, another important issue is the
role of marketing. In the UK, for example, most of the airports during this
corporatisation and commercialisation period developed marketing departments which
started to use pricing tactics and promotional campaigns to attract new customers
(airline or retail companies) and which began to undertake market research (Humphreys
41
Ch 3
1999). In other countries, this phenomenon also occurred in those airports which had
been corporatised.
3.1.3 THE PERIOD OF PRIVATISATION

The literature on agency theory and strategic management suggest that ownership
influences a firm’s performance because different owners pursue different goals and
have different incentives (Vickers and Yarrow 1998; Oum et al. 2006). Under
government ownership and management, a firm is operated by bureaucrats whose
objective function is a weighted average of social welfare and their own personal
agendas. Under private ownership, by contrast, the firm maximises profit (i.e. the
shareholders’ value) (Cullinane et al. 2011). A recent common-sense view is that
government-owned firms are less efficient than their private sector counterparts
operating in similar situations. Consequently, the effects of ownership on a firm’s
productive efficiency have been an important topic of research. The 1990s were a
decade when airport privatisation in western countries became a reality. Privatisation of
an airport may be defined simply as the whole or part moves from public to private
ownership, with substantial involvement of private sector management and operation
(Humphreys 1999). Another common-sense view suggests that privatisation reduces the
need for public sector investment and that free access to commercial markets will bring
improved efficiency, greater competition and a wider share of ownership (Graham
2009).
A number of reasons why these governments sought to devolve responsibility for

airport ownership at this time are as follows: (Parker 1999; Humphreys and Francis
2002a; Graham 2005; Oum et al. 2008).
(1) It reduces dependency on government resources.
(2) Airport expansion increases catchments and influence.
(3) A publicly owned airport is unable to attract finances or to attract market
investment.
(4) It focuses on customer requirements.
(5) It focuses on policy and regulations.
According to Oum et al. (2006) and Gillen (2011), ownership and governance form can
be classified into eight different categories, which are as follows:
(1) Government owned and operated (e.g. Finland, and some airports in the US).
42
Ch 3
(2) Mixed private–government ownership, with the private sector owning a majority
share (e.g. Denmark, Austria, and Switzerland).
(3) Mixed government–private ownership, with the government owning a majority
share (e.g. Hamburg, France, China, and Kansai-Japan).
(4) Government ownership but contracted out to an airport authority under a long term
lease (e.g. Chile, Hamilton and some airports in the US).
(5) Multi-level governments who form an authority to own and operate airports in the
region (e.g. some airports in the UK).
(6) 100% government corporation ownership and operation (e.g. Singapore, Hong
Kong, and Taiwan).
(7) Fully private ownership (e.g. BAA).
(8) Independent non-profit corporations (e.g. Canada).
This research, in order to include the all possibilities, the sample airports, which are
selected in this research, should try to cover these eight types of airport ownership. In
addition, privatisation does come along with some hazards. The theoretical arguments
for and against privatisation of publicly owned organisations, particularly when a share
flotation is being considered, are well known. For example, it may create a private
monopoly which overcharges, may deliver poor standards of service, may invest
inadequately and may give insufficient consideration to externalities and other
disadvantages (Beesley 1997). This may also happen in the airport industry. Therefore,
when privatising airports, the government usually set up several regulations or
regulatory bodies to manage private airports.
3.2. AIRPORT OWNERSHIP EVOLUTION IN DIFFERENT AREAS

The evolution of airport ownership in North America, Europe, and Asia has been
experienced in different periods of time. In order to understand the differences in the
main countries on different continents, the airport ownership evolution of The US,
which was the first country to implement airport commercialisation, the UK, which was
the first country to undertake fully airport privatisation policy, and China, which will be
the most important country in the air transport industry in the next 20 years, are
described in the following subsections.
43
Ch 3
3.2.1. NORTH AMERICA: THE UNITED STATES OF AMERICA

In general, airports in the US are almost the most privatised in the world, despite the
fact that all of the major commercial airports are owned by government entities.
Compared to airports elsewhere in the world and even to the airports in the countries
that have recently privatised their airports, the major US airports have experienced an
extensive degree of private control over virtually every aspect of airport planning,
design, finance, operations, pricing and access, but not ownership (Neufville 1999).
The high level of private participation in the management and strategic development of
major commercial airports in the US is summarised in Table 3.3. While the degree of
involvement of private companies in the control of airports varies widely from state to
state and from city to city, the overall situation is that major American commercial
airports are run through a form of partnership between the federal government, local
civic interests, and private companies.
Table 3.3: Level of privatisation for major commercial US airports

Elements of Control Typical Status Details
Government leads, but private interests influence

the decision through their willingness to provide
Planning for expansion Government/Private
financing or to accept plans under majority-in-
interest clauses in leases.
Airlines and users have significant, often

Design of projects Largely Private decisive, control over design. Private consultants
typically execute designs.
Mix of public (mostly federal) and private

Financing Largely Private sources, with a significant fraction of money
coming from bonds issued in capital markets.
Operation of facilities largely done by airlines

Operation of facilities Largely Private
and other third parties.
Price is set in negotiation with major users and is

Pricing of services Government/Private
subject to legal controls on increases.
Principle of open access to all qualified users

Availability of services Government/Private generally holds, subject to airlines and other
third parties controlling the use of their facilities.
Ownership of properties Government Municipal or regional agencies.

Source: Neufville (1999).
44
Ch 3
Public airports in the US are owned and operated under a variety of organisational and
jurisdictional arrangements. Commercial airports might be owned and operated by a
city, county, or state; by the federal government or may be owned by more than one
jurisdiction (e.g. a city and a county). In some cases, a commercial airport is owned by
one or more of these governmental entities but operated by a separate public body, such
as an airport authority that is specifically created for the purpose of managing the
airport. Regardless of ownership, the legal responsibility for the airport’s day-to-day
operation and administration can be vested in any of five kinds of governmental or
public entities, which include a municipal or county government, a multipurpose port
authority, an airport authority, a state government, or the federal government (Wells
and Young 2004).
In the US, airport privatisation typically involves the lease of airport property and/or
facilities to a private company to build, operate, and/or manage commercial services
offered at the airport. However, no commercial airport property has been completely
sold to a private entity. From a service perspective, although no US commercial airport
has been sold to a private entity, many publicly-owned airports have extensive private
sector involvement. Most of the services that are now performed at large commercial
airports (such as airline ticketing, baggage handling, cleaning, retail concessions, and
ground transportation) are provided by private firms (Wragg 2009). Some estimates
indicate that as many as 90% of the people working at the nation’s largest airports are
employed by private firms. The remaining 10% of the employees include local and
state government personnel performing administrative or public safety duties, federal
employees, such as Federal Aviation Administration (FAA) air traffic controllers and
Transportation Security Agency (TSA) security screeners, or other public employees,
which are made up primarily of military personnel (FAA 2011). From a financial
perspective, many airports in the US are now relying more on private financing for
capital development. Airports have sought to diversify their sources of capital
development funding, including the amount of private sector financing. Several reasons
have motivated this interest in expanding the role of the private sector at commercial
airports in the US (Wells and Young 2004). Firstly, privatisation advocates believe that
private firms will provide additional capital for development. Secondly, proponents
believe that privatised airports will be more profitable because the private sector will
operate them more efficiently. Lastly, advocates believe that privatisation will
45
Ch 3
financially benefit all levels of government by reducing demand on public funds and by
increasing the tax base.
Since 1997, the FAA has implemented the Pilot Programme on Private Ownership of
Airports. Under this programme, five public-use airports are operated under a private
management group. The airports selected to participate in the program include Stewart
International Airport in Newburgh, New York; Brown Field in San Diego, California;
Rafael Hernandez Airport in Aguadilla, Puerto Rico; New Orleans Lakefront Airport in
New Orleans, Louisiana; and Niagara Falls International Airport in Niagara Falls, New
York. However, so far, this programme has been met with limited success, with only
Stewart International Airport fully completing the privatisation process (Wells and
Yang. 2004). However, in 2011, only three airports (Puerto Rico’s Luis Munoz Marin
International Airport, Briscoe Field in Gwinnett County, and Hendry County Airglades
Airport) in the entire US have active applications in the privatisation program (Assaf
and Gillen 2012). The enthusiasm toward full airport privatisation has appeared to wane
since the late 1990s, as the overall economy of the US has declined. As mentioned
above, however, the overall progress has not been very successful as compared with
progress in the UK. However, the concepts that drive private enterprises toward
competitive and efficient operations are becoming embraced by publicly owned and
managed airports. Consequently, more efficient organisational structures and
management responsibilities have resulted in more streamlined and efficient airport
management organisational structures.
3.2.2. EUROPE: THE UNITED KINGDOM

The UK was the first country to embark on a path of full airport privatisation following
the introduction of the 1986 Airport Act. Before this, airports in the UK had depended
on subsidies from the UK tax payers. Until the mid-1980s, UK airports were regarded
as public utilities to be owned and subsidised by the government. Since then there has
been a significant shift and most airports in the UK are now full or partly funded by the
private sector.
The trend towards airport privatisation began in the UK around 25 years ago. The
consequences of this process provide an important case study for policy makers and
practitioners worldwide as they seek to assess whether or not to commercialise,
46
Ch 3
privatise or retain ownership of their airports. Until 1987, most of the airports in the UK
were owned by either the central or local governments. The 1986 Airports Act
privatised the airports of the British Airports Authority (BAA) and transformed UK
municipal airports into commercial companies. Since then, the pattern of airport
ownership has evolved unevenly over time as airports have been commercialised and
privatised in a variety of forms (Humphreys 1999). The aims of privatisation in the UK
were to improve efficiency, reduce government involvement in the industry, reduce
subsidies to the public sector, reduce the financial burden on government of the Public
Sector Borrowing Rate (PSBR), provide access to private investment, widen share
ownership, gain political advantage, and introduce commercially focused management
(Morgan 1995). These reasons are similar to those of the US.
Before privatisation, most UK major airports were operated by the BAA, which was
established by the passing of the Airport Authority Act 1966 to take responsibility for
four state-owned airports. In the next few years, the authority acquired responsibility
for Glasgow airport, Edinburgh airport Aberdeen airport and Southampton airport. As
part of Margaret Thatcher's moves to privatise government-owned assets, the Airports
Act (1986) mandated the creation of the BAA plc as a vehicle by which stock market
funds could be raised. The initial capitalisation of the BAA plc was £1,225 million. At
the time of privatisation, all of BAA’s issued share capital was sold by the government,
except for a retained special golden preference share3 (which still exists) (Parker 1999).
The state corporation was privatised without restructuring on the grounds that a unified
company would have the financial resources to fund future investment needs. The main
impact of privatisation was not, therefore, in the product market but was rather in the
capital market. The BAA became subject to pressure from a threat of takeover by
another company that identified possible efficiency gains. At the same time, the
continued existence of the government’s golden share in BAA may have reduced the
takeover threat.
More recently, the BAA has expanded into international operations, including retail
contracts at Boston Logan International Airport and Baltimore-Washington
3
This share is often retained only for some defined period of time to allow a newly privatised company
to become accustomed to operating in a public environment, unless ownership of the organization
concerned is deemed to be of ongoing importance to national interests, for example for reasons of
international security (Parker 1999).
47
Ch 3
International Thurgood Marshall Airport (through a subsidiary called BAA USA, Inc.),
and a total management contract with the City of Indianapolis to exclusively run the
Indianapolis International Airport (as BAA Indianapolis, Inc.) (BAA 2011). In 2005,
BAA took a 75% stake in Budapest Ferihegy, the largest airport in Hungary, which was
being privatised by the Hungarian government. In 2007, the decision was made to sell
the stake in Ferihegy, which was done when a consortium led by HOCHTIEF AirPort
of Germany purchased the stake.
In 2006, BAA was taken over by a consortium led by Grupo Ferrovial. Consequently,
the company was delisted from the London Stock Exchange in 2006, and the company
name was subsequently changed from BAA plc to BAA Limited. In 2008, Gatwick
airport was put up for sale. In October 2009, BAA announced that Gatwick had been
bought by Global Infrastructure Partners (GIP). (Grupo Ferrovial 2007; BAA 2011;
Gatwick 2011). Furthermore, in March 2009, the UK Competition Commission ruled
that BAA must sell Stansted within two years to either Glasgow or Edinburgh airport.
In 2012, BAA announced the sale of Edinburgh Airport to GIP, and after losing a case
in the Court of Appeal, BAA announced they would sell Stansted in the near future
(BAA 2012). This brief summary of the history of BAA shows that the airport
ownership can be transferred to the capital market easily.
The second part of the Airports Act (1986) required that all airports with a turnover of
more than £1 million in two of the previous three years become companies. Prior to this,
airports had been run directly by their local government owners. Under this condition,
16 airports were covered by this part of the Act (Graham 2005). However, the most far-
reaching impact of the Airports Act (1986) was to place airports under an ownership
structure that enabled local authorities to sell their shares and become fully privatised
companies. Although the UK government has never directly forced airports to privatise,
the reduction of the money available for public sector borrowing since 1992/3 has
forced most airports to seek private capital to finance expansion (Humphreys 1999).
There is no doubt that private companies always emphasise their profits. Therefore, the
introduction of various commercialised forms of ownership (including in some cases
the full privatisation of airports) has led many airport managements to increase their
focus on non-aeronautical sources of revenue (Humphreys and Francis 2002b).
48
Ch 3
The emphasis on commercial revenue has led to the increased development and
utilisation of revenue-generating space and the rapid development of airport sites with
business parks, hotels, freight facilities and maintenance facilities. However, these
changes (as increased retail outlets) can reduce terminal capacity to process passengers
(Humphreys and Francis 2002b).
Commercial pressures have also resulted in the pattern of airport ownership becoming
increasingly dynamic in the UK, with many significant changes taking place in the
ownership and governance of regional airports. Details of the changes at the original 16
airports that were commercialised under the Airports Act (1986) are discussed in
Humphreys (1999). In 1997, the Airports Act Part II introduced a mixed pattern of
ownership structures (see Figure 3.3). By 1997, four airports were fully privatised (i.e.
East Midlands, Bournemouth, Southend and Cardiff) and three airports entered a part
public, part private ownership structure (i.e. Birmingham, Bristol, Liverpool). The
remaining nine airports remained in public ownership (i.e. Manchester, Blackpool,
Norwich, Humberside, Leeds Bradford, Luton, Newcastle, Teesside and Exeter). Since
1997, the private sector has taken an increased role in UK airport ownership structures,
and along with this, the rate at which ownership has been transferred from one owner to
another has also increased (see Figure 3.4). Only two airports have remained with the
same owners. The predominant ownership structure has shifted towards public and
private partnerships, with seven airports adopting this structure. Five airports are fully
privately owned, while four have remained in public ownership. Detail are provided as
follows:
49
Ch 3
Figure 3.3: The ownership structure of UK airports in 1997
Source: Ison et al. (2011).
50
Ch 3
Figure 3.4: The ownership structure of UK airports in 2010
Source: Ison et al. (2011).
Table 3.4 shows the trend of ownership in the UK. As mentioned early, in the past 20
years, airport ownership has been transferred several times. Both East Midlands and
Bournemouth have moved from the private sector back into full public ownership (i.e.
Manchester Airport Group: MAG). Humberside has remained in public ownership, but
a majority of its shares are held by the MAG. Birmingham has stayed partly privatised
and Liverpool has stayed fully privatised, but the private owners have changed. Exeter
51
Ch 3
and Blackpool have moved from the public sector to the private sector. Meanwhile,
Norwich, Leeds Bradford, Durham Tees Valley, Newcastle and Luton have gone from
being publicly owned to being partly privately owned. Since 1987, only Manchester
airport has retained the same owners. In addition, the airports that were not part of the
original 16 have offered commercial services. In a further trend, all but two of the
airports are now partly owned by larger (in some cases international) airport groups. As
mentioned in the previous section, the privatisation of a public sector entity results in a
monopoly. The UK government has tried to take the role of regulator to prevent an
enterprise from abusing its position. The most common form of intervention has been
the regulation of the price an enterprise can charge for its products or services (Bishop
and Thompson 1992). However, the importance of whether or not ownership is public
or private may be misplaced. Some people have suggested that the nature of
competition and the form of regulation is more important than ownership in achieving
the economic aims of privatisation (Graham 2011). Privatisation has been successful in
some of its other aims. The amount of public money required to subsidise nationalised
industries has been drastically reduced; the strain on public sector borrowing has been
removed; access to private finance has been provided, and the role of government has
changed from owner/operator to regulator with the power to intervene in the public
interest. Privatisation was introduced in the UK to control the PSBR by addressing the
inefficiencies of loss-making public sector industries (Pirie 1985). Although the public
sector deficit had disappeared by 1987/8, privatisation was still pursued as a politically
attractive means to finance tax cuts without reducing public expenditures (Thompson
1990). Given the lack of a case for privatisation in terms of improved efficiency, it
appears that privatisation was pursued as an ideology by the UK government. The
model shows how a government can privatise swiftly and maintain political popularity.
How far then do municipal airports reflect these general trends?
In order to deal with airport privatisation, the UK government authority has attempted
to set up several regulations. The principal aspects of these regulations are airport
licensing and safety, economic regulations, international obligations, traffic regulation,
aviation security and noise. This section looks at the framework for economic
regulation of airports. The regulatory system aims to provide safeguards against
distortion of the air travel market through predatory pricing or other monopoly abuses
by airport operators. It also aims to incentivise cost control and efficiency (Gillen 2011).
52
Table 3.4: The changing of UK airport ownership
1980s 1990s 2000s Ownership
Airport transfer
86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12
Southampton SOU Mr Somer PdS BAAplc Ferrovial 3
Aberdeen ABZ PS BAA plc Ferrovial 2
Edinburgh EDI PS BAA plc Ferrovial GIP 3 From 04/2012
Global
Gatwick LGW PS BAA plc Ferrovial GIP 2 Infrastructure
Partners
Heathrow LHR PS BAA plc Ferrovial 2
Announced to
Standsted STN PS BAA plc Ferrovial 2 sell on 2012
Prestwick PIK PS BAA plc Prestwick Aviation Holdings Stagecoach Infratil 3
Liverpool LPL Public Sector British Aerospace (76%) Peel Airports (76%; 100% from 2001) 2
Manchester MAN Public Sector Manchester Airport Manchester Airport Group (MAG) public enterprise 0
Eastern Announced
Humberside HUY Public Sector Manchester Airport and from 2001MAG 2 on 08/2012
group
EastMidlands EMA Public Sector National Express MAG 2
Bournemouth BOH Public Sector National Express MAG 2
Cardiff CWL Public Sector TBI TBI (Abertis) 2
Belfast BHD Public Sector TBI TBI (Abertis) 2
30 years management contract with London Luton Airport
Luton LTN Public Sector London Luton Airport Operations Ltd (Abertis) 0
London Luton Airport Operations Ltd From 1998 Operations Ltd (TBI)
Birmingham BHX Public Sector EuroHub (Birmingham) Limited. (48.25%) 1
Bristol BRS Public Sector First Group (51%) MEIF1 (50%); Teachers’(49%)* 2
Newcastle NCL Public Sector Local government (51%) Copenhagen Airport (49%) 1
Durham Tees Vantage Airport
MME Public Sector Peel Airports (75%) 2
Valley Group (65%)
Balfour
Blackpool BLK Public Sector MAR Balfour Beatty 2 Beatty
Norwich NWI Public Sector Omnipot (80.1%) 1
Exeter EXT Public Sector RCA 1
Leeds
LBA Public Sector Bridgepoint Capital. 1
Bradford
Inverness INV Public Sector Highlands and Islands Airports Limited 0
Newquay NQY Public Sector 0
London City LCY Mowlem Dermot Desmod AIG & GIP GIP(75%) 3
Doncaster
DSA Peel Airports 0
Sheffield
* Bristol Airport is 50% owned by Macquarie European Infrastructure Fund 1 (MEIF 1), with approximately 49% held by Ontario Teachers’ Pension Plan (Teachers’).
53
Ch 3
At present, the CAA’s Economic Regulation Group regulates the 45 airports which
exceed the £1million turnover threshold under the Airports Act and the Airports
(Northern Ireland) Order 1994. The economic regulation of airports by the CAA dates
from the Airports Act (1986). The objectives of the CAA are to further the reasonable
interests of airport users, to promote the efficient, economic and profitable operation of
airports, to encourage investment in new airport facilities to satisfy anticipated user
demand, and to impose the minimum amount of regulation consistent with these duties.
When measuring the success of UK airport privatisation and commercialisation against

with other countries, it is clear that this has been largely successful in some points (such
as encouraging enterprise in the operation of major airports, air transport facilities
should not in general be subsidised by the taxpayer and should normally operate as
commercial undertakings). However, it is difficult to separate out the impact of
commercialisation from privatisation and it can be concluded that much could have
been achieved by commercialisation alone. In this the UK experience is very different
from the policies which are undertaken in the US. This research can help to find out
what kind of ownership can improve airport efficiency.
3.2.3. ASIA: CHINA

Although in 2011, China had 142 civilian airports, the market was skewed towards the
largest 10 airports, which together possess a 60% share of passenger volumes. Seven of
these are located along the eastern seaboard (i.e. Beijing, Shanghai Pudong, Shanghai
Hongqiao, Hangzhou, Guangzhou, Shenzhen, and Hainan). The pattern is similar for
cargo, with the five leading airports accounting for 64% of total volume (CAAC 2010).
These airports employ nearly 50,000 people and have assets worth approximately US$
4.8 billion (CAAC 2010). The majority of these airports remain fully government
owned through the CAAC and local government entities. In 1984, the Civil Aviation
Administration of China (CAAC) was established; prior to this, China had no
commercial aviation sector. However, it was not until 1988 that the first airport reform
was initiated. The subsequent process can be broadly divided chronologically into three
stages (Yang et al. 2008).
In the first stage, between 1988 and 1994, the primary objective was to reform airport
management. Previously, the Chinese government had controlled all airport activities
54
Ch 3
through its economic planning institutions. China’s economic reforms involved

restructuring state-owned enterprises and industries and injecting new impetus into the
economy. Through the implementation of the “Temporary Provision of Airport
Management” in 1989, airports were separated from airlines. Airports were then
defined under the ‘Measures to Change the Operational Mechanisms of State-owned
Aviation Enterprises’ as enterprises to protect airport operation from central
government control.
The second stage of reform, which lasted from 1995 to 2001, included the development
of airport joint-equity and further localisation reforms. At this time, attention turned to
the ownership structure of airports. Many airports began to operate along market lines
following the provision of the ‘Standardisation Management of Civil Aviation
Enterprises’ in 1997. Most airports underwent joint-equity reform, and airport
businesses activities were extended to capital operations. Several airports were listed on
either domestic or foreign stock markets. In a bid to stimulate enthusiasm and interest
among local governments in regard to development of the industry, devolution was
extended. A total of 35 airports were transferred their management to local
administrations during this period.
In the third stage, which began in 2002, the liberalisation of the airport industry
accelerated and blanket devolution was implemented. Most of China’s airports had
established some form of internal governance structure and had become joint-equity
enterprises. In 2002, the CAAC transferred ownership to the provincial government.
Management control over these jobs and assets largely passed to its provincial offices.
Although only a handful of airports were listed or had foreign investments, several
airports announced that they would welcome a strategic partner. A number of Chinese
airports were looking to foreign investors, not only to provide capital but also to
provide international management expertise.
The role of the CAAC has continued to evolve since 2002. Having once owned and
managed the entire aviation sector, the CAAC now discharges more administrative and
regulatory functions. Today its jurisdiction spans the following areas (Fung et al 2008):
55
Ch 3
(1) Central planning for airports

The CAAC has the power to approve construction or redevelopment of airports
facilities, including terminals and runways. It liaises with other government bodies in
the planning of supporting infrastructure.
(2) Setting domestic and international aviation tariffs

The CAAC’s control over tariffs means that it retains significant influence over their
largest source of revenues. This situation is expected to change and will affect the
future financial performance of Chinese airports.
(3) Administering airport construction fees

The CAAC collects these fees in a centrally administered fund and then disburses the
money back to the airports.
(4) Setting and monitoring standards

The CAAC formulates all standards governing safety, security and other operational
issues within airports.
(5) Airspace administration and air traffic control

The CAAC grants airlines the rights to use certain routes and administers air traffic
within China’s borders. These functions are likely to remain under the close supervision
of the CAAC. The CAAC is also involved in representing China in international
negotiations related to civil aviation airspace.

These reforms, which have developed since 1988, have brought about greater
opportunities and greater business interest in the sector. In turn, they have dramatically
changed the ownership pattern of China’s airports. In the last few years, ownership of
most airports in China has been transferred from the central government to local
authorities. The most profitable airports have been partially privatised and listed on the
stock market (e.g. Shanghai and Beijing airports) (Fung and Chow 2011). However,
most airports in China are still majority owned by the government and, unlike their
counter-parts in other countries, are still highly subject to government intervention in
their daily operations and management (in addition to regulatory requirements). The
56
Ch 3
following section describes current types of privatisation methods in China (Yang and
Hong 2010):
(1) Foreign investment

After the implementation of a localisation program focusing on ownership transfer from
the central government to local authorities, the CAAC has further allowed foreign
investors to take equity stakes in China’s airports. For example, in April 2005, the
Airport Authority Hong Kong (AAHK) agreed to invest a stake of 35% in Hangzhou
Xiao Shan International Airport (ranked number nine among Chinese airports by the
number of passengers handled). After AAHK’s investment in Hangzhou Airport,
airports in Ningbo, Nanjing, Chengdu and Kunming were reportedly negotiating with
foreign investors on stake sales. In 2005, German airport operator Fraport AG (which
manages Frankfurt Airport) signed a strategic partnership agreement to buy 25% of
Ningbo Lishe International Airport (Zhang 2008).
(2) Publicly listed company

The extent of airport privatisation in China has been relatively limited compared to that
of other countries, especially in Europe. The most common process of privatisation in
China has been to issue shares in the stock market to introduce private capital intended
to support the expansion and upgrade of airport facilities. In most cases, the local
government has remained a majority shareholder and is still in control of the board of
the airport company. Since 2000, six Chinese airports (i.e. Shenzhen, Shanghai,
Xiamen, Hainan, Beijing and Guangzhou) have been listed on the stock market.
(3) Public Private Partnership (PPP)

In contrast to the core business of passenger terminal management and aircraft handling
(i.e. the aeronautical part of aviation), the Chinese government has always been more
receptive to opening the market of the non-core aviation business (i.e. non-aeronautical)
to private operators, which is considered less essential (such as retail in passenger
terminals and ground handling services). Consequently, in China, airport assets and
property are usually managed by the airport company, which is 100% or majority
owned by the government, while the non-aeronautical part of the airport business is
now often contracted out to private companies (Zhang 2008). In China, due to the lack
of a legal framework in the management of concessions, the major types of PPP models
57
Ch 3
in commercial activities usually involve short-term sub-contracting of services and

mid-term leasing. For example, the retail spaces in Shanghai International Airport are
leased out to private operators, and their performance is reviewed regularly. Meanwhile,
the maintenance of their terminal facilities is contracted out. Shanghai International
Airport has also established a joint-venture company with Frankfurt Airport to provide
training to their airport employees (Fung and Chow 2011). These are examples of
attempts by the government to partially privatise the operations and maintenance of
airports, and they demonstrate that the Chinese government is continuing its effort to
develop a market economy. This is also the first step in the process of granting more
autonomy to state-owned enterprises.
(4) The airport corporations

The Chinese government has allowed mergers and acquisitions between airports in the
last few years. Consequently, several large airport corporations have been formed in
China to achieve economies of scale and the synergy by which to improve management
and financial strength. Although the size of most airport corporations in China is still
relatively small when compared with other international airport operators (e.g. BAA),
the creation of airport corporations managing more than one airport signifies the
Chinese government’s effort in promoting operational autonomy and a strategy to
achieve balanced developments between the regions (Gong et al. 2012). Capital
Airports Holding Company (CAH) can be considered a success story of airport mergers
and acquisitions in the Chinese airport industry. At the end of 2008, CAH was holding
stakes worth RMB 67 billion in more than 30 airports in China located in many parts of
the country (Yang and Hong 2010).
After 20 years of evolution, airport ownership in China has significantly diversified.

Some Chinese airports are built and owned by municipalities while others are owned
and controlled by provincial governments. Some have been sold or handed over to
airport groups while others have been incorporated and have become leading to cross-
region and cross-industry multiple owners. There have been domestic and international
Initial Public Offerings (IPOs) of several Chinese airports (Gong et al. 2012).
Meanwhile, foreign ownership by joint venture has started to be introduced. This
process of commercialisation and privatisation has gradually transformed airports into
financially self-sufficient and profit-making businesses. China’s airports are no longer
58
Ch 3
run as municipal facilities maintained by subsidies from central or local governments.

Table 3.5 shows the ownership structure of Chinese major airports.
Table 3.5: Major airports in China

Ownership Airports
All the airports in Tibet;

All the airports under Beijing Capital Group including
Central government ownership Capital Airport and Tianjin Airport;
All airports under Hubei, Jiangxi, Jilin, Guizhou, Chongqing
airport groups and Heilongjiang.
Central government controlled - Capital Airport*, Shenyang Airport and

mixed ownership Dalian Airport.
Trunk airports and all feeder airports in most provinces
(e.g. Shanxi, Shaanxi, Hunan,
Local government ownership
Henan, Yunnan, Harbin and Xinjiang,
Yinchuan and Qinghai Airports, etc).
Local government controlled Shanghai, Baiyun**, Xiamen*, Shenzhen*,
mixed ownership Hangzhou*, Zhuhai* , Meilan **
Source: Yang et al. (2008). Organised by the author.
Note:*Airport has foreign investment. ** Airport has airlines investment.
Reforms in China have transformed some airports from loss making entities that are
reliant on large public subsidies into profitable, customer-orientated businesses.
Airports have been able to diversify and to put more emphasis on expanding their non-
aeronautical activities (Yang et al. 2008). However, despite these changes, such as
foreign investment, publicly listed airport companies, airport corporatisation, and
public private partnership, China’s aviation industry still lags behind those of many
developed countries, and it continues to face a number of challenges (such as a low-cost
airline wave). In particular, institutional reform has been slow, and it has often not
supported larger policy objectives. Unbalanced development among different regions is
problematic, and many airport operations remain unprofitable. In the long-term, central
and local authorities may have to allow more diversification and commercialisation of
airport ownership structures and further reduce perceived commercial risk. At the same
time, the lack of transparent performance variables to gauge policy affects is hampering
objective assessment of reform. This research can provide an opportunity to determine
whether it is appropriate to undertake a Western style airport privatisation policy in
Eastern airports.
59
Ch 3
3.3. SUMMARY
The review of the changing nature of airports in this chapter has described the changing
patterns of airport ownership in three very different countries. The airport ownership
patterns in the periodicity of airport ownership and governance evolution that have
been described in this chapter (see Section 3.1) can help us to understand the current
airport ownership trends around the world. Figure 3.5 shows the evolution of airport
ownership structure among these three countries.
Figure 3.5: Evolution of airport ownership structure

Airport ownership evolution period
Corporatisation and
Government control Privatisation
Commercialisation
China
U.S.
U.K.
Airports in China are currently in the early stage of a second period of development,
due to the outsourced nature of airport operations and the government still retaining
ownership. Since the late 1990s, the Chinese government has embarked on a policy of
floating state-owned airlines and airports in the stock markets in order to improve their
efficiency and performance. Even after a localisation program which was started in
1988 and completed in 2003, among these 142 commercial airports, so far, only six
Chinese airport companies have been listed on stock exchanges in Hong Kong,
Shanghai and Shenzhen (Gong et al. 2012). However, the state still holds majority
ownership in these listed companies. When comparing the evolution in China with the
other two countries, the FAA in the US has only tried to privatise a few airports, and
most of the day-to-day operations in the majority of US airports have been
commercialised. Therefore, airport ownership evolution in the US could be said to be in
the early stage of privatisation. In addition, UK airports have been transferred between
owners several times. In this case, the evolution period should be classified as being in
the middle of privatisation.
60
Ch 3
Sample airports from different ownerships are selected for the purposes of this research
to answer the second Research Question 2. Therefore, after reviewing the changing
nature of airports in this chapter, we can see that US airport ownership structures are
very similar. Consequently, the sample airports should be selected from Europe and the
Asian-Pacific region.
61
Ch4
Research design and methodology
CHAPTER 4
RESEARCH DESIGN
AND METHODOLOGY
This chapter examines the methodological considerations on the understanding of

airport efficiency by using different analysis methods. It connects the previous chapters
and the following chapters, which develops the research structures and presents the
entire analytical process of this research. This chapter consists of eight sections. A
general discussion on research philosophies, approaches, and strategies is presented in
Section two. The data analysis methods that are used in this study are described
individually from sections three to six. The data collection methods that are adopted in
this study are described in Section seven. The research framework of this study is
discussed in the last section.
62
Ch4
4.1 RESEARCH DESIGN

Churchill (1976) compared research design to the architect’s blueprint for a house,
which is nothing more than the framework for research. The choice of research design
reflects the decision about the priority that is given to a range of dimensions of the
research process (Bryman and Bell 2011). In essence, a research design provides
guidance for the collection and analysis of the data in a study, which ensures the
relevance of the work to the proposed problem and employment of economical
procedures (Churchill and Iacobucci 2002). This will influence the choice of
methodology, the data collection method that is used, and the justifications of the
research outcomes. Following the development of these concepts, the next step is to
identify an appropriate research design to structuralise and conceptualise the newly
evolved disciplines of the research.
In academic research, methodology is defined as a body of knowledge that allows a

researcher to underpin the research questions through the use of various types of
evidence that can be gathered (Clark et al. 1984). According to Avison and Fitzgerald
(1995 p. 63):
“a methodology is a collection of procedures, techniques, tools, and documentation

aids…. But a methodology is more than merely a collection of these things. It is usually
based on some philosophical paradigm; otherwise it is merely a method, like a recipe.”
To identify the presuppositions and consequences of the applied procedures, a research

methodology is important to any study (Miller 1983). Näslund (2002) recommended
that the selection of a research method should be based on the research paradigm, or
question, due to the fundamental nature of the research processes, which are generally
involved with a particular research strategy or method. The following section presents
the philosophical position and approach of this research, which highlights the influence
of the research method selection. Saunders et al. (2009) described the research process
as being like an onion, where assumptions must be made at each individual stage of the
research approach, and where each of these stages is represented as a layer of an onion
(as showed in Figure 4.1). The research process that is used in this current study is
described from the outside to the inside of the onion.
63
Ch4
Figure 4.1: The research onion
Positivism
Experiment
Survey
Realism
Case
Cross-‐sectional Study
Deductive
Data collection
Grounded
Interpretivism
Theory
Longitudinal
Ethnography Inductive
Archival
Pragmatism
Time horizons
Techniques and Strategies Approaches Philosophies
procedures
Source: Saunders et al. (2009); Organised by author.
4.2 RESEARCH PROCESS APPROACH

Figure 4.1 shows that there are five major categories in the research process, which are:
research philosophies, research approaches, research strategies, time horizons, data
collection, and data collection techniques and procedures. Research philosophy is
concerned with the distinction of science from non-science, which procedures should
be followed, and establishes the conditions for a scientific explanation that should be
established (Smith 2000). The research approach is defined as a choice between testing
and building theory. Research strategy is a general plan for answering the research
questions (e.g. survey or case study). The time horizon is related to a snapshot or diary
approach. Finally, the data collection method chooses how the data will be gathered
(Saunders et al. 2007).
4.2.1 RESEARCH PHILOSOPHY

A research philosophy is defined as an assumption of how knowledge is developed and
analysed (Saunders et al. 2007; Maylor and Blackmon 2005; Levin 1988). The axiom
of “knowledge”, which is driven by research paradigms, can be explained in terms of
64
Ch4
ontology, epistemology, and methodology (Denzin and Lincoln 1994; Guba and
Lincoln 2005).
Ontology is the branch of metaphysics that is concerned with the nature of existence.
Social ontological considerations are mainly concerned with questions about the nature
of reality; for example, whether an objective reality exists or not, or whether social
entities can and should be considered as social constructions built up from the questions
and actions of social actors (Bryman and Bell 2011). There are two aspects of ontology,
which are objectivism and constructivism. Objectivism is an ontology which asserts
that social phenomena and their meanings have an existence that is independent of
social actors, while constructivism is a position which asserts that social phenomena
and their meanings are continually being accomplished by social actors (Saunders et al.
2009).
Epistemology is concerned with the question of what is regarded as acceptable

knowledge in a discipline. It asks whether the principles, procedures, and ethos of
natural science can and should be applied to the social world. Broadly speaking, there
are two opposing philosophical perspectives on epistemological consideration,
positivism and interpretivism (Bryman and Bell 2011). Positivism share some features
with the natural sciences in which it is believed that natural scientific methods can, and
should, be extended to the study of human mental and social life. It is also believed that
once reliable social scientific knowledge has been established, it will be possible to
apply it to control or regulate the behaviour of individuals or groups in society (Benton
and Craib 2001). On the contrary, interpretivists share a view that the subject matter of
the social sciences is fundamentally different from that of the natural sciences.
Methodology examines how we gain knowledge about the world, and Guba and
Lincoln (1994) indicated that a methodological question is constrained by both
ontological and epistemological considerations. Saunders et al. (2000) also claimed that
a research philosophy is a rather profound thought that has not normally been paid
attention to, but which governs the way that researchers go about doing research.
65
Ch4
Guba and Lincoln (2005) pointed out that there are three main research paradigms that
can be explained through ontological, epistemological or methodological positions,
namely: positivism, critical realism and constructivism (see Table 4.1).
Table 4.1: Comparisons of philosophical research paradigms

Constructivism/
Elements Positivism Critical Realism
Interpretivism
Ontology ‘Naïve realism’, in which an Critical realism – Relativism – local and
understandable reality is ‘real’ reality but specific constructed realities.
assumed to exist, which is only imperfectly The social world is produced
driven by immutable natural and and reinforced by humans
laws. The true nature of reality probabilistically through their actions and
can only be obtained by testing apprehendable. interactions.
theories about actual objects,
processes or structures in the
real world.
Epistemology Dualistic/objectivist. Modified dualist/ Transactional/ subjectivist.
Verification of hypothesis objective. Understanding of the social
through rigorous empirical Critical tradition/ world from the participants’
testing. Search for universal community. perspective through
laws of principles. Tight Findings probably interpretation of their
coupling among explanations, true. meanings and actions.
predictions and control. Researchers’ prior
assumptions, beliefs, value
and interests always
intervene to shape their
investigations.
Methodology Hypothetical-deductive Modified Hermeneutical/ dialectical.
experiments/ manipulative. experimental/ Interpretive case study.
Verification of hypotheses. manipulative. Action research. Holistic
Mainly quantitative methods. Falsification of ethnography.
hypotheses.
May include
quantitative
methods.
Inquiry Aim Explanation: prediction Understanding,
and control reconstruction.
Nature of Verified hypotheses established Non-falsified Individual and collective
Knowledge as facts or laws. hypotheses that are reconstructions, sometimes
probable facts or coalescing around
law. consensus.
Knowledge Accretion – “building More informed and
Accumulation blocks” adding to sophisticated
“edifice of knowledge”: reconstructions.
generalisations and Vicarious experience.
cause-effect linkages.
Source: Guba and Lincoln (2005).
In the past, research into airport performance has been predominantly influenced by
economic approaches (i.e. mathematical modelling, simulation and sensitivity analysis)
and, to a lesser degree, by behavioural approaches (i.e. questionnaires, interviews and
66
Ch4
case studies) (Lai et al. 2012). Both approaches are primarily based on the scientific
approach of positivism (Mentzer and Kahn 1995). If the research reflects the principle
of positivism, then the researcher will adopt the philosophical stance of the natural
scientist. Saunders et al. (2007) advised that positivist researchers prefer working with
an observable social reality and that the end product of this research is a law-like
generalisation similar to those produced by physical or natural scientists. A highly
structured methodology is expected to allow researchers to quantify their observations
and to analyse those observations through complicated statistical techniques (Saunders
et al. 2007). Positivistic research revolves around implicit assumptions which formulate
a reference framework by which to understand social reality (Giddens 1974).
Whatever the outcome of a positivist social investigation, the goal of analysis can and
must be able to formulate law or law-like generalisations of the same kind as those that
have been established in relation to social reality (Giddens 1974). Generally, the
positivist approach focuses on the testing of theories and provides new material for the
development of laws. There are strong connections between theory and research, which
carries the implication that “it is possible to collect observation in a manner that is not
influenced by pre-existing theories” (Bryman 2003 p.14).
As much of the debate is based on how methods are developed in natural science and
transferable to the social sciences, the positivist approach gives a clear sense of
separating subjective and objective data interpretation (McKenzie 1997). Under these
assumptions, it is believed that social phenomena can be scientifically observed and
measured. Along with the emphasis on objectivity, the attained knowledge through
scientific methods is viewed as resulting in greater strength in terms of reliability.
Furthermore, the positivist approach asserts that results based on a data set will be bias-
free (In the current context, bias is commonly caused by personal interpretations and
values that may influence conclusions drawn from a set of data). On the other hand,
constructivism/interpretivism views of knowledge can only be reached through
understanding of subjective meanings in social actions.
In between these two extremes (positivism and constructivism / interpretivism), another

emerging research paradigm is critical realism, which views the world as having three
67
Ch4
components namely: reality, actual and empirical (Sayer 2000). A critical realist
believes that the existence of the ‘true’ domain involve objects and structure which
requires casual power to be uncovered. However, the statement of ‘truth’ is not treated
as an absolute matter. Instead, a mechanistic form, such as a relationship or the degree
of practical adequacy, is more involved in uncovering the logic (Sayer 2000).
According to Sayer (2000), critical realism acknowledges that social phenomena are
intrinsically meaningful, and hence that meaning is not only externally descriptive but
also constitutive.
In comparison to positivism, a critical realist does not consider ‘actual’ as a complete

representation of the ‘real’. According to Reed (1997), social action is the results of the
choice of agents, which is influenced by the generative mechanism of structures. A
critical realist will not only explain the outcomes but also will reference the impacts of
the specific conditions or context. Therefore, the results from the activation of structure
at one point in time might not be replicated in the future due to the conditions in which
decisions are made.
In retrospect, qualitative researchers are greatly influenced by different intellectual

traditions, whereas quantitative researchers are strongly influenced by a natural science
approach to what should count as acceptable knowledge (Bryman and Bell 2007). With
a particular emphasis on the accumulation of knowledge and discrete steps that follow
forms of a pattern, the current research falls into the positivist paradigm. The
ontological position of the current research suggests that reality is an external objective
which exists beyond our knowledge and comprehension. In this research, transport and
logistics management are viewed as an objective entity, and therefore a decision was
made to adopt an objectivist stance to the study of particular aspect of logistics and
transport management. It is also believed that the social world is constructed through
people’s experience and knowledge and that only a certain facet of truth can be
encapsulated as opposed to the whole phenomenon. In essence, philosophical
paradigms have demonstrated ways of knowing and understanding the social world.
According to the literatures review in Chapter 2, the essential assumption of an airport

is that airports are auxiliary facilities and that their functions are to support airlines and
transportation industries. Consequently, airports are required to offer efficient services
68
Ch4
to maintain the competitiveness of these industries. Many positivist researchers assume

that airports, airport operation systems, or airport industries are tangible objects which
are independent of society. They do not consider the human factor as being one of the
determinants affecting airport performance. They attempt to find regularities and
general laws in airport management and often adopt quantitative methods that rely on
numerical and secondary data, and they select their variables through a literature review.
In addition, they formulate a production function using econometric techniques and
advanced efficiency analysis tools. Therefore, their definition of airport performance
can be identified as following a positivistic perspective.
This positivistic perspective could be considered to be influenced by the critical realist

approach, which attempts to study the existence of phenomena in the natural world.
Critical realism argues that a common approach can only reflect an environmental
change which corresponds to the empirical or actual level; it cannot recognise the
structural and functional changes which correspond to the real level of reality. This
approach also argues that airports have become pivotal nodes and will play a crucial
role in the logistics chain, while the previous approach suggests that airport services are
offered by a demand that is derived from other players in the logistics chain. Therefore,
a critical realist approach suggests that airports are situated in the centre of a
performance measurement framework as a point linking the members of a logistics
chain. The methodology of critical realism tends to deploy qualitative methods, where
critical realism mainly requires interpretive and qualitative methods to explain
mechanisms and structures. However, there is only limited evidence in the literature for
the use of this approach on airport performance evaluation and in port performance
evaluation. Two relevant studies about this philosophy in regard to airport performance
are summarised below in Table 4.2.
4.2.2 RESEARCH APPROACH

This section aims to examine the decisions involved in selecting a proper research
approach based on the positivist research paradigm. As noted earlier, positivist
researchers focus on explicit testing of theories or hypotheses with actual objects,
processes, or structures in the real world (Guba and Lincoln 2005). Therefore, the first
decision that is to be made is whether the research should use a deductive approach or
an inductive approach (Saunders et al. 2007).
69
Ch4
Table 4.2: Critical realist approach of airport efficiency evaluation research

Methodolo
Authors Aims of Research Objects of Research
gy
This paper examines how
Semi- Interview of airport managers from
Francis benchmarking is used by
structured European airports with over one
et al. airport managers as a means
face-to-face million passengers per annum
(2002) of internal performance
interviews. (sample size around 200).
comparison and improvement.
In-depth interviews: Passengers in
terminal waiting areas of a major
The purpose of this paper is to USA South-western airport (100
contribute to the development passengers).
In-depth
of a conceptual model of Focus groups: Frequent flyers in
Fodness and interviews
service quality in airports by Los Angeles, Dallas and Miami. 72
Murry Focus group
conducting an empirical frequent flyers (six focus groups in
(2007) Verbatim
investigation into the total and two in each location).
comments.
passengers’ expectations for
Verbatim comments: Visitors to
this service industry.
the web site of a major South-
Western airport (1,500 comments).
The deductive approach is a theory testing process, which commences with an

established theory or generalisation and seeks to determine if the theory applies to
specific instances. The inductive approach is a theory development process that starts
with observations of specific instances and then seeks to establish generalisations about
the phenomenon under investigation (Spens and Kovács 2005). The inductive approach
(which is also known as the logic of the ethnographer) is the mirror image of the
deductive approach. Therefore, theory is the outcome of the research, which involves
portraying generalisable inferences out of observations (Bryman and Bell 2003).
Research using an inductive approach is likely to study a small sample of subjects,
while the deductive approach studies a large number of subjects (Saunders et al. 2007).
Knowledge through literature is not necessarily needed at the starting point; instead,
empirical observations of the world through logical argumentation are used to lead to
theoretical generalisation (Spens and Kovács 2006). The general differences in these
two approaches are presented in Figure 4.2 and Table 4.3. The distinct differences
between the deductive and inductive approaches are to be found at the starting point of
the research process, the aim of the research, the point in time at which hypotheses or
propositions are developed, and whether they are further applied or not (Spens and
Kovács 2006).
70
Ch4
Figure 4.2: The difference between a deductive and an inductive approach

Prior
Theoretical New
theoretical
framework knowledge
knowledge
Suggestion of
Real-life Application
hypothesis or
observations / testing
propositions
Deductive
Research Approach
Inductive
Source: Spens and Kovács (2006).
Table 4.3: Differences between a deductive and an inductive approach

Deductive Approach Inductive Approach
• Gaining an understanding of the meanings

• Scientific principles.
that humans attach to events.
• Moving from theory to data.
• A close understanding of the research
• The need to explain causal relationship context.
between variables.
• The collection of qualitative data.
• The collection of quantitative data.
• A more flexible structure to permit
• The application of controls to ensure changes in research emphasis as the
validity of data. research progresses.
• The operationalisation of concepts to • A realisation that the researcher is part of
ensure clarity of definition. the research process.
• A highly structured approach. • Less concerned with the need to
• Researcher independence from what is generalise.
being researched.
• The necessity to select samples of
sufficient size in order to generalise
conclusions.
Source: Saunders et al. (2000).
This research adopts a deductive approach, which seeks to test the relationship between
airport ownership and efficiency. Deductive positivism is often regarded as the
predominant research approach, which is also true in logistics and transport research
(Naslund 2002; Aastrup and Halldorsson 2008; Wagner and Kemmerling 2010). By
using a deductive approach, a researcher can firstly generate a probability sample from
the entire population with a known degree of accuracy. Secondly, the operationalisation
of complex constructs with establishment of casual links between the constructs of
interest can be simplified. Because the deductive approach is widely used within
transportation research, it is unlikely that the research will be misunderstood and
subsequently under-valued.
71
Ch4
Another key consideration is to decide the purpose of the present study; that is, whether
it is explanatory, exploratory or descriptive. Saunders et al. (2007) described the three
classifications of research purposes as: firstly, explanatory studies, which aim to
establish causal relationships between variables in a situation or a problem; secondly,
exploratory research, which aims to explore new insights, ask questions, and assess
phenomena in a new light; and thirdly, descriptive studies, which aim to portray an
accurate profile of persons, events or situations. The distinctions between these
purposes are not absolute, and more than one purpose can be found in any given study
depending on the research question (Churchill and Iacobucci 2002). An explanatory
study (which is also known as a causal study) conducts experiments to investigate the
cause and effect of two or more measured variables (Churchill 1976). Evidence is
provided in regard to a structural causal relationship between variables by means of
concomitant variations and time order, which results in the elimination of other possible
explanations (Churchill and Iacobocci 2002).
An exploratory study operates is a more flexible and creative way to discover

unexpected meanings (Kinnear and Taylor 1991; Saunders et al. 2007). This is
particularly helpful in clustering a vague problem statement into smaller, more precise
sub-problem statements in the form of specific hypotheses. Exploratory research is
appropriate to problems about which little is known. In addition, it allows researchers to
be flexible with respect to the methods used for gaining insights and developing
hypotheses (Churchill and Iacobocci 2002).
Unlike both explanatory and exploratory studies, a descriptive study acts as an

extension of (or a forerunner to) exploratory or explanatory study (Saunders et al. 2007).
A descriptive study is used to describe the characteristics of certain groups, to estimate
the proportion of people in a specified population who behave in a certain way, and to
make specific predictions (Churchill 1976). Churchill and Iacobucci (2002) emphasised
that a good descriptive study must presuppose existing knowledge in regard to the
phenomenon studied. In direct contrast to an exploratory study, a descriptive study
requires clear specification and is arguably more rigid.
Regarding to this thesis, exploratory study would be appropriate in investigating the

under-explored gap in between airport privatisation policy and airport efficiency with
72
Ch4
the use of precise measurements through application of DEA and AHP data analysis
techniques, possibility of inconspicuous result would hinder the overall research
outcome.
4.2.3 RESEARCH STRATEGY

Quantitative and qualitative studies form two distinctive clusters of research strategy.
The distinction goes beyond the fact that quantitative researchers employ measurement
and that qualitative researchers do not (Bryman and Bell 2011). Quantitative research
emphasises quantification in the data collection process; its analysis entails a deductive
approach, and it incorporates the practices and norms of positivism. Meanwhile,
qualitative research emphasises words rather than quantification in the data collection
process, its analysis entails an inductive approach, and it tends to emphasise the ways in
which individuals interpret their social world. With regard to research strategy, from
the viewpoint of Bryman and Bell (2009), this research adopts a quantitative strategy to
evaluate airport performance by using quantifiable secondary data.
From other researchers’ viewpoints, research strategy means a general plan of how to
answer the research questions that are established by the researcher (Saunders et al.
2007). A variety of strategies are presented (as shown in Table 4.4). There are six
commonly used research strategies (i.e. experiment, survey, case study, grounded
theory, ethnography, and archival research), which are described below.
An experimental research strategy measures the efforts of manipulating one variable on

another (Robson 1993). A survey collects information in a standardised form from
groups of people (Robson 1993). In addition, a survey is an effective tool to get
opinions, attitudes and descriptions as well as cause-and-effect relationships (Ghauri
and Grønhaug 2002). A case study develops detailed intensive knowledge about a
single case or a small number of related cases (Robson 1993). Grounded theory
generates a theory from data gathered by a series of observations. Theory is grounded
in a continual reference to the data (Saunders et al. 2007). Similar to a case study,
ethnography, through involvement with a group, seeks to provide a written description
of the implicit rules and traditions of that group (Robson 1993). The last research
strategy, archival research, makes use of administrative records and documents as the
principal source of data (Saunders et al. 2007).
73
Ch4
AHP is adopted in this research to obtain weights of variables; hence, a survey is

employed as part of the research strategy. In addition, DEA is used to calculate airport
efficiency by analysing secondary data; consequently, archival research is also
considered to be an appropriate research strategy for use in this research.
Table 4.4: Research strategy and data collection method

Literature Research strategy Data collection method
Documentation
Experiment
Archival records
Survey
Interviews
Yin (2003) Case study
Direct observations
Archival analysis
Participant observation
Historical analysis
Physical artefacts
Experiment
Questioning
Survey
Observation
Thomas (2004) Case study
Documentation
Ethnography
Recording
Action research
Experiment
Survey Observation
Saunders et al. Case study Interviewing
(2007) Grounded theory Questionnaire survey
Ethnography Secondary data
Archival research
Structured interviewing
Bryman and Bell Quantitative
Self-completion questionnaire
(2011) Qualitative
Structured observation
4.2.4 TIME HORIZON

An important question to be asked in planning a research is if it is to be carried out at a
particular time, or, whether it is a series of representations of events which are taken
over a given period. According to Bryman and Bell (2007), a cross-sectional design is
built on the idea that it is a social survey that connects in peoples’ minds with
questionnaires in regard to two or more variables at a particular time. A cross-sectional
design is the most widely used design in social research when quick results are required
(De Vaus 2001).
In longitudinal design, the data are collected for each item or variable for two or more
distinct time periods. The subjects or cases analysed are the same, or at least
comparable, from one period to the next, and finally, the analysis involves some
comparisons of data between or among the periods under consideration (Burton 2000).
74
Ch4
In this research, the researcher seeks to describe the impact of airport performance in
different regions in the same year. Therefore, positing the key determinants for
designing the time horizon (i.e. time constraints, the abilities of subjects and the nature
of the research objectives and questions) means that a cross-sectional study design with
archival research strategy is adopted for use in this research.
4.3 DATA ANALYSIS METHODS

In the past, the efficiency of airports has generally been measured and compared on the
basis of the number of passengers, the amount of cargo, and the number of aircraft
movements. While this approach is valid, there is an assumption that all inputs and
outputs have equal weighting. In reality, however, the relative importance of each of
these inputs may vary between different airport stakeholders. This section aims to
describe which kinds of data analysis methods are adopted in this research.
To bring in qualitative judgements, MCDM will be used. Referring to section 2.2,

MCDM includes AHP, TOPSIS, PROMETHEE, and ELECTRE. Among these
methods, only TOPSIS and AHP were found to be used in airport efficiency evaluation
research. This research uses AHP presented as follows:
Tremendous efforts have been spent, and significant advances have been made towards
the development of numerous MCDM methods for solving different types of decision
problems (Yeh et al. 1999; Triantaphyllou 2000). Despite this, there is no universally
accepted approach for the general MCDM problem (Yeh et al. 2000), and the validation
of the decision outcome remains generally an open issue. The outcome is quite often
dependent on the method used. Besides, methods should enhance the Decision Makers’
(DMs’) learning about the problem (Zeleny 1983) as well as eliciting the DMs’
preferences (French, 1980).
The AHP is an appealing methodology by which to evaluate qualitative and

quantitative criteria systematically (Saaty 1980). It is very flexible in regard to allowing
the decision maker to structure the hierarchy to fit individual needs and preferences and
enables the DM to develop a trade-off among multiple criteria implicitly in the course
of structuring and analysing a series of pair-wise judgement comparison matrices
75
Ch4
(Zeleny 1983). Moreover, the AHP can combine tangible and intangible aspects to
obtain, in a ratio scale, the priorities associated with the alternatives of a problem.
Therefore, the major strength of the AHP is that it enables the systematic structuring of
any complex multi-player, multidimensional problem (Saaty 1980; Zelen 1983).
In Section 2.2, the literature showed that DEA, SFA and TFP have been applied widely
in airport efficiency evaluation literature, with DEA being the most popular. In regard
to SFA, some of the advantages over DEA are that it accounts for noise and can be used
to conduct conventional tests of hypotheses (Coelli et al. 2005). It also has some
disadvantages, such as the need to specify a distributional form for the inefficiency
term and the need to specify a functional form for the production function (or cost
function). TFP is usually measured by using either least squares econometric methods
or other index numbers. Some of the advantages of index numbers over least-squares
econometric methods are that only two observations are needed; they are easy to
calculate and the method does not assume a smooth pattern of technical progress, while
the principal disadvantage is it requires both price and quantity information (Coelli et al.
2005).
This research (as mentioned in Chapter 1) attempts to establish an objective and reliable
AEES and also is an attempt to compare the results generated by using different
analysis methods. Among these frontier analysis methods, DEA is the one that can
easily be combined with other methods and is the most reliable approach. Some of the
strengths of DEA include (Lewin and Minton 1986; Chen and Yen 2005):
• DEA analysis can combine many measures without the need to set prior weights for
various parameters to produce an overall efficiency measure.
• In contrast to conventional econometric techniques, DEA generates an intangible
‘efficiency’ frontier to make a comparison of efficiency in an optimal sense.
Therefore, a slack analysis, which provides the inefficient DMU information
necessary to raise outputs and reduce inputs in order to improve their efficiency, can
be easily conducted.
• In DEA, two or more input and output measures can be specified simultaneously. In
addition, since DEA is unit invariant, no normalisation or transformation of the input
and output variables are required.
76
Ch4
Despite the many strengths of DEA, there are also some limitations (Lewin and Minton
1986; Bowlin 1987; Zhang and Bartels 1998; Cooper et al. 2006; Lozano and Gutierrez
2011):
• The result will be influenced by the homogenous level of the measured DMUs.
• DEA cannot handle negative data.
• DEA does not consider random error and accepts instead that all errors come from
inefficiency; hence, the DEA is easily influenced by extreme values. If there are
significant variations between DMUs, then the efficiency score will be significantly
changed.
• The quantity of DMUs and the choice of input and output variables will influence
the DEA efficiency score, which causes a change of the feature and the position of
the efficiency frontier; therefore, the response is quite sensitive. Accordingly, choice
of the key element for DEA is very important.
• An insufficient number of DMUs for a DEA model will tend to rate all DMUs 100%
efficient because of an inadequate number of degrees of freedom. A rule of thumb
for maintaining this when using DEA is to obtain at least two DMUs for each input
or output measure.
4.4 ANALYTICAL HIERARCHY PROCESS (AHP)

Decision making within the real world inevitably includes the consideration of evidence
that is based on several criteria, rather than on a preferred single criterion (Beynon
2002). In business, decision making practices increasingly involve multi-criteria
decisions that are made by groups of DMs (Triantaphyllou 2000). The solution to a
multi-criteria decision problem can provide a recommendation to DMs who are faced
with a choice of Decision Alternatives (DAs). Multi-Criteria Decision Making (MCDM)
is currently one of the most well-known branches of decision making methodology
(Salo 1993) (details are described in Section 2.2.2).
4.4.1 THE RATIONALE OF AHP

The AHP is a decision making technique that was developed by Thomas Saaty in the
1970s. It depends on the study of both mathematics and psychology. The first step of
AHP is to decompose the elements that are related to the decision into goals, criteria,
and alternatives. The next step is to study these elements using both qualitative and
77
Ch4
quantitative analysis techniques (Ramanathan 2001). The AHP method is widely used
in both individual and group decision-making environments (Bolloju 2001). It is also
used to determine the relative ranking of DAs. It is built on human-beings’ intrinsic
ability to structure their perceptions (or their ideas) hierarchically, to compare pairs of
similar things against a given criteria or a common property, and to judge the intensity
of their preference for one thing over another (Forman and Peniwati 1998). These pair-
wise comparisons are determined by using scale values, which are processed in order to
derive their final weight values (priority values). However, in many decision problems
the information available from the DMs is often imprecise due to the use of inaccurate
estimates of criteria values and due to subjective errors that arise from the inconsistent
judgement of DMs (Pan and Rahman 1998).
4.4.2 THE AHP PROCESS

The AHP is aimed at integrating different measures into a single overall score for
ranking DAs (Önü and Soner 2008). Its main characteristic is that it is based on pair-
wise comparison judgements. The operational process is illustrated in Figure 4.3. There
are five main steps in the AHP (Saaty 2008):
(1) Define the decision object.
(2) Classify the variables which affect the decision and build a multi-level structure.
The top level is the goal of this decision; the intermediate levels are criteria and
sub-criteria for comparing DAs, and the lowest level are alternatives (as shown in
Figure 4.4).
(3) Make comparisons between each criterion in an upper level and the same criterion
in the level below it in terms of relative importance; that is, forge a set of pair-wise
comparison decision matrices. Let A represent an n x n pair-wise comparison
matrix, which can be expressed as:
1 𝑎!" …. 𝑎!!

1 𝑎!" 1 …. 𝑎!!
𝐴 = 𝑎!" = , (4.1)
⋮ ⋮ ⋱ ⋮
1 1
𝑎!! 𝑎!! … . 1
where 𝑎!! = 1 and 𝑎!" = 1 𝑎!" , 𝑖, 𝑗 = 1, 2, … , 𝑛.
𝑎!" > 0
78
Ch4
Let 𝐶! , 𝐶! , ⋯ , 𝐶! denote the set of criteria, while 𝑎!" represents a quantified

judgement on a pair of criteria 𝐶! and 𝐶! . Saaty (1980) constituted a measurement
scale for pair-wise comparisons. In addition, in order to create a contrast indicating
the degree to which one criterion is more important than another, a scale of
numbers (see Table 4.5) is settled. The values of 1, 3, 5, 7, and 9, in the original 1-
9 scale by Saaty (2008), represent equal importance, weak importance, essential
importance, demonstrated importance, and extreme importance, respectively; while
the values 2, 4, 6, and 8 are used to compromise between the values referenced
above.
Figure 4.3: The operational process in AHP

Establish Preliminary
evaluation perspectives
Selecting evaluation indictors
Questionnaire design
Questionnaire collection Feedback and revise
Set up comparative matrix in pairs
Obtain quantitative and qualitative

NO
performance data
Consistency examination
Yes
Weight value
Optimisation formula
Source: Saaty (2008); Organised by author.
79
Ch4
Figure 4.4: A simple structure of AHP

Goal
Criterion 1 Criterion 2 Criterion 3 Criterion 4
Alternative 1 Alternative 2 Alternative 3
Table 4.5: The fundamental scale of absolute numbers

Intensity of
Definition Explanation
Importance
1 Equal importance. Two activities contribute equally to the objective.
2 Weak or slight.
Experience and judgement slightly favour one
3 Moderate importance.
activity over another.
4 Moderate plus.
Experience and judgement strongly favour one
5 Strong importance.
activity over another.
6 Strong plus.
Very strong or
An activity is favoured very strongly over another;
7 demonstrated
its dominance demonstrated in practice.
Importance.
8 Very, very strong.
The evidence favouring one activity over another is
9 Extreme importance.
of the highest possible order of affirmation.
Source: Saaty (2008).
(4) To calculate the importance degree, the normalisation of the geometric mean
method is used to determine the important degrees of the DMs requirements
(Escobar et al. 2004). Let 𝑊! denoted the importance degree (weight) for the
𝑖 !! criteria, then:
! ! ! ! !
𝑊! = !!! 𝑎!" !!! !!! 𝑎!" , 𝑖, 𝑗 = 1, 2, ⋯ , 𝑛 , (4.2)
where n is the number of criteria.
80
Ch4
In addition, the maximum eigenvalue 𝜆!"# can be calculated by Equation (4.3)

and Equation (4.4):
1 𝑎!" …. 𝑎!! 𝑊! 𝑊!′

1
𝑎!" 1 …. 𝑎!! 𝑊 ′
A*𝑤! = ∗ ! = 𝑊! , (4.3)
⋮ ⋮ ⋱ ⋮ ⋮ ⋮
1 1 … . 1 𝑊 𝑊 ′
𝑎!! 𝑎!! ! !
! ! !
𝜆!"# = 1 𝑛 × 𝑊! 𝑊 + 𝑊! 𝑊 + ⋯ + 𝑊! 𝑊 . (4.4)
! ! !
(5) The next step is to test the matrix consistency through calculation, modifying it if
necessary in order to get an acceptable consistency. In line with the premise of the
consistency test, the eigenvector is calculated corresponding to the maximum
eigenvalue 𝜆!"# of the pair-wise comparison matrix. The weight is then defined
between each criterion and that in its upper level. The overall ranking weight
between each criterion is then determined. The final step is to make the decision.
Ø Calculate 𝐶𝐼 (which stands for Consistency Index, 𝜆!"# ) using the maximum
eigenvalue of the pair-wise comparison matrix, n as the size of matrix:
𝜆!"# − 𝑛
𝐶𝐼 = 𝑛−1. (4.5)
Ø In Equation (4.5), if 𝐶𝐼 = 0, then the evaluation for the pair-wise comparison

matrix is implied to be completely consistent. In particular, the closer the
maximal eigenvalue is to n, then the more consistent the evaluation is found to
be. Generally, a Consistency Ratio (𝐶𝑅) can be used as a guide to check for
consistency (Saaty, 1996). Table 4.6 shows the order of the matrix and the
average 𝑅𝐼 according to study of Aguarón and Moreno-Jiménez (2003), which
is used to calculate Equation (4.6). The formulation for 𝐶𝑅 is:
𝐶𝑅 = 𝐶𝐼 𝑅𝐼 . (4.6)
If 𝐶𝐼 < 0.1, then the consistency of matrix is tolerant; otherwise the matrix
should be modified.
81
Ch4
Table 4.6: Average Random Index (RI) for corresponding matrix size
N 1 2 3 4 5 6 7 8 9 10 11 12
RI 0.00 0.00 0.58 0.90 1.12 1.24 1.32 1.41 1.45 1.49 1.51 1.48
Source: Saaty (1996)
4.4.3 ALTERNATIVE SCALES WITHIN AHP

The utilisation of alternative comparison scales within AHP is discussed in this section.
As described above, a series of pair-wise comparisons and a unit scale are used in AHP
to play a fundamental role in quantifying a DM’s preference judgements. To date, the
Saaty 1-9 scale is the 9-unit scale that has been most used in the AHP. Furthermore,
some authors have tried to argue the appropriateness of the 1-9 scale, but the AHP
literature has not addressed the answer as to which of the available alternative scales
are most appropriate for the process of pair-wise comparisons. The influence of
alternative scales on the results of the AHP analysis is assessed in this section.
As mentioned in the previous section, the pair-wise comparisons made are quantified
by using numerical scale values, and they are processed in order to derive the final
weight values (Wi…Wn). DMs are asked to compare two DAs at a time: ‘Which one of
these two DAs is preferred?’ and ‘How strongly is it preferred?’ The DMs give weight
ratios to indicate the strength of preferences by using linguistic terms, where the values
of the pair-wise comparisons are determined according to the instructions depicted in
the 1-9 scale (Saaty 1980). The importance of scales for preference elicitation has been
emphasised by a number of previous practical works, and substantial empirical work is
still required to characterise the specific strengths and weakness of scales (Hämäläinen
and Salo 1997). Harker and Vargas (1987) stressed that the method of preference
revelation that is used in this present study is entirely independent of the scale of
measure. There have been several research studies that have suggested that the integers
from one to nine (i.e. 1-9 scale) should be avoided when using AHP. Firstly, the 1-9
scale is a limited range of numbers that cannot correctly describe the preference ratios
because of the weight of ratios that are above 9 (Harker and Vargas 1987; Schoner and
Wedley 1989; Dyer 1990; Salo and Hämäläinen 1993; Pöyhönen et al. 1997).
Given that all methods are in some way scale dependent, there has been considerable
discussion as to the correct scale to be used in the AHP and whether an unbounded
82
Ch4
scale should be used. Firstly, difficulty arises if Decision Alternative1 𝐷𝐴! is strongly
preferred to 𝐷𝐴! , and 𝐷𝐴! is ‘strongly preferred’ to 𝐷𝐴! . The scale value in Table 4.5
for strongly preferred is 5. Therefore, to maintain consistency, it would have to be rated
as 5×5 = 25 times preferred in comparison to 𝐷𝐴! . Hence, with a scale bounded by
the largest value 9, this consistent judgement is not permitted.
Additionally, if the respondent evaluates a comparison that is beyond the boundaries of

the scale, then the respondent is forced to modify the judgement and revise it so that it
is within the upper and lower limits of the scale. Therefore, in order to represent the
usual AHP relative comparisons, all pair-wise comparisons that exceed 9 are truncated
to 9. In addition, in the AHP, the respondents are forced to provide integer numbers
within a range of 1 to 9 although the actual judgement is not necessarily an integer.
Consequently, all pair-wise comparisons are truncated to the nearest integer value
(Carmone Jr. et al. 1997).
Thirdly, the 1-9 scale creates and deals with a very unbalanced scale of estimation. For
matrices with reciprocal elements in the Saaty AHP, half of the non-diagonal elements
are in the range of 1 to 9, and the other half are in the range of 1/9 to 1/1, which is
smaller compared with the former (i.e. 1 to 9). In the reciprocals, the range is
1 1 − 1 9 = 0.889, compared with 9 − 8 = 1 in the integers (Ma and Zheng 1991;
Mon et al. 1994; Triantaphyllou et al. 1994). Fourthly, although the use of the discrete
scale of 1 to 9 has the advantage of simplicity, it does not take into account the
uncertainty that is associated with the mapping of one’s judgement onto a number
(Brugha 2000; Leung and Cao 2000).
To overcome the deficiencies of the 1-9 scale, various judgement scales for a pair-wise
comparison have been proposed and evaluated. Among these, Kok and Lootsma (1984)
developed a geometric scale. A geometric scale quantifies the intensities of the
preference based on psychophysical arguments (Saaty 1987). The geometric scale has
been advocated over the Saaty 1-9 scale because of its transitivity and larger value span
found in many situations, resulting in more robust selections (Legrady et al. 1984). In
addition, Lootsma (1989) used a class of ratio scales based on a geometric progression.
Ma and Zheng (1991) considered the 1-9 scale in relation to its representation of
83
Ch4
language, stating that the suitability of a scale should be measured by the consistency
between the scale and the language. Therefore, they suggested that scales could have
their values evenly distributed in the interval [1/9, 1], while the values in the interval [1,
9] could be simply the reciprocals of the values in the interval [1/9, 1].
Donegan et al. (1992) suggested the use of a scale that is partly linear and partly
harmonic (∅ mapping scale), which resolves the unmathematical nature of the 1-9 scale.
An evaluation of 78 different scales appears in a study by Triantaphyllou et al. (1994),
which reveals that there is no single scale that can outperform all the other scales.
Furthermore, the same findings indicated that a few scales are very efficient under
certain conditions. Therefore, an appropriate scale for a successful application of pair-
wise comparisons needs to be selected. Beynon (2002) suggested that the original 1-9
scale is ineffective, pointing out that alternative scales offer a good opportunity to
follow linguistic scales, such as the 10/10 to 18/2 and 9/9 to 9/1 (Ma and Zheng 1991),
∅ mapping (Donegan et al. 1992) and 1.1 to 1.9 (Saaty 1987) scales.
Some related work has been carried out with verbal probability assessments where the
verbal expression seems to be best modelled by interval judgement rather than point
estimates (Beyth-Marom 1982; Hamm 1991; Timmermanns 1994). Hershey et al.
(1982) showed that a linear transformation of an interval scale can drastically alter the
results. Therefore, provided that these results can be generalised to ratio comparisons of
relative preference, it is possible that exact numbers in the AHP should be replaced by
intervals of numbers (Pöyhönen et al. 1997). Saaty and Vargas (1987) proposed an
interval judgement for the AHP as a way to model the subjective uncertainty in the
DM’s preferences. Meanwhile, Arbel (1989) developed efficient algorithms for
synthesising interval judgements into dominance relations on the DAs.
Previous studies have revealed some of the limitations of the use of a 1 to 9 scale.
Consequently, in this research, the 1 to 9 scale is contrasted with alternative scales. The
alternative scales that are used in this study are illustrated in the following section.
84
Ch4
4.4.4 THE FEATURES OF ALTERNATIVE SCALES

As shown in Table 4.7, the alternative scales that are considered in this section, apart
from the original 1 to 9 scale, include two geometric scales, two scales from Ma and
Zheng (1991) and the ∅ mapping scale (Donegan et al. 1992). There are also other
scales which are numerically close to these five alternative scales but which have
different theoretical motivations (Lootsma 1989).
In Table 4.7, the first column lists the verbal statements of preference taken from Saaty
(1980). The second column is the 1-9 scale that was proposed by Saaty (1980). The
third and fourth columns are geometric scales based on series of powers of ‘𝑒’ and ‘2’
(the 𝑒1−9 scale and the 21−9 scale), which were applied by Legrady et al. (1984), who
suggested that a geometric scale with powers of a suitable base number was more
appropriate. Moreover, scaling of words and phrases expressing grades of approval or
disapproval has shown that the response range (i.e. the ratio of the extreme stimuli) can
easily be calculated up to 100; however, the 1-9 scale allows a response range of 9 only.
Table 4.7: Definition of alternative scales

Verbal Scale
statement of ø
importance
1-9 𝑒1−9 21−9 9/9-9/1 10/10-18/2
mapping
Equal 1 𝑒0 =1.0000 20 =1 9/9=1.0000 1=1.0000 1.0000
- 2 𝑒0.5 =1.6487 21 =2 9/8=1.1250 11/9=1.2222 1.1180
Moderate 3 𝑒1.0 =2.7183 22 =4 9/7=1.2857 3/2=1.50000 1.2536
- 4 𝑒1.5 =4.4817 23 =8 9/6=1.5000 13/7=1.8571 1.4142
Strong 5 𝑒2.0 =7.3891 24 =16 9/5=1.8000 7/3=2.3333 1.6125
- 6 𝑒2.5 =12.1825 25 =32 9/4=2.2500 3=3.0000 1.8708
Very strong 7 𝑒3.0 =20.0855 26 =64 9/3=3.0000 4=4.0000 2.2361
- 8 𝑒3.5 =33.1155 27 =128 9/2=4.5000 17/3=5.6667 2.8284
Extreme 9 𝑒4.0 =54.5982 28 =256 9/1=9.0000 9=9.0000 4.1231
The fifth and the sixth columns in Table 4.7 report the two 9-unit scales from Ma and
Zheng (1991), namely 9 10 − 𝑘 , and 9 + 𝑘 11 − 𝑘 with 𝑘 = 1, ⋯ ,9 called ‘9/9
to 9/1’ and ‘10/10 to 18/2’ scales, respectively. The sixth column is also called a
balanced scale (Salo and Hamalainen 1997), which is based on the idea that the local
weights should be evenly dispersed over the weight range [0.1, 0.9] (Pöyhönen et al.
1997). The seventh column is a set of scale values that was introduced in Donegan et
85
Ch4
!!!
al.’s (1992) ∅ mapping scale. They use the formula “∅: 𝑡 → exp (𝑡𝑎𝑛ℎ!! ( !
))” to
map into a range where “real” arithmetic applies (i.e. into a true ratio scale so that the
transformation of the pair-wise comparison values can be used as entries to a
comparison matrix which can then be handled in the usual AHP fashion). In this
research, the weights of each variable using these six alternative scales are compared in
the next chapter.
4.5 DATA ENVELOPMENT ANALYSIS (DEA)

This section aims to describe the DEA method, including its characteristics, the models
which are employed in the research, and the progress of its evaluation.
The DEA method was first proposed by Charnes et al. (1978); they described the DEA
methodology as
‘a mathematical programming model applied to observed data that provides a new way
of obtaining empirical estimates of external relationships such as the production
functions and/or efficiency production possibility surfaces that are the cornerstones of
modern economics.’
Since then, numerous applications employing the DEA methodology have been
presented and involve a wide area of contexts: education, health care, banking, armed
forces, sports, transportation, agriculture, retail stores and electricity suppliers (Gattoufi
et al. 2004). Originally, this method was designed to evaluate the efficiency of DMUs,
which use multiple inputs to produce multiple outputs, without a clear identification of
the relation between them. DEA has progressed throughout a variety of formulations
and uses in other kind of industries. Gattoufi et al. (2004) cited more than 500 articles
in a comprehensive bibliography and stated that DEA methodology is an important
analytical tool whose acceptance is no longer in doubt.
This research does not intend to cover the basic aspects of DEA models. A good
introduction to DEA notation, formulation and geometric interpretation can be found in
Charnes et al. (1994), Ali and Seiford (1993) and Coelli et al. (2005). As discussed
86
Ch4
therein, a model can be described by the envelopment surface, orientation of the model,
invariance of units, and efficiency measurement.
The DEA method was first proposed by Charnes et al. (1978), who employed a
mathematical programming model called the Charnes, Cooper, and Rhodes model (the
CCR model) to measure the technical efficiency of DMUs using the Pareto optimum
concept. Charnes et al. (1978) assumed that a situation involved Constant Return to
Scales (CRS); namely, that increasing the input of a part would simultaneously increase
the output. The problem with calculating DEA scores can, therefore, be viewed as a
linear programming issue. The CCR model is usually applied in the first stage of DEA.
It is also the first step in entering the DEA field. The CCR model includes both input
and output-oriented models. This model is basically assumed to have constant returns to
scale; however, each DMU might operate on different returns to scale, which may
cause inefficiency.
Banker et al. (1984) extended the CCR model, which they termed the Banker, Charnes,
and Cooper model (the BCC model). The BCC model assumes the existence of
Variable Returns to Scale (VRS). The key part of these two models (i.e. the CCR and
BCC) is that Charnes et al. (1978) included Pareto optimality into the model, in which
each DMU selects the optimum input and output multiplier for the purpose of
maximising its own efficiency, and the only constraint is that the value of the selected
multipliers must not exceed 1 to satisfy the constraint that the maximum efficiency
value is 1.
The choice of a DEA model depends on some assumptions regarding the data set to be
employed and on some prior results about the industry to be studied. The data set has to
describe the activities of the units in the best possible way. It is especially important to
have some idea about the hypothetical returns to scale that exist in the industry. This
knowledge is going to determine the envelopment CRS or VRS of the model. Once the
selection of envelopment surface has been made, an orientation of the model to
determine the measurement of the efficiency is needed. There are three basic
orientations: input, output and output/input. An input orientation focuses on the
proportional decrease of the input vector; the output orientation adjusts the proportional
87
Ch4
increase of the output vector, and the output/input orientation does not discriminate the
importance of possible increases in output or decreases in input.
The units involved in this research determine the selection of the orientation, and it is
very important to have in mind what the real possibilities of managers are. In the
structure conduct-results tradition, the investigator must try to establish what the
conduct of agents and the structure of the market are in order to determine a possible
orientation for the model. In DEA analysis, it is generally assumed that there are n
production units to be evaluated, using amounts of m different inputs to produce
quantities of s different outputs. Specifically, the oth production unit consumes xio units
of input i (i = 1 to m) and produces yro units of output r (r = 1 to s). The oth production
unit can now be described more compactly with the vectors (Xo , Yo), which denote,
respectively, the vectors of input and output values for DMUo.
Next, we consider the dominance comparisons for this production unit using the data
set as a reference. DEA considers the dominance of the linear combinations of the n
production units, i.e. 𝑘 𝜆! 𝑋! , 𝑘 𝜆! 𝑌! , with the scalar restricted to be non-
negative. The production unit o is dominated, in terms of inputs, if at least one linear
combination of production units shows that some input can be decreased without
worsening the rest of the inputs and outputs. The production unit o is dominated in
terms of outputs if at least one linear combination of production units shows that some
output can be increased without worsening the rest of the inputs and outputs. Thus, the
method serves to partition a set of production units into two subsets: the efficient
production units and the inefficient ones. The method also serves to calculate the level
of inefficiency of a given inefficient production unit. Airport managers can affect the
efficiency of the airport using their inputs (such as runways, terminal buildings,
employees, etc.) in different manners. In this research, an output orientation is
employed. Once an airport has invested in the building of new runways or new
terminals, it is difficult for managers to disinvest to save costs, therefore invalidating
the input orientation (Martin and Roman 2001).
88
Ch4
In this sense, it is more credible to use airport facilities as intensively as possible since
variables of production are fixed or semi-fixed. Formally, the DEA output efficiency
for the unit o is calculated through the following linear programming Equation (4.7):
max!, ! ,!! !! 𝑧! = 𝜙 + 𝜀×1𝑠 ! + 𝜀×1𝑠 ! (4.7)

Subject to: 𝑌𝜆 − 𝑠 ! = 𝜙𝑌! ,
𝑋𝜆 + 𝑠 ! = 𝑋! ,
1𝜆 = 1
𝜆, 𝑠 ! , 𝑠 ! ≥ 0,
where X and Y are the input and output matrixes, respectively; Xo and Yo are the input
and output vectors of the unit o, respectively; 𝜙 and 𝜆 are parameters calculated in the
model, and represent the maximum proportional output that can be attained and the
linear convex combination that dominates the oth unit, respectively; 𝜀 and 𝑠 ! , 𝑠 ! are the
Archimedean constant and the slack variables, respectively.
The model compares the production unit o with all the convex linear combinations of
production units. The linear programming problem is solved for every airport in the
sample in order to obtain its relative performance. The efficiency measure obtained is
considered the technical efficiency and is calculated as the inverse of the maximum
proportional output that can be obtained for the indicated inputs.
4.6 LITERATURE OF INTEGRATED AHP/DEA MODEL

Although the combined AHP and DEA approach has attracted comparatively less
attention (Ho 2008), an integrated AHP/DEA model has been developed and used for a
number of purposes, including: supply chain performance evaluation (Guo et al. 2006),
facility layout design in manufacturing systems (Yang and Kuo 2003; Ertay et al.
2006), warehouse operator selection (Korpela et al. 2007), improvement and
optimisation of railway systems (Azadeh et al. 2007), and bridge risk assessment
(Wang et al. 2008).
Ho (2008) has done an extensive review of integrated AHP and its applications, and has
reported that only four papers have employed combined AHP and DEA. In addition to
89
Ch4
these papers, the combined AHP/DEA approach has been used in a number of different
ways (Lozano and Villa 2007; Ramanathan 2006; Jing-Yuan et al. 2006). For example,
Chen and Chen (2007) used DEA with a Balanced Scorecard (BSC) performance
evaluation in the semiconductor industry, where they used AHP to obtain the weights
for the four perspectives given by the BSC although in a research and development
context, the application of an integrated AHP/DEA model has not yet been used and it
has not been reported in the literature. Hsu (2005) used Fuzzy DEA with BSC for
multinational research and development project performance assessment. While the
DEA was originally designed for classification, it has been widely used to measure
overall relative productivity and efficiency in relation to allocation decisions. The
following paragraphs describe the characteristics and differences of the AHP/DEA
approaches that have been developed in previous literature.
Xing and Tseng (2002) argued that the coefficient from the DEA model could be
replaced with the weight from the AHP. To do this, the variable weight must first be
determined. The efficiency of DMUs can then be obtained. However, the need to
simultaneously obtain weight and efficiency makes achieving this difficult. On the
other hand, in support of the integration of AHP and DEA, Sinuany-Stern et al. (2000)
indicated that the AHP and DEA have the same characteristics for cases of single input
and output. The basic idea is to employ a cross-evolution concept of AHP for ranking
DEA DMUs and then to extend this to multiple inputs and outputs based on the DEA
(Lee and Tseng 2006).
Takamura and Tone (2003) developed a combined AHP and DEA approach to deal
with the relocation of several government agencies out of Tokyo. Firstly, the AHP was
used to obtain the relative importance weightings of both criteria and attributes.
Secondly, based on the AHP weightings, DEA was adopted to measure the
effectiveness of alternative locations. Meanwhile, Yang and Kuo (2003) proposed a
combined AHP and DEA approach to solve a facility layout design problem. A
computer-aided layout planning tool (called Spiral) was adopted to generate a number
of alternative layouts in advance. The relative importance weightings of alternative
layouts were obtained by using the AHP pair-wise comparison with respect to three
qualitative factors: flexibility, accessibility, and maintenance. DEA was then used to
90
Ch4
solve the layout design problem by simultaneously considering both the qualitative and
quantitative performance data leading to the identification of performance frontiers.
Saen et al. (2005) proposed a combined AHP and DEA approach to measure the
relative efficiency of slightly non-homogeneous DMUs. Due to the fact that some
DMUs may lack one or more features (i.e. input and/or output), they used AHP to
estimate the missing features so that they could build a DMU that was as close to
reality as possible. To do this, two alternatives were compared with respect to the
attribute of the higher levels, which were: the DMU which lacked the feature(s) and the
series means of other DMUs. The data for the mean of other DMUs was obtained by
taking the mean of each feature of all DMUs with the exception of the one that had a
missing value. The data was assumed to be normally distributed. Meanwhile, Ertay et al.
(2006) applied the combined AHP and DEA approach to aid in facility layout design;
their approach was very similar to that presented in Yang and Kuo (2003). Firstly, a
computer-aided layout planning tool (called VisFactory) was adopted to generate a
number of alternative layout designs. Secondly, the AHP was used to obtain the relative
with respect to two qualitative factors, which were flexibility and quality. Thirdly, the
DEA was used to evaluate the designs by simultaneously considering both qualitative
and quantitative data. The best design was then selected. In addition, the flow distance,
adjacency, and shape ratio that were proposed by Yang and Kuo (2003) were
considered, as were the material handling vehicle utilisation and material handling costs.
Korpela et al. (2007) developed an approach to select a warehouse operator network by
combining the AHP and DEA. The outcome of the AHP analysis was a preference
priority for each alternative operator describing the expected performance level.
Additionally, Feng et al. (2004) combined AHP and DEA to measure the efficiency of
university management activities. Their study demonstrates that one of the basic
concepts of AHP is the importance of pair-wise comparison. To support this, several
studies have indicated that AHP can be applied to form an AHP/DEA ranking model
for the purpose of improving DEA usability (Feng et al. 2004; Friedman and Sinuany-
Stern 1998; Lee and Tseng 2006; Sinuany-Stern et al. 2000). The advantage of the
AHP/DEA ranking model is that the comparative weight (or importance) can be
derived from inputs/outputs via an AHP pair-wise comparison (Lee and Tseng 2006;
Sinuany-Stern et al. 2000). While most studies of the AHP/DEA model have focused
on investigating the efficiency of DMUs, a method by which to structure the
91
Ch4
appropriate weight of the input/output variables when DMUs achieve maximum

efficiency has been relatively less studied, and, therefore, this is one of the research
motivations of the present study (see Research Question 2).
4.7 DATA COLLECTION METHOD

There are practical reasons for the intensive use of survey-based methods to collect data
in air transport research. Firstly, they allow for the involvement of various functions or
locations within a firm, which are a convenient for gathering data. Secondly, surveys
allow the researcher to reach upper managerial and executive levels. Lastly, surveys
can ensure anonymity, and their standardised wording can help to avoid drawbacks,
such as interviewer bias (Wagner and Kemmerling 2010). The following section briefly
outlines the data collection methods and questionnaire development that are employed
in this research.
The research gaps revealed from the literature review in Chapter 2 (Section 2.5)
indicated that existing research has generally involved the subjective selection of
evaluation variables by the authors. This gap can be filled easily by adopting more
objective methods when choosing evaluation variables, such as interviews. Therefore,
in this research, the AHP method is used, which includes semi-structured and structured
interviews surveys to be used to develop an AEES.
4.7.1 SEMI-STRUCTURED INTERVIEW

Interviews are generally believed to be an appropriate method for an exploratory study
to find out what is happening and to seek new insights (Robson 2002; Saunders et al.
2007). While in unstructured interviews, there is no predetermined list of questions to
work through, in semi-structured interviews, the researcher has a list of questions and
themes to be covered although these may vary from case to case. This means that the
researcher can omit some questions in particular interviews and may change the order
of questions depending on the flow of the conversation (Saunders et al. 2007). This
research employed a semi-structured interview because the situations which the
interview aimed to explore were relatively constrained. In other words, the exploration
has to be made within the research questions and the AEES this research was intended
to develop. However, the numbers of questions were minimised in order to make the
92
Ch4
interview more flexible and less directive and to obtain deeper understanding of
interviewees’ perceptions. One of the main objectives of conducting interviews is to
ensure the validity and the reliability of the variables of the evaluation system. To
address subjectivity concerns, which is often one of the major criticisms faced by
qualitative research, this study has applied the AHP method during the process of
variable selection. The characteristics and application of the pilot semi-structured
interviews are described in Table 4.8.
Table 4.8: The characteristics of semi-structured interviews

Characteristics of
Application in this Research
Semi-Structured Interviews
A semi-structured interview may be used for The semi-structured interview in this research is
an exploratory study in order to understand used to help the researcher understand the
1.
the causal relationship between variables relationship between the variables and airport
(Copper and Schindler 1998). efficiency.
If an interviewee is a manager in an airport, then
The researcher should have a list of themes the interview questions emphasised the financial
and questions on the fairly specific topics to and service variables of the airport.
2. be covered. The researcher may omit some In this pilot study, the questions were spread
questions in particular interviews (Bryman into different perspectives because the
and Bell 2007; Saunders et al. 2007). interviewed expert is a researcher on airport
efficiency.
When facing different interviewees, different

questions will be asked to help researchers to
Additional questions may be required to
comprehend or amend research questions and
explore research questions and objectives
3. objectives.
given the nature of events within particular
In this pilot study, the expert can give some
organisations (Saunders et al. 2007).
suggestions for the research questions and other
areas of this research.
The semi-structured interview scheme allowed for the addition of more questions and
asking these questions differently depending on the responses from interviewees. The
interviews were recorded by note-taking and recording and then transcribed for analysis.
Sampling interviewees and analysis are addressed in Chapter 5. While ethical issues,
such as anonymity and confidentiality, occur when collecting primary data rather than
secondary, these issues are generally considered more significant when conducting
qualitative research as this involves direct interaction with persons as compared to
quantitative research. Ethical issues are also presented in the section addressing
sampling issues.
93
Ch4
Analysis the qualitative data requires a different approach from that used when
analysing quantitative data. Bryman and Bell (2007) indicated that clear-cut rules about
how qualitative data analysis should be carried out have not been developed because
qualitative data takes the form of a large corpus of unstructured textual material.
However the basic concept for qualitative data analysis is categorisation and
characterisation (Saunders et al. 2009). Several of the aspects the interviewees
described in the interviews can be sorted into various categories. In this research,
interviewees were asked to help identify the categories of variables which are
appropriate to this AEES and also were asked to help select appropriate variables. For
this reason, the author attempted to develop a preliminary AEES. This categorisation is
supported by existing literature (addressed in Chapter 2). However some categories can
emerge from the feedback of interviewees. After finishing this semi-structured
interview, a structured questionnaire was undertaken in the next step.
4.7.2 STRUCTURED INTERVIEW/ SURVEY

Structured interviews are often designed to emphasise the greater generality in the
formulation of interviewer’s concerns of the informants’ perspectives (Bryman 2008).
A structured interview can be conducted in a structured questionnaire. The advantage of
structured interviews lies in the uniformity of the interviewees’ behaviour; hence, other
people than the researcher are able to replicate the interview in a similar situation
(Ghauri and Grønhaug 2002). The formality of the answers comes in a uniform length
which can lend itself nicely to being coded, quantified and compared (Denscombe
2003). In a normal structured interview, the limitation of sample size and analysis
technique should follow some statistical order. In addition, unless there are some other
reasons, the interviewees in structured interview are usually interviewed on one
occasion only (Bryman 2008). The aim of a structured interview is to gather data from
a statistically representative sample of the population in a controlled environment
(Bryman 2008).
Due to the quantitative nature of the questions, a structured interview is designed to

categorise, select and rank the appropriateness of variables that are adopted in an
evaluation system. The interview questions intend to identify which variables are
appropriate for use when evaluating airport financial and customer efficiency from each
interviewee’s point of view. In order to avoid any misunderstanding of the question, the
94
Ch4
researcher provides the definition of each variable in the questionnaire in order to

increase both response and completion rates. The data from structured interviews can
then be analysed through a quantitative approach to identify any significant differences
between variables, such as through the use of ANOVA (Analysis of Variance) or other
statistical methods. The results of the data analysis can help the researcher identify the
most appropriate variables. In this research, the ANOVA method is not used to analyse
the questionnaires, instead AHP analysis is used to analyse the questionnaire.
Because of the volume of respondents, visiting every prospective respondent would be

difficult due to budgetary and time constraints. Therefore, in this research, a
questionnaire survey was selected as the main empirical data collection method.
According to Maylor and Blackmon (2005), a survey is a useful technique by which to
capture facts, opinions, behaviour or attitudes from a range of respondents. However,
according to Saunders et al. (2007), there are various types of survey methods that
should be taken into account when implementing this specific method (see Figure 4.5).
Figure 4.5: Methods of questionnaire administration

On-line questionnaire
Self-
Postal questionnaire
Administered
Delivery and collection

Questionnaire questionnaire
Telephone questionnaire
Interviewer-
administred
Structured questionnaire
Source: Saunders et al. (2007).
There are two main forms of questionnaires, namely: self-administered questionnaire

sand interviewer-administered questionnaires. The main difference between these two
forms is the involvement of an interviewer. Self-administered questionnaires are
completed by the prospective respondent without any aid from the interviewer while
interviewer-administered questionnaires require verbal or face-to-face contact between
the interviewer and the interviewee (such as telephone questionnaires, structured face to
face interviews, or questionnaires) (Maylor and Blackmon 2005). The interviewer-
95
Ch4
administrated questionnaire is one of the most common techniques that is used in all
types of business and areas of management research (Maylor and Blackmon 2005;
Aastrup and Halldorsson 2008). The flexibility of an interviewer-administrated
questionnaire is that it focuses on a specific subject and the possible extension of its
meaning. The interviewer-administrated questionnaire enables researchers to gain more
freedom to probe beyond the answer through a form of dialogue with the respondent
and, therefore, affords them the opportunity to collect additional information (May
2001; Bryman and Bell 2007). However, in large samples, this particular method is
expensive in terms of time and cost, especially when the prospective respondents are
geographically dispersed. Alternatively, a self-administered questionnaire method has
more advantages in terms of convenience (i.e. time, cost, and location for both
interviewer and interviewee); it is also less obstructive (i.e. absence of interviewer
effects) to interviewers (Bryman and Bell 2007). There are a number of disadvantages
which need to be considered when using self-administered questionnaires, such as lack
of clarification when needed and less opportunity to collect additional data (Maylor and
Blackmon 2005). In this research, some of experts are interviewed using a structured
questionnaire.
Of the three self-administered methods, the postal questionnaire was adopted for use in
this research, rather than an on-line questionnaire or a delivery and collect
questionnaire. According to Bech and Kristensen (2009), among older respondents (i.e.
those aged between 50 and 75) a postal or mail survey will have a typical response rate
of around 30 % higher than an on-line survey. There are two main reasons why on-line
surveys have a lower response rates than postal questionnaires including ‘survey fatigue’
or a lack of internet access on the part of the recipient. Due to the fact that the average
ages of experts who were interviewed in this research were around 50, it was
determined that the main questionnaire survey should be based on a postal
questionnaire supported by a structured questionnaire.
This study is aimed at the development of an AEES, and in order to enhance reliability
and validity of the variables, the proposed approach employs two methods to select
variables for the airport performance evaluation system. DEA is adopted to conduct
quantitative analysis and the AHP method is employed to acquire the weight of
individual variables. Therefore, two forms of data collection processes are used. The
96
Ch4
first is a semi-structured survey, and the second is a questionnaire survey. The former is
intended to help the author conduct a pilot study that can acquire expert opinions to
enable the establishment of evaluation variable sets that are obtained by means of in-
depth interviews (semi-structured interviews). The latter are intended to obtain the
weight of each variable resulting from handing out an AHP questionnaire (or by
conducting so called structured interviews).
In addition, some secondary data, acquired from the Air Transport Research Society
(ATRS) and the annual report from individual airports, are also used in this research
when undertaking the DEA analysis.
4.8 RESEARCH STRUCTURE AND SURVEY DESIGN

After discussing several concepts on research design in the previous sections, the
research process is described in this section. The framework of the research procedures
that are used in this study and the tasks of each stage are summarised and presented in
Figure 4.6.
Figure 4.6: Framework of research procedures
4.8.1 Stage I: Variables collection

A preliminary variable set is constructed in the first stage, which is based on the
research presented in Chapter 2 (Section 2.3.4). In the preliminary variables set, seven
inputs (i.e. number of employees, number of gates, number of runways, size of terminal
97
Ch4
area, length of runway, operational expenditures, and non-operational expenditures)

and five output variables (i.e. number of passengers, amount of freight and mail,
aircraft movements, aeronautical revenue, and non-aeronautical revenue) were chosen.
These seven input and five output variables are selected from the most widely used
variables among the 66 previous studies of airport efficiency (see Table 2.3 and Table
2.4).
Generally speaking, AHP is a method that uses both semi-structured interviews and
structured interviews to acquire data. In this stage, a pilot study (semi-structured
interview) is conducted after setting up the preliminary evaluation system in order to
enhance the reliability and validity of the preliminary variables. One of the advantages
of conducting a pilot study is that it can give advance warning about where the main
research could fail, where the research protocols may not be followed, or whether the
proposed methods or instruments are either inappropriate or too complicated (Polit et al.
2001). The suggestions that were made during the pilot study in this research came
from three experts who were selected from different areas: academia, practice, and the
airport authority. The accessibility of the airport data needed to be considered when
deciding the final version of variable sets. The final variables were then set up after
these semi-structured interviews were completed (further details on this stage are
presented in Chapter 5).
The selection of the sample airports is another important task in this stage. There are
two main principles for selection of the sample airports. One of the objectives of the
research questions was to examine if airport ownership or governance have an
influence on an airport performance. Therefore, the first principle for sample selection
was that the samples should cover eight types of airport ownership, as put forward by
Oum et al (2006) and Gillen (2011) and detailed in Section 3.3.1. However, one aspect
of airport ownership, independent non-profit corporations, can only been found in
Canada and is therefore outside of the geographical scope of this thesis. The second
principle is that the sample airports should be similar in nature, as derived from the
limitations of DEA. Therefore, in this research, the sample airports consisted only of
primary airports in two regions (i.e. Europe and the Asia-Pacific region) due to the
availability and accessibility of the data and the variety of airport ownerships.
According to these two principles and the survey from “Air Traffic Data of 2010”,
98
Ch4
which is published by the Airport Council International (ACI), the sample size is
limited to the busiest 12 airports from each region. Table 4.9 shows 24 sample airports
which are studied in this research and the seven different types of airport ownership
mentioned in Chapter 3 (see Section 3.1.3) are covered.
Table 4.9: Sample airports in this research

Europe Ownership Category Asia-Pacific Ownership Category
Amsterdam Private company

Bangkok (BKK) Private company
(AMS) (Public majority)
Barcelona
Public-owned company Beijing (PEK) Private company
(BCN)
Frankfurt Private company
Guangzhou (CAN) Private company
(FRA) (Public majority)
Public-owned, operated by Public-owned

Istanbul (IST) Hong Kong (HKG)
a private company company
London Public-owned
Private company Incheon (ICN)
(LGW) company
Public-owned
London (LHR) Private company Kuala Lumpur (KUL)
company
Madrid Public-owned Private company
Osaka (KIX)
(MAD) company (Public majority)
Munich Public-owned Public-owned

Tokyo (HRT)
(MUC) company company
Public-owned
Paris (CDG) Private company Shanghai (PVG)
company
Paris (ORY) Private company Singapore (SIN) Public-owned
Private company
Rome (FCO) Private company Shenzhen (SZX)
(Public majority)
Zurich (ZRH) Private company Sydney (SYD) Private company
4.8.2 STAGE II: WEIGHTS CALCULATION

In this stage, the weights of each variable are developed using basic AHP 1-9 scales
and other alternative scales. By making pair-wise comparisons at each level of the
hierarchy, the participating experts can develop relative weights and set priorities. They
can then differentiate the importance of the variables with continued feedback and
revisions.
In this stage, an AHP questionnaire, which focuses on managers from airport

companies, air transport researchers, or officers in civil aviation authorities, is carried
99
Ch4
out to determine what the most important variables for these interviewees are when
evaluating airport operational efficiency. The final version of the selected variables is
used to establish the AEES. Some of questionnaires were handed out by post and some
of were conducted using face to face structured interviews (a copy of the questionnaire
that was used is attached in Appendix II).
4.8.3 STAGE III: EFFICIENCY EVALUATION

The third stage of this research is aimed at applying variable sets to assess the
efficiency of the sample airports. In general, the performance is evaluated in terms of
relative efficiency by applying the basic DEA model and an integrated AHP/DEA
model. Four main steps are followed in this stage; firstly, computing the relative scores
of variables from raw data and analysing in different groups; secondly, computing the
relative weighted scores of all individual airports based on weights which are derived
from AHP analysis in alternative scales; thirdly, evaluating the efficiency in terms of
the relative weighted score for each airport (in this step a DEA model is used to
evaluate airport efficiency); and finally, measuring the operating efficiencies of primary
airports in Europe and Asia-Pacific region by applying integrated AHP and DEA
models. Identifying the benchmark airport and the relatively inefficient airports is the
main task in this step. Additionally, comparison of airport performance using a set of
standardised benchmarking and reporting variables benefits both airport managers and
stakeholders who wish to make judgements about airport performance, both
comparatively and individually. Furthermore, sensitivity analysis is also conducted in
this research, and another variable set (six variables) and analysis techniques
(AHP/DEA-AR model) are implemented in this stage.
4.9 SUMMARY
This chapter was devoted to the methodological issues of this thesis. Firstly, a general
discussion of research philosophies, approaches and strategies was presented. This
study is based on the positivist paradigm, and it recognises the airport industry as an
objective and external entity. Consequently, this research applies a deductive approach
and adopts surveys as a research strategy. It employs the quantifiable quantitative data
analysis method (i.e. DEA models and the AHP/DEA model) to determine the
relationship between airport privatisation policy and efficiency. Because a number of
100
Ch4
airports in different countries are selected as the sample in this research, a cross-
sectional research design is employed to answer the research questions. To acquire the
relative importance of each variable from experts in different areas, in-depth interviews
(i.e. semi-structured interviews), structured questionnaires, and postal questionnaires
are adopted by the researcher as data collection methods. The data analysis methods are
also described. The reasons of using alternative scales are initially discussed. Secondly,
the characteristics of AHP and the nature of alternative scales within AHP are outlined.
Thirdly, the characteristics of DEA, the basic two DEA models are described. Finally,
the framework of the research procedures and methods are presented.
101
Ch 5
Empirical Analysis: AHP Analysis
CHAPTER 5
EMPIRICAL ANALYSIS:
AHP ANALYSIS
The last chapter provided an extensive explanation of the methodological tools and
analysis methods employed in this research. The aim of this chapter is to present the
findings of the AHP questionnaire. This chapter includes two main sections. The first
section addresses the results of the pilot questionnaire that will be used to confirm the
complete AEES which is going to be applied in this research. The second section
presents an overview of the relative AHP weights of the variables and their
classification into different groups. It then concludes with a brief summary of the
analysis of this research.
102
Ch 5
5.1 PRELIMINARY EVALUATION VARIABLES

According to Figure 4.6, the preliminary variable set is going to be established in Stage
I. From the literature review (Section 2.3.4), a range of input and output variables was
suggested. A preliminary selection of nine input variables and five output variables was
made (as shown in Table 5.1) by the author. The following principles have been taken
into consideration when the variables were chosen:
Ø According to the assumption of DEA, the number of DMUs should be at least more
than the product of the input and output variables (Charnes et al. 1985). Ali et al.
(1987) and Bowlin (1987) advised that the number of DMUs should be at least
twice the sum of the input and output variables.
Ø The variables in different hierarchies groups should not have more than seven
variables (Saaty 1980).
Ø The selected variables must be comprehensive enough to provide an effective
evaluation of the operational performance of international airports.
Ø As noted in Section 2.3.4, the selection of variables in the preliminary variable sets
is based on the most widely used variables in the past 20 years (see Table 2.3 and
Table 2.4).
The preliminary variables can be classified into two hierarchies, which include an input
perspective (whose main-criteria consist of airport capacity and financial concerns) and
an output perspective (whose main-criteria includes service performance and financial
performance).
5.2 THE PILOT STUDY PROCESS

The aim of conducting pilot interviews is to ensure the validity and reliability of the
preliminary variables. To address the concern of subjectivity, which is usually one of
the major criticisms faced by research, the AHP method is used after the process of
selecting the variables. Following the guidelines of the research framework (which was
described in Section 4.8), three experts were invited to join individual semi-structured
interviews. They were invited from academia, the airport authority (airport company),
and the civil aviation authority (department of transport). Three of them have been
involved in air transport relevant business or research for more than 20 years.
103
Ch 5
Table 5.1: The preliminary variables of airport efficiency evaluation system

Input
Main
Sub-criteria Definition of the criteria
criteria
The number of full-time equivalent employees directly
Number of employees
employed by the airport.
The number of gates with jet ways and other non jet-way
Number of gates
gates.
Number of runways The available number of runways at each airport.
Airport Number of check-in The number of the desks where passengers check in their
capacity desks bags and cases and have their tickets checked at an airport.
Number of car parking The number of the spaces where people can park their cars
spaces in an airport.
Size of terminal area The total area of passenger terminals.
Length of runway The average runway length of every runway in each airport.
Non-Operational The debt services, capital expenditure and other non-

expenditure operating expenses.
Airport
The financial resources needed to run an airport, including
finances
Operational salaries and benefits, communications and utilities, supplies,
expenditure materials, repairs and maintenance, services and other
expenses.
Output
Main
Sub-criteria Definition of the criteria
criteria
The number of passengers arriving or departing at an airport

Number of passengers
(including terminal passengers and transit passengers).
Service Amount of freight and The weight of property carried on an aircraft and the weight
performance mail of Post Office mail carried.
The number of landings or take-offs of aircraft engaged in

Aircraft movements the transport of passengers, cargo or mail on commercial
terms.
The revenues that are generated by aviation activities such

Aeronautical revenues as landing fees, terminal fees, apron charges, fuel flowage,
Financial fixed base operators (FBOs), rentals and utilities.
performance
Non-aeronautical
The rents, concessions, parking, rental cars, catering, etc.
revenues
104
Ch 5
5.2.1 QUESTIONS IN THE PILOT EXPERT INTERVIEW

A series of questions were asked in the expert pilot interview in order to understand the
entire picture of the current airport industry. The expert questionnaire had three sections:
firstly, some questions about the airport industry were asked; secondly, the experts
were asked about some general ideas regarding an airport performance evaluation;
finally, they were asked to help to select appropriate input and output variables.
The order of the interview questions is presented as follows:

1. The current situation about airport industry:
a. Do you think the airport industry is dramatically changing?
b. Do you think the competition between airports has become more intensified?
c. From your viewpoint, what strategy is the most useful to improve airport
operational efficiency (e.g. Privatisation or Commercialization)?
2. The needs of the airport evaluation system:

a. What are the key variables related to airport efficiency?
b. Do you think every airport needs to measure performance regularly?
c. Do you think every airport needs to develop its own evaluation system?
d. From your viewpoint, which benchmarking technique is the most useful to
evaluate airport operational efficiency?
3. The appropriateness of variables:

a. Which variables are appropriate for measuring airport efficiency in each level?
(Please provide your answer in the following table).
b. Other suggestions about variables.
5.2.2 INTERVIEW RESULT ANALYSIS

From pilot interviews, experts can help to amend variable sets. Based on these
suggestions, the researcher can improve the evaluation system objectively.
Ø Main-Criteria
All experts agreed that the author’s classification of the main criteria into airport
capacity and airport financial concerns from the input perspective and service
performance and financial performance from the output perspective are suitable.
105
Ch 5
Ø Sub-Criteria
Based on the literature review, the preliminarily evaluation variables were selected
by the author. There are nine variables in the input and five variables in the output.
However, according to the limitations of the DEA, the number of DMUs and
variables should be considered. In this study, 24 airports are selected as DMUs (see
Section 4.8). As mentioned in Section 5.1, the number of DMUs should be at least
greater than the product of the input and output variables or at least twice the sum of
the input and output variables. Therefore, in this research, the appropriate number of
variables should not exceed more than six variables in regard to two different
perspectives in order to achieve the first requirement. Consequently, the number of
variables must be reduced. The suggestions about variables in the sub-criteria are
described as follows:
5.2.2.1 INPUT PERSPECTIVE

Ø Airport Capacity
All of the experts agreed that the number of employees, number of gates, number of
runways, and length of runways are suitable for the AEES in this research. The
number of employees can help researchers to understand how many employees are
employed in different airports in order to provide similar services. The number of
gates can help researchers to determine how many aircraft can be served at a specific
time. The number of runways can assist researchers in determining how many
aircraft movements can be handled during a specific period of time. The length of
runways affects the ability to handle larger aircraft; in which case, the airport can
serve more passengers.
Some experts did show their concern about particular variables. For example, only
one expert thought that the number of check-in desks is appropriate for airport
efficiency evaluation, while the other two experts suggested that the size of the
terminal area is similar to this variable; therefore, the researcher only kept one of
these two variables. None of the experts felt that the number of parking spaces is an
important variable for airport evaluation efficiency. One of them suggested instead
the addition of a variable about the number of public transportation routes.
106
Ch 5
Ø Airport Finances
All of the experts agreed that this research should only include the variable of
operational expenditure. This is considered to be appropriate because some airports
(such as Amsterdam and Dubai) have recently tried to develop the concept of an
“Airport City”. Consequently, this new kind of airport may need to produce more
expenditure related to non-operational expenditure, which is not related to airport
operation.
5.2.2.2 OUTPUT PERSPECTIVE

Ø Service performance
The author used three variables (i.e. number of passengers, amount of freight and
mail, and aircraft movements) as sub-criteria when evaluating service performance.
All of the experts felt that the philosophy of a business operation is to learn how to
acquire maximum benefit by applying limited resources. Consequently, these three
variables can help the researcher to determine the efficiency of an airport. Therefore,
these three variables are all suitable.
Ø Financial performance
Currently, non-aeronautical revenue is becoming more important for every airport.
Therefore, all of the experts suggested that total revenue is appropriate when
evaluating airport efficiency.
After the interviews, the experts provided some suggestions to help the researcher to
improve the AEES with reliable and objective variables. Eventually, ten variables were
selected, in input perspective, airport capacity and financial considerations were kept as
the main criteria. Among the sub-criteria, six variables were selected. The number of
car parking spaces, the number of check-in desks, and the non-operational expenditure
were removed because they are less relevant or have similar definitions to those of
other variables.
From the output perspective, the main criteria of service performance and financial
performance were kept. Regarding, the sub-criteria, all of variables were reserved
because of less relevance, aside from non-aeronautical revenue. The new set of ten
variables is illustrated in Figure 5.1. However, in order to conduct sensitivity analysis
107
Ch 5
and answer Research Question 4: Does the number of input and output variables affect
the results for airport efficiency?, another variable set was also employed in this
research. In this variable set, only two hierarchies and six variables were chosen (the
details are shown in Figure 5.2). The variable selection in this set is also based on the
most widely used variables in the past 20 years and suggestions from experts (see
Section 2.4).
Figure 5.1: AHP hierarchies of Variable Set I (ten variables)
Airport Evaluation System

Level 1:
Goal
Input Output
Airport Capacity Airport Service Financial

Level 2: Aspect Finances Performance Performance
Main-‐Criteria
Length of Operational Aircraft Total

Runway Expenditures Movements Revenues
Size of Amount of

Terminal Area Freight and Mails
Number of
Runways Number of
Level 3:
Passengers
Sub-‐Criteria
Number of
Gates
Number of
Employees
Figure 5.2: AHP hierarchies of Variable Set II (six variables)
Airport Evaluation System

Level 1:
Goal
Input Output
Size of Aircraft
Terminal Area Movements
Number of Amount of

Level 2: Runways Freight and Mails
Main-‐Criteria
Number of Number of
Employees Passengers
108
Ch 5
5.3 RELATIVE IMPORTANCE EVALUATION OF VARIABLES

After constructing an AHP hierarchy of airport efficiency evaluation, in stage II, a
structured interview and postal survey (AHP questionnaire) were applied in order to
explore the weights and priority analysis of each variable. A questionnaire entitled
“Airport Efficiency Evaluation Questionnaire” was devised based on the hierarchy
shown in Table 5.1. This questionnaire was designed in terms of AHP, a principle that
makes pair-wise comparisons at the same level. The measurements were made on a
scale from 1 to 9. The definition of each scale is shown in Table 4.5, and the actual
questionnaire is attached in Appendix II. Between both variable sets, only one variable
(i.e. total revenue) is different in regard to the output perspective. Therefore, when
designing the questionnaire for the VSII, the author only needed take the main criteria
of financial performance into consideration. The interviews did not have to address the
relative importance of aircraft movement, amount of freight and mail, and number of
passengers. In the meantime, the relative weights can be introduced from VSI.
In stage II, the data collection was comprised of gathering responses from 36 airport
management specialists in two distinct categories: In practice, questionnaires were
posted to 24 sample airport companies or authorities. In the area of academia, 11
scholars in related fields such as those in charge of airport management or
transportation management were selected from North America, Asia, and Europe, as
selected from an ATRS experts list.
All respondents were given clear instructions prior to filling out the questionnaire. The
survey was carried out in two rounds. The first round was from 2011/03/01 to
2011/03/15; the second round was from 2011/04/01 to 2011/04/15. Details about the
respondents and the response rate are displayed in Table 5.2. There were a total of 35
questionnaires sent out, and 25 questionnaires were collected, giving a response rate of
71.43%.
The weights of each variable were calculated using the answers of the respondents and
were calculated using Super Decision (an AHP software program) and cross checked
using equations programmed with Maple. If the Consistency Index (CI) was < 0.1, then
the consistency in the respondent’s questionnaire was considered to be acceptable;
109
Ch 5
otherwise, the questionnaire has to be returned for re-answering or should be excluded

from effective responses, which explains the reliability of the weight. In this research,
the effective rate was 62.86%, which was achieved for 22 acceptable questionnaires (as
shown in Table 5.2). When using Super Decision, the geometric mean method for
solving the eigenvalue of the pair-wise comparison matrix was used to calculate the
local weight of each level using a spread sheet (Saaty 2003).
Table 5.2: Effectiveness of questionnaires

Copies of Response Copies of Rate of
Region Sent out
return rate effectiveness effectiveness
North
3 2 66.67% 2 66.67%
America
Asia 4 4 100.00% 4 100.00%
Academia
Europe 4 2 50.00% 1 25.00%
Sub-total 11 8 72.73% 7 63.64%
EU 12 8 66.67% 7 58.33%
Practice Asia 12 9 75.00% 8 66.67%
Sub-total 24 17 70.83% 15 62.5%
Total 35 25 71.43% 22 62.86%

Source: organised by author.
5.3.1 THE WEIGHTS ANALYSIS OF VARIABLE SET I (VS I): MAIN-CRITERIA

Table 5.3 introduces the pair-wise comparative matrix for the hierarchy that is
concerned with comparing the relative importance matrix of airport capacity and airport
finances. After calculating the geometric mean of 22 questionnaires, among these two
criteria, all the experts felt that airport finances are more important than airport capacity.
On the left side of Table 5.3, the results show the weight of airport capacity to be
0.3759 and the weight of airport finances to be 0.6241 (the calculations are shown in
Section 4.4).
This means that among these 22 experts, when evaluating airport efficiency, they felt
the importance of airport capacity to be less than airport financial concerns. In addition,
the CI and the CR in this comparison matrix were all equal to 0 and smaller than 0.1;
therefore, the results are considered to be reliable.
110
Ch 5
Table 5.3: Pair-wise comparison matrix and weights for level 2

Input Output
Main Airport Airport Service Financial
Weights Main criteria Weights
criteria capacity finances performance performance
Airport Service
1 0.6022 0.3759 1 1.0019 0.5005
capacity performance
Airport Financial
1.6605 1 0.6241 0.9981 1 0.4995
fiance performance
𝐶𝐼 = 0 < 0.1; 𝐶𝑅 = 0 < 0.1 𝐶𝐼 = 0 < 0.1; 𝐶𝑅 = 0 < 0.1
Table 5.4 shows how the results of the main criteria can be broken down into three
different groups: experts from academia, practice, and overall experts. On the input side,
all of the experts from these two groups felt airport finances to be more important than
airport capacity, and the experts from practice were more concerned about this topic (i.e.
64.5% and 52.45%). There was a difference of opinion on the output side. The experts
from academia argued that service performance in an airport should be emphasised
more than financial performance when evaluating airport efficiency; however, the
airport managers were found to place more emphasis upon financial output than the
academic researchers, probably because airport managers have a responsibility to the
stakeholders (although they still take service output into account).
Table 5.4: Local weights in level two in different groups

Input Output
Service Financial
Airport capacity Airport finances
performance performance
Practice 0.3549 0.6450 0.4133 0.5867
Academia 0.4755 0.5245 0.7349 0.2561
Overall 0.3759 0.6241 0.5005 0.4995
5.3.2 THE WEIGHTS ANALYSIS OF VS I: SUB-CRITERIA

The variables under airport capacity indicate the resources inputted by the airport
authority. Five variables were selected for this variable (i.e. number of employees,
number of gates, number of runways, size of terminal area, and length of runway).
Table 5.5 shows the weights of these sub-criteria. Among the five variables in this area,
the experts felt the number of gates to be the most important variable, followed by size
111
Ch 5
of terminal area and number of runways. The range between these two variables was
less than 0.03 (i.e. 0.1173 and 0.0946). There was less importance gap among other
variables but only number of employees. The CI = 0.0137 and CR = 0.0122 were both
smaller than 0.1; therefore, the results in this comparison matrix are reliable. Among
these input variables, operating expenditure was found to be the most important
variable (0.6241).

Local With Respect to Airport
Sub-criteria (A1) (A2) (A3) (A4) (A5)
Weights Capacity Variables
(A1)
Number of 1 0.2731 0.2140 0.2560 0.3257 0.0606 0.0228
employees
(A2)
Number of 3.6619 1 1.5640 1.4509 2.2048 0.3122 0.1173
gates
(A3)
Number of 4.6721 0.6394 1 0.8472 1.6301 0.2299 0.0864
runways
(A4)
Size of 3.0699 0.6892 1.1804 1 2.0497 0.2516 0.0946
terminal area
(A5)
Length of 3.0699 0.4535 0.6134 0.4879 1 0.1457 0.0547
runway
(B1)
Operating - - - - - 0.6241 0.6241
Expenditure
𝐶𝐼 = 0.0137 < 0.1; 𝐶𝑅 = 0.0122 < 0.1
Table 5.6 shows that the experts in academia felt the size of the terminal area to be the
most important variable in regard to airport capacity, followed by the number of gates
and number of runways. However, the experts selected from practice felt the number of
gates to be the most of essential variable in regard to airport capacity, followed by size
of the terminal area and number of runways.
The possible reason why the experts from academia emphasise this variable more than
others is that they might be concerned about the feelings of the passengers (basically,
they believe that the main aim of an airport is to satisfy passengers). On the other hand,
the experts from practice felt the number of gates to be the most important variable for
an airport because the airport authorities think that more gates means that they can
serve more aircraft. However, the gaps between the second and third-ranked variables
112
Ch 5
in the two different groups were only slightly different. Nevertheless, experts from
these two groups placed less emphasis on the influence of employee numbers on airport
efficiency. Among these input variables, the experts felt operating expenditure to be the
most important variable not only in academia but also in practice.
Table 5.6: Local weights in level 3 in different groups

(A4)
(A1) (A2) (A3) (A5) (B1)
Size of
Number of Number of Number of Length of Operating
terminal
employees gates runways runway expenditure
area
Practice 0.0576 0.3336 0.2239 0.2354 0.1495 0.6450
Academia 0.0586 0.2660 0.2153 0.3233 0.1368 0.5245
Overall 0.0606 0.3122 0.2299 0.2516 0.1457 0.6241

The variables in service performance indicate how many entities (i.e. passengers,
aircrafts, and cargo and mail) can be served by an airport. In this aspect, three variables
were selected: number of passengers, amount of freight and mail, and aircraft
movements. Table 5.7 shows the weights of these three variables as they were awarded
by the experts when considering output in an airport. The number of passengers was
deemed the most important variable, followed by the amount of freight and mail and
aircraft movements, because the main concept of an airport is to transport passengers
and freight. In addition, the range between the first and the other two ranked variables
was quite significant. The CI = 0.0055 and CR = 0.0095, both were smaller than 0.1,
indicating the results to be reliable. Among these output variables, total revenue was
found to be the most important variable (0.4995).
Table 5.8 shows that all of the experts in these two group agreed that when reviewing
an airport output, the number of passengers is the most essential variable, followed by
the amount of freight and mail and aircraft movements. The importance of the number
of passenger was found to be much higher than that of the other variables.
113
Ch 5

With Respect
Sub-Criteria (C1) (C2) (C3) Local Weights to Service
Performance
(C1)
Number of 1 1.7045 3.2193 0.5206 0.2605
passengers
(C2)
Amount of freight 0.5867 1 2.3735 0.3296 0.1649
and mail
(C3)
0.3106 0.4213 1 0.1498 0.1498
Aircraft movements
(D1)
- - - 0.4995 0.4995
Total revenue
𝐶𝐼 = 0.0055 < 0.1; 𝐶𝑅 = 0.0095 < 0.1
Table 5.8: Local weights in level 3 in different groups

(C1) (C2) (C3)
Number of Amount of Freight Aircraft
Passengers and Mail Movements
Practice 0.5389 0.3232 0.1379
Academia 0.4715 0.3801 0.1480
Overall 0.5206 0.3296 0.1498
5.3.3 WEIGHT ANALYSIS OF VARIABLE SET II (VS II): MAIN-CRITERIA

In order to conduct sensitivity analysis in this research, VS II (six variables) were
adopted to calculate the weight of each variable.

In this variable set, three input variables were selected: number of employees, number
of runways, and size of terminal area. Among these input variables, experts felt the size
of the terminal area to be the most important variable, followed by the number of
runways and number of employees. From Table 5.9, it can be seen that the importance
of the number of employees is much lower than that for the other two variables. In
comparison with VS I, the number of employees is once again the less important
variable in the input aspect. The CI = 0.0126 and CR = 0.0217 are both smaller than 0.1.
114
Ch 5
Table 5.9: Pair-wise comparison matrix and weights in input aspect

Criteria (A1) (A2) (A3) Local Weights
Number of
(A1) 1 0.2140 0.2560 0.1047
employees
Number of
(A2) 4.6721 1 0.8472 0.4362
runways
Size of
(A3) 4.6721 3.9066 1 0.4590
terminal area
𝐶𝐼 = 0.0126 < 0.1; 𝐶𝑅 = 0.0217 < 0.1
The results from Table 5.10 can be analysed in three groups. The experts from
academia felt the size of the terminal area to be the most important input variable,
followed by the number of runways and number of employees. However, the experts
from practice felt the number of runways to be the most essential input variable,
followed by the size of terminal area and number of employees. The overall ranking of
these variables were the same as VS I (Table 5.5). The difference between the first and
second ranked variables was only slightly different. Nevertheless, all of the experts
from these two different groups believed that the number of employees does not
significantly influence airport efficiency.
Table 5.10: Weights in different groups

(A1) (A2) (A3)
Number of Number of Size of
Employees Runways Terminal Area
Practice 0.1035 0.4505 0.4459
Academia 0.1071 0.4059 0.4870
Overall 0.1047 0.4362 0.4590

In this variable set, three output variables were selected: number of passengers, amount
of freight and mail, and aircraft movements. These three variables were the same as the
variables in the service performance area of VSI. In VSII, financial performance was
not taken into consideration. Therefore, there was no need to ask experts to answer
another AHP questionnaire survey among these three variables. The relative weights of
these three variables are shown in Table 5.7. The results from Table 5.8 are also
115
Ch 5
presented there. All of the experts from these two groups agreed that the variable of
aircraft movements is less important for airport efficiency.
5.4 OVERALL WEIGHTS OF AIRPORT PERFORMANCE EVALUATION

The overall weights of each variable determine their integral priority with respect to
airport efficiency as a whole, which can be calculated from the local weight of each
aspect and their variables using the weighting method and the AHP approach
assumption.
These weights additionally enable policy makers to have a more detailed understanding
of the relative importance among these variables in the situation as a whole, and the
overall weights can also serve as an important reference when they make a decision on
the priorities for strategy formulation and when they decide on action items when there
are multiple goals to be considered with limited resources. In this research, these
weights are conducted using DEA analysis. By means of this, an objective AEES can
be established, and a more accurate airport efficiency result can be found.
5.4.1 OVERALL WEIGHTS OF VS I

The overall weights of the VS I are shown in Table 5.11. Among these variables, the
input side operational expenditure, which belongs to the financial area, is the highest
overall weighted variable. This indicates that operational expenditure should be taken
into consideration in the first place for airport efficiency evaluation. The second overall
weighted variable is the number of gates, and the third weighted variable is the size of
the terminal area, which suggests that when evaluating airport efficiency, these two
variables should be considered over others in regard to airport capacity.
On the output side, total revenue, which belongs to the area of financial performance, is
the highest overall weighted variable. This indicates that total revenues should first be
taken into consideration on the output side for airport efficiency evaluation. The second
overall weighted variable is the number of passengers, and the third weighted variable
is the amount of freight and mail. This suggests that when evaluating an airport, the
airport authority should consider these two highest variables first.
116
Ch 5
Table 5.11: Weights of airport efficiency evaluation system (VS I)

Input
Local Overall
Main-Criteria Weights Sub-Criteria
weights weights
Number of
(A1) 0.0607 0.0228 (6)
employees
(A2) Number of gates 0.3122 0.1173 (2)
Airport
0.3759 (A3) Number of runways 0.2299 0.0864 (4)
capacity
(A4) Size of terminal area 0.2516 0.0946 (3)
(A5) Length of runway 0.1457 0.0547 (5)
Airport Operational
0.6241 (B1) 0.6241 0.6241 (1)
finance expenditure
Output
Local Overall
Main-Criteria Weights Sub-Criteria
weights weights
Number of
(C1) 0.5206 0.2605 (2)
passengers
Service Amount of freight

0.5005 (C2) 0.3296 0.1649 (3)
performance and mail
(C3) Aircraft movements 0.1498 0.0750 (4)
Financial
0.4995 (D1) Total revenue 0.4995 0.4995 (1)
performance
5.4.2 OVERALL WEIGHTS OF VS II

The overall weights of the VS II are shown in Table 5.12. Among these variables, in
terms of the input perspective, the size of terminal area is the highest overall weighted
variable, which indicates that it should be taken into consideration first in evaluation of
the input aspect of airport performance. This is followed by the weighted variables,
number of runways and number of employees.
In terms of the output perspective, number of passengers is the highest overall weighted
variable, which indicates that it should first be taken into consideration in the
evaluation of the output aspect of airport efficiency. This is followed by the overall
weighted variables amount of freight and mail and aircraft movements.
117
Ch 5
Table 5.12: Weights of airport efficiency evaluation system (VS II)

Input Output
Criteria Weights Criteria Weights
(A1) Number of Number of

0.1047 (3) (B1) 0.5206 (1)
employees passengers
(A2) Number of (B2) Amount of

0.4362 (2) 0.3296 (2)
runways freight and mail
(A3) Size of (B3) Aircraft

0.4590 (1) 0.1498 (3)
terminal Area movements
5.5 OVERALL WEIGHTS BASED ON AN ALTERNATIVE SCALE

In this section, five other alternative scales which can be used in the AHP method are
compared with the Saaty 1-9 scale. Each of the pair-wise comparison matrices which
were mentioned in previous sections need to be transformed into the alternative
judgement scale values since these different scales are 9-unit and bounded scales and,
therefore, it is easy to map pair-wise comparisons from exact positions on the 1-9 scale
to any of these scales.
For example, referring to Table 4.7, in respect of the 1-9 scale, the verbal statement
moderate importance, is associated with the scale values 3 (1-9), 2.7183 (𝑒 !!! ),
4(2!!! ), 1.2857 (9/9-9/1), 1.5000 (10/10-18/2), and 1.2536 (ø mapping) respectively. If
the scale value is 5, then the position on the Saaty 1-9 scale is the 5th. The value on the
5th position of the alternative judgement scale value is 5 (1-9), 7.3891 (𝑒 !!! ), 16 (2!!! ),
1.8000 (9/9-9/1), 2.3333 (10/10-18/2), and 1.6125 (ø mapping). The preference
evaluation procedure outlined in Section 4.4 can also be used in determining the
alternative judgement scales. The weights can easily be calculated on the alternative
judgement scale by transforming the importance from the questionnaire.
5.5.1 WEIGHTS OF THE VARIABLES IN VS I

The differences in the set of weights between the alternative scales and the comparisons
in respect to the two aspects in the main criteria are provided in Table 5.13. The variety
is identified clearly in Figure 5.3. In Table 5.13, the weight of the airport finances is the
118
Ch 5
most important among these experts, but it also shows that there are differences in the
weight which is calculated using alternative scales.
Airport capacity ranges from 0.4564 (which is calculated by scale 4) to 0.3059 (which
is calculated by scale 3). The weight of financial variable ranges from 0.6941 (which is
calculated by scale 3) to 0.5355 (which is calculated by scale 6). The weight of criteria
for airport capacity and finances are computed by means of alternative judgement
scales, and they are only slightly different. By conducting scale 2!!! , the gap is the
found to be the largest, and the gap in scale 6 is found to be the smallest.
Table 5.13: Weights of main-criteria variables in input perspective

1 2 3 4 5 6
!!! !!! ∅ mapping
Main-Criteria 1-9 𝑒 2 9/9-9/1 10/10-18/2
(A)
0.3759 0.3564 0.3059 0.4564 0.4374 0.4645
Airport capacity
(B)
0.6241 0.6436 0.6941 0.5436 0.5626 0.5355
Airport finances
Figure 5.3: Weights of main-criteria Figure 5.4: Weights of sub-criteria

0.8 0.4
0.7 0.35 Sub-
0.6 0.3 criteria
Main-
0.5 criteria 0.25 (A1)
0.4 0.2 (A2)
(A) (A3)
0.3 0.15
0.2 (A4)
(B) 0.1
0.1 (A5)
0.05
0
1 2 3 4 5 6 scale 0
Scale
1 2 3 4 5 6
In Table 5.13, if the 6 scales are separated into two groups (scale 1-3) and (scale 4-6),
the results show identical ranking orders and similar sets of weights in these two groups.
This is perhaps because of the verbal statement of importance. From Table 4.7, it can
be seen that the values which represent importance among scale (1-9), (𝑒 !!! ), and
(2!!! ), are similar and that the values among scale (9/9-9/1), (10/10-18/2), and (ø
mapping) are similar, too. In this case, finances can be positioned at moderate
importance as compared to airport capacity. From Table 4.7 the moderate importance is
119
Ch 5
associated with the scale values 3(1-9), 2.7183 (𝑒 !!! ), 4(2!!! ), 1.2857(9/9-9/1),
1.5000(10/10-18/2), and 1.2536 (ø mapping) respectively. If using these value to do the
calculation, the results will show the weights computed by scale (1-9), (𝑒 !!! ), and
(2!!! ) will close, and the weights computed by scale (9/9-9/1), (10/10-18/2), and (ø
mapping). This situation also occurs as shown in the following tables.
The weights of the sub-criteria from the input perspective are calculated and listed in
Table 5.14. The variety can be identified clearly in Figure 5.4. In Table 5.14, the overall
weights of each variable are represented in the value in brackets, and the number in
each column means the rank of each sub-criterion. From Figure 5.4, by conducting
scale 3, the difference is the largest and in scale 6 is the smallest. The weight of the
number of gates is the most important variable, even when calculated by alternative
scales as an airport capacity variable. On the other hand, the number of employees is
the least important variable. The ranking of these five variables in different scales are
all the same. Table 5.14 also shows that the weights that are computed by scales 1 to 3
are close and that the weights which are computed by scale 4 to 6 are also close.
Table: 5.14: Weights of sub-criteria variables in input perspective

1 2 3 4 5 6
∅
Sub-Criteria 1-9 𝑒 !!! 2!!! 9/9-9/1 10/10-18/2
mapping
(A1) 0.0607 0.0349 0.0166 0.1143 0.0953 0.1320
Number of (0.0228) (0.0124) (0.0051) (0.0522) (0.0417) (0.0613)
employees 6 6 6 6 6 6
(A2) 0.3122 0.3236 0.3621 0.2417 0.2578 0.2339
Number of (0.1173) (0.1153) (0.1108) (0.1205) (0.1128) (0.1087)
gates 2 2 2 2 2 2
(A3) 0.2299 0.2316 0.2273 0.2220 0.2254 0.2173
Number of (0.0864) (0.0825) (0.0695) (0.1053) (0.0986) (0.1009)
runways 4 4 4 4 4 4
(A4) 0.2516 0.2725 0.2842 0.2336 0.2419 0.2260
Size of (0.0946) (0.0971) (0.0869) (0.1106) (0.1058) (0.1050)
terminal area 3 3 3 3 3 3
(A5) 0.1457 0.1374 0.1099 0.1885 0.1796 0.1908
Length of (0.0547) (0.0490) (0.0336) (0.0825) (0.0786) (0.0886)
runway 5 5 5 5 5 5
(B1)
0.6241 0.6436 0.6941 0.5436 0.5626 0.5355
Operational
1 1 1 1 1 1
expenditure
120
Ch 5

The weights of the main criteria for output are listed in Table 5.15, and the variations
are further illustrated in Figure 5.5. The weights between criteria service performance
and financial performance using scale 3 indicate that the variety is the largest and that
in scale 1 it is the smallest. By means of scale 1, it can be seen that service performance
of an airport is more important than financial performance. On the other hand, by
adopting the rest of the five scales have reverse results.
Table 5.15: Weights of main-criteria variables in output perspective

1 2 3 4 5 6
Main ∅
1-9 𝑒 !!! 2!!! 9/9-9/1 10/10-18/2
Criteria mapping
(C)
0.5005 0.4943 0.4921 0.4976 0.4973 0.4984
Service
1 2 2 2 2 2
performance
(D)
0.4995 0.5057 0.5079 0.5024 0.5027 0.5016
Financial
2 1 1 1 1 1
performance
Figure 5.5: Weights of main criteria Figure 5.6: Weights of sub-criteria

0.51 0.7
0.6
0.505 Sub-
0.5 criteria
0.5
0.4 (C1)
0.495 (C)
0.3 (C2)
(D)
0.49 (C3)
0.2
0.485 0.1
0.48 0
1 2 3 4 5 6 Scale 1 2 3 4 5 6 Scale
Pöyhönen and Hämäläinen (2001) revealed that the rank of variables may remain the
same even when there remains a difference in the criteria weights. However, from
Figure 5.5, a very interesting point needs to be discussed. In the previous figures
(Figure 5.3 and Figure 5.4), the rank of the variable is not changed when conducting
alternative scales, but in Table 5.15 and Figure 5.5, the criteria for service performance
and financial performance are in different order when using the 1-9 scale to calculate
the weight. From this, although the rank changes, the evidence still shows that
alternative scales can help provide different weights. In this case, the difference
between service performance and financial performance by using not only scale 1-9 but
121
Ch 5
also other scales is slight. In general, all experts felt these two variables both to be very
important when evaluating airport efficiency.
The weights for the main criteria from the output perspective are listed in Table 5.16,
and the differences are illustrated in Figure 5.6. The weights among these three
variables on alternative judgement are only slightly different. By conducting scale 3,
the difference is the largest, and scale 6 is the smallest. In addition, the ranking of these
five variables in different scales are all the same, and among these output variables,
total revenue is the most import variables when evaluation airport efficiency.
Table 5.16: Weight of sub-criteria variables in output perspective

1 2 3 4 5 6
∅
Sub-Criteria 1-9 𝑒 !!! 2!!! 9/9-9/1 10/10-18/2
mapping
(C1) 0.5206 0.5402 0.6078 0.3962 0.4245 0.3850
Number of (0.2605) (0.2670) (0.2991) (0.2213) (0.2111) (0.1919)
passengers 2 2 2 2 2 2
(C2)
0.3296 0.3327 0.3104 0.3454 0.3462 0.3432
Amount of
(0.1649) (0.1644) (0.1527) (0.1803) (0.1722) (0.1711)
freight and
3 3 3 3 3 3
mail
(C3) 0.1498 0.1271 0.0818 0.2584 0.2292 0.2717
Aircraft (0.075) (0.0628) (0.0402) (0.1171) (0.1140) (0.1354)
movements 4 4 4 4 4 4
(D1)
0.4995 0.5057 0.5079 0.5024 0.5027 0.5016
Total
1 1 1 1 1 1
revenue
5.5.2 WEIGHTS OF THE VARIABLES IN VS II

The weights of variables from the input perspective are calculated using alternative
scales and are listed in Table 5.17. The differences are identified clearly in Figure 6.7.
The weights among these criteria (which are computed by means of alternative
judgement scales) are distributed significantly. By conducting scale 3, the variety is the
largest, and using scale 6, it is the smallest. Using alternative scales, the number of
runways is the most important variable as compared to others.

Table 5.17 also displays the weight of variables in the output. The differences are
identified clearly in Figure 5.7. The weight among these three variables is distributed
122
Ch 5
smoothly, and the rankings for each variable are the same. From Figure 5.7 and Figure
5.8, it can also be seen that the weights computed by scales 1 to 3 are similar, and the
weights computed by scales 4 to 6 are similar.
Table 5.17: Weights of the variables in VS II

Input
1 2 3 4 5 6
10/10- ∅
Criteria 1-9 𝑒 !!! 2!!! 9/9-9/1
18/2 mapping
(A)
0.1077 0.0634 0.0303 0.1985 0.1678 0.2280
Number of
3 3 3 3 3 3
employees
(B)
0.4590 0.4825 0.5052 0.4067 0.4226 0.3899
Number of
1 1 1 1 1 1
runways
(C)
0.4334 0.4541 0.4645 0.3949 0.4095 0.3821
Size of
2 2 2 2 2 2
terminal area
Output
1 2 3 4 5 6
10/10- ∅
Criteria 1-9 𝑒 !!! 2!!! 9/9-9/1
18/2 mapping
(D)
0.5206 0.5402 0.6078 0.3962 0.4245 0.3850
Number of
2 2 2 2 2 2
passengers
(E)
0.3296 0.3327 0.3104 0.3454 0.3462 0.3432
Amount of freight
3 3 3 3 3 3
and mail
(F)
0.1498 0.1271 0.0818 0.2584 0.2292 0.2717
Aircraft
4 4 4 4 4 4
movements
Figure 5.7: Weights of criteria in input aspect Figure 5.8: Weights of criteria in output aspect
0.6 0.7
0.6
0.5 Sub-
0.5 criteria
0.4
(A) 0.4 (C1)
0.3 (B) 0.3 (C2)
0.2 (C) (C3)
0.2
0.1
0.1
0 0
1 2 3 4 5 6 1 2 3 4 5 6 Scale
5.5.3 ALTERNATIVE SCALES IN DIFFERENT GROUPS: VS I
This section discusses the results in three different groups: experts in academia, experts
in a practical area, and overall experts.
123
Ch 5
5.5.3.1 MAIN-CRITERIA
In regard to the input perspective of VS I (see Table 5.18), when employing alternative
scales, experts in practice felt the airport finances to be more important than airport
capacity. This was the case with academia, too. These results are very similar to the
overall results. The experts all believed that on the input side, finances in an airport
operation should be given more consideration than airport capacity. Table 5.18 also
shows that the weights which were computed using scales 1 to 3 are close and that the
weights computed using scales 4 to 6 are also close.
Table 5.18: Weights of main-criteria in different scales by groups: input perspective

1 2 3 4 5 6
∅
Main-Criteria 1-9 𝑒 !!! 2!!! 9/9-9/1 10/10-18/2
mapping
Practice
0.3549 0.3245 0.2657 0.4445 0.4219 0.4555
Airport capacity
2 2 2 2 2 2
0.6450 0.6756 0.7343 0.5555 0.5781 0.5445
Airport finances
1 1 1 1 1 1
Academia
0.4222 0.4290 0.4023 0.4821 0.4711 0.4838
Airport capacity
2 2 2 2 2 2
0.5778 0.5710 0.5977 0.5179 0.5289 0.5161
Airport finances
1 1 1 1 1 1
Overall
0.3759 0.3564 0.3059 0.4564 0.4374 0.4645
Airport capacity
2 2 2 2 2 2
0.6241 0.6436 0.6941 0.5436 0.5626 0.5355
Airport finances
1 1 1 1 1 1
From the output perspective (see Table 5.19), the experts from academia believed that
service performance in an airport should be given more consideration than financial
performance. Although this result is different from the experts in a practice, the gap is
not very significant. The overall results are influenced by experts selected from practice
rather than from academia. From Table 5.19, the weights in different groups can help to
explain why the overall weight and rank which are calculated using scale 1 is different
from those of other scales. Experts in academia felt service performance to be much
more important than financial performance (i.e. 0.7439 and 0.2561), but experts in
practice felt that financial performance is important than service performance (i.e.
0.5867 and 0.4133). After combining the weights in both groups, the results show that
124
Ch 5
the importance of service performance is a little higher than that of financial

performance. The situation is also occurred among the other scales.
Table 5.19: Weights of main-criteria in different scales by groups:

output perspective
1 2 3 4 5 6
∅
Main-Criteria 1-9 𝑒 !!! 2!!! 9/9-9/1 10/10-18/2
mapping
Practice
Service 0.4133 0.3933 0.3542 0.4830 0.4542 0.4741
performance 2 2 2 2 2 2
Financial 0.5867 0.6067 0.6458 0.5170 0.5458 0.5259
Academia
Service 0.7439 0.7625 0.8344 0.6193 0.6193 0.5679
Financial 0.2561 0.2375 0.1656 0.3807 0.3807 0.4321
Overall
Service 0.5005 0.4943 0.4921 0.4976 0.4973 0.4984
Financial 0.4995 0.5057 0.5079 0.5024 0.5027 0.5016
5.5.3.2 SUB-CRITERIA
In the input perspective of VS I, there is a different opinion about the importance of the
number of gates and the size of the terminal area among the experts from academia and
practice. The ranking of other variables in different scales are all the same and are only
different on weight values (further details are shown in Table 5.20).
In the output perspective of VS I, the ranking of other variables in different judgement

scales are all the same and are only different in regard to weight values. (Details are
shown in Table 5.21). Table 5.20 and Table 5.21 also show that the weights computed
using scales 1 to 3 are similar and that the weights computed by scales 4 to 6 are close.
125
Ch 5
Table 5.20: Weights of sub-criteria of different scales by groups:

input perspective
1 2 3 4 5 6
∅
Sub-Criteria 1-9 𝑒 !!! 2!!! 9/9-9/1 10/10-18/2
mapping
Practice
0.0576 0.0311 0.0140 0.1089 0.0904 0.1279
Number of employees (0.0205) (0.0101) (0.0037) (0.0484) (0.0381) (0.0582)
6 6 6 6 6 6
0.3336 0.3506 0.4015 0.2498 0.2688 0.2401
Number of gates (0.1184) (0.1138) (0.1067) (0.1111) (0.1134) (0.1093)
2 2 2 2 2 2
0.2239 0.2289 0.2226 0.2235 0.2257 0.2180
Number of runways (0.0795) (0.0742) (0.0591) (0.0994) (0.0952) (0.0993)
4 4 4 4 4 4
0.2354 0.2486 0.2489 0.2276 0.2332 0.2217
Size of terminal area (0.0836) (0.0807) (0.0661) (0.1012) (0.0984) (0.1010)
3 3 3 3 3 3
0.1495 0.1407 0.1129 0.1902 0.1820 0.1924
Length of runway (0.0531) (0.0457) (0.0300) (0.0845) (0.0768) (0.0876)
5 5 5 5 5 5
Operational 0.6450 0.6756 0.7343 0.5555 0.5781 0.5445
expenditure 1 1 1 1 1 1
Academia
0.0670 0.0439 0.0230 0.1267 0.1067 0.1410
6 6 6 6 6 6
0.2679 0.2675 0.2798 0.2243 0.2349 0.2210
Number of gates (0.1131) (0.1148) (0.1126) (0.1081) (0.1106) (0.1069)
3 3 3 3 3 3
0.2410 0.2340 0.2308 0.2183 0.2238 0.2156
4 4 4 4 4 4
0.2877 0.3257 0.3650 0.2462 0.2604 0.2352
2 2 2 2 2 2
0.1365 0.1289 0.1013 0.1845 0.1742 0.1872
Length of runway (0.0576) (0.0553) (0.0408) (0.0890) (0.0820) (0.0906)
5 5 5 5 5 5
Operational 0.5778 0.5710 0.5977 0.5179 0.5289 0.5161
Overall
0.0607 0.0349 0.0166 0.1143 0.0953 0.1320
6 6 6 6 6 6
0.3122 0.3236 0.3621 0.2417 0.2578 0.2339
Number of gates (0.1173) (0.1153) (0.1108) (0.1205) (0.1128) (0.1087)
2 2 2 2 2 2
0.2299 0.2316 0.2273 0.2220 0.2254 0.2173
4 4 4 4 4 4
0.2516 0.2725 0.2842 0.2336 0.2419 0.2260
3 3 3 3 3 3
0.1457 0.1374 0.1099 0.1885 0.1796 0.1908
Length of runway (0.0547) (0.0490) (0.0336) (0.0825) (0.0786) (0.0886)
5 5 5 5 5 5
Operational 0.6241 0.6436 0.6941 0.5436 0.5626 0.5355
126
Ch 5
Table 5.21: Weights of sub-criteria of different scales by groups:

output perspective
1 2 3 4 5 6
10/10- ∅
Sub-Criteria 1-9 𝑒 !!! 2!!! 9/9-9/1
18/2 mapping
Practice
0.5389 0.5604 0.6327 0.4026 0.4339 0.3904
Number of passengers (0.2227) (0.2204) (0.2241) (0.1885) (0.1971) (0.1851)
2 2 2 2 2 2
0.3232 0.3251 0.2974 0.3457 0.3455 0.3434
Amount of freight
(0.1335) (0.1279) (0.1053) (0.1619) (0.1569) (0.1628)
and mail
3 3 3 3 3 3
0.1379 0.1144 0.0699 0.2517 0.2206 0.2662
Aircraft movements (0.0570) (0.0450) (0.0248) (0.1179) (0.1002) (0.1263)
4 4 4 4 4 4
0.5867 0.6067 0.6458 0.5318 0.5458 0.5259
Total revenues
1 1 1 1 1 1
Academia
0.4804 0.4954 0.5511 0.3826 0.4043 0.3736
2 2 2 2 2 2
0.3417 0.3466 0.3359 0.3445 0.3472 0.3427
Amount of freight
(0.2326) (0.2433) (0.2574) (0.1930) (0.2044) (0.1886)
and mail
3 3 3 3 3 3
0.1779 0.1580 0.1130 0.2729 0.2485 0.2837
4 4 4 4 4 4
0.3193 0.2979 0.2336 0.4396 0.4112 0.4497
Total revenue
1 1 1 1 1 1
Overall
0.5206 0.5402 0.6078 0.3962 0.4245 0.3850
2 2 2 2 2 2
0.3296 0.3327 0.3104 0.3454 0.3462 0.3432
Amount of freight
(0.1649) (0.1644) (0.1527) (0.1803) (0.1722) (0.1711)
and mail
3 3 3 3 3 3
0.1498 0.1271 0.0818 0.2584 0.2292 0.2717
4 4 4 4 4 4
0.4995 0.5057 0.5079 0.5024 0.5027 0.5016
Total revenue
1 1 1 1 1 1
5.5.4 ALTERNATIVE SCALES IN DIFFERENT GROUPS: VS II

In the input perspective of VS II, there is a difference of opinion about the importance
of the number of runways and the size of the terminal area among the experts from
academia and practice. The ranking of other variables in different scales are all the
same and are only different with regard to weight values (further details are shown in
Table 5.22).
127
Ch 5
In the output perspective of VS II, the ranking of all variables in different scales are all
the same and only differ in weight values (further details are shown in Table 5.23).
Table 5.22: Weights of different scales by groups: input perspective

1 2 3 4 5 6
10/10- ∅
Sub-Criteria 1-9 𝑒 !!! 2!!! 9/9-9/1
18/2 mapping
Practice
0.1035 0.0550 0.0248 0.1891 0.1569 0.2194
Number of employees
3 3 3 3 3 3
0.4505 0.4751 0.4914 0.4167 0.4238 0.3921
Number of runways
1 1 1 1 1 1
0.4459 0.4699 0.4839 0.3942 0.4192 0.3885
Size of terminal area
2 2 2 2 2 2
Academia
0.1071 0.0646 0.0309 0.2050 0.1717 0.2319
Number of employees
3 3 3 3 3 3
0.4059 0.3851 0.3670 0.3744 0.3793 0.3652
Number of runways
2 2 2 2 2 2
0.4870 0.5504 0.6021 0.4206 0.4490 0.4030
1 1 1 1 1 1
Overall
0.1077 0.0634 0.0303 0.1985 0.1678 0.2280
Number of employees
3 3 3 3 3 3
0.4590 0.4825 0.5052 0.4067 0.4226 0.3899
Number of runways
1 1 1 1 1 1
0.4334 0.4541 0.4645 0.3949 0.4095 0.3821
2 2 2 2 2 2
128
Ch 5
Table 5.23: Weights of different scales by groups: output perspective

1 2 3 4 5 6
10/10- ∅
Sub-Criteria 1-9 𝑒 !!! 2!!! 9/9-9/1
18/2 mapping
Practice
Number of 0.5389 0.5604 0.6327 0.4026 0.4339 0.3904
Amount of freight and 0.3232 0.3251 0.2974 0.3457 0.3455 0.3434
mail 2 2 2 2 2 2
0.1379 0.1144 0.0699 0.2517 0.2206 0.2662
Aircraft movements
3 3 3 3 3 3
Academia
Number of 0.4804 0.4954 0.5511 0.4083 0.4043 0.3629
Amount of freight and 0.3417 0.3466 0.3359 0.3481 0.3472 0.3428
mail 2 2 2 2 2 2
0.1779 0.1580 0.1130 0.2436 0.2484 0.2943
Aircraft movements
3 3 3 3 3 3
Overall
Number of 0.5206 0.5402 0.6078 0.3962 0.4245 0.3850
Amount of freight 0.3296 0.3327 0.3104 0.3454 0.3462 0.3432
and mail 3 3 3 3 3 3
0.1498 0.1271 0.0818 0.2584 0.2292 0.2717
Aircraft movements
4 4 4 4 4 4
5.6 CROSS DISCUSSION AND ANALYSIS

Comparing the weight of each variable with those reported in the literature review in
Chapter 2, there are some points that can be discussed.
Firstly, from the literature, the variable number of employees has been the most widely
used parameter when evaluating airport efficiency (in total, 27 of 66 papers). However,
in this research, experts from both academia and practice felt this variable to be the
least important variable because the experts felt that the number of employees can be
included in operational expenditure. Therefore, other variables (such as size of terminal
area, number of runways, and operational expenditure) were all considered to be first
priority.
Furthermore, Table 5.11 also shows that the experts who were interviewed in this
research felt that more emphasis should be placed on the finances, but the literature
review shows that only around 10 papers used financial variables. This might be a
129
Ch 5
result of data accessibility or research limitations. It is widely believed that financial

data is the most difficult data to acquire. Secondly, the weights of each variable show
significant differences when using alternative scales to calculate variable weights. From
the results, there is proof that when conducting alternative scales in AHP analysis, the
weights of each variable will change. In addition, in order to determine the further
influence of alternative scales on airport efficiency, it is necessary to combine the
weights of each variable with the DEA model.
5.7 SUMMARY
This chapter demonstrated how the earlier concepts and analysis methods (see Chapter
2 and Chapter 4) were applied to acquire the weights of each variable. This chapter also
provided a description of the interview and questionnaire techniques that were adopted
in this study. The results of AHP analysis indicate that experts from the practical area
placed more emphasis on the value of financial variables than the experts from
academia. The use of alternative judgement scales in this research and their application
were also described in this chapter. From this comparison of the results using a 1-9
scale, it appears that different scales can obtain different weights. Therefore, this
provides strong evidence to support the view that no single benchmark for the choice of
scales should be used in this research. In the next chapter, the weights that are
calculated by alternative scales are combined with the DEA model to compute the
relative efficiency of the 24 sample airports.
130
Ch 6
Airport Efficiency Analysis I
CHAPTER 6
AIRPORT EFFICIENCY ANALYSIS I:

BASIC DEA MODELS AND
INTEGRATED AHP/DEA MODEL

This chapter presents an analysis of the relative efficiency scores of the sample airports
by means of variable weights. The first section describes the concepts and analysis
process for the efficiency scores. The second and third sections present the efficiency
scores as computed using basic DEA models. Section four calculates the efficiency
scores by using an integrated AHP/DEA model. Finally, the empirical results cross
discussion and conclusions are provided.
131
Ch 6
6.1 CONCEPTS OF AIRPORT EFFICIENCY ANALYSIS

A set of balanced panel data are required to evaluate the efficiency of the 24 sample
airports examined in this research. The data set is acquired from two major sources,
which include the 2010 ATRS airport benchmarking report and the annual reports
published by the individual airports. In order to conduct sensitivity analysis, this
research uses two variables sets (a ten variables set and a six variables set) to calculate
airport efficiency.
6.1.1 THE DEA MODELS IN THIS RESEARCH

There are two basic models in the DEA method (the CCR and BCC). The choice of a
DEA model depends on the assumptions regarding the variable sets to be employed and
on the prior results about the industry to be studied. The variable sets have to describe
the activities of the units in the best possible way. It is particularly important to have
some idea about the hypothetical returns to scale that exist in the industry (Martin and
Roman 2001). The literature review in Section 4.6 indicated that DEA can be carried
out under the assumption of CRS for the inputs and outputs in the DEA-CCR model or
by introducing a scale constraint into the model under conditions of VRS in a DEA-
BCC model. The VRS scores calculate pure technical efficiency only. However, the
CRS index is composed of a non-additive combination of pure technical and scale
efficiencies (which is usually called technical efficiency) (Cooper et al. 2006; Barros
and Dieke 2007).
6.1.2 THE ORIENTATION OF THE DEA MODEL

Once the selection of a DEA model has been made, an orientation of the model to
determine the measurement of the efficiency is needed. There are two basic orientations,
which include both input and output. An input orientation focuses on the proportional
decrease of the input vector while the output orientation adjusts the proportional
increase of the output (Martin and Roman 2001). The samples involved in this research
determine the selection of the orientation. An output orientation is employed in this
research because once an airport has invested in something such as the building of new
runways or new terminals, it is difficult for airport authorities to disinvest to save costs
by amending their input variables (i.e. runways, terminal buildings, or employees) to
invalidate the input orientation (Gillen and Lall; 1997; Oum 2006).
132
Ch 6
6.1.3 THE DEA ANALYSIS PROCESS

The process for analysing the operating efficiency is divided into five parts:
(1) Validity test of the variables
Isotonicity is one of the tests that can be performed to check the validity of the
inputs and outputs chosen. This concept means that the outputs should be
significant and positively correlated with the inputs (Charnes et al. 1985). In this
research, a correlation coefficients analysis is applied to determine the relations
between the input and output variables.
(2) Analysis of operating efficiency

The five evaluation items include the categories of airport ownership and
governance, the technical efficiency (i.e. CRS efficiency), the pure technical
efficiency (i.e. VRS efficiency), the scale efficiency, and the returns to scale. The
technical efficiency can be obtained from the CCR model while the BCC model
can be used to obtain the pure technical efficiency. The technical efficiency is then
divided by the pure technical efficiency in order to get the scale efficiency. The
obtained efficiency values are used then to analyse the operating efficiency of each
airport. In addition, Coelli et al. (2005) revealed the concept that the assumption in
a DEA-CCR model is that all DMUs are operating at an optimal scale. However,
imperfect competition, government regulations, and constraints on finance may all
cause a firm to not operate at an optimal scale. When not all DMUs are operating at
the optimal scale, the use of the DEA-BCC model specification is more suitable to
assess airport efficiency. Therefore, in the following section, more emphasis is
placed on the results from the DEA-BCC (i.e. output-oriented) model.
(3) Clustering analysis referral

The purpose of the clustering analysis referral is to view the relatively efficient
airports, which are then used as the references and frequencies that improve the
efficiencies of relatively inefficient enterprises (Zhu 2009). The more an airport is
referred, the more efficient the airport is in the DMU group; therefore,
improvements in the operating flow of an inefficient airport can be more reachable.
133
Ch 6
(4) Slack variable analysis

A slack variable analysis can provide guidance for the researcher or manager to
find any improper resource allocation and utilisation in the DMU; hence, a
relatively inefficient DMU can help determine how to adjust inputs in order to
increase outputs. Analysis of the different variables is separated into the CCR
model and the BCC model. The CCR model (i.e. the slack variable analysis of CRS
efficiency) represents the long-term direction of the DMUs while the BCC model
(the slack variable analysis of VRS efficiency) represents the short-term
improvment direction of the DMUs (Cooper et al. 2006).
(5) Hypothesis testing

Having established the efficiency rankings of the sample airports, the hypotheses
related to the rankings obtained need to be tested. The Mann–Whitney U-test,
which tests for differences between the efficiency scores, is adopted in the current
research. Golany and Roll (1989) recommended the use of the Mann–Whitney U-
test for the non-parametric analysis of DEA results that is used in this study
because the efficiency scores do not follow a standard normal distribution. In this
research, this test can help us answer Research Question 2:
Would an airport privatisation policy (airport ownership) influence the

performance of an airport’s operation?
6.2 AIRPORT EFFICIENCY ANALYSIS: VS I

Two different variable sets are applied in the DEA analysis in this research (as
described in Chapter 4). VS I (which includes ten variables) is conducted to evaluate
airport efficiency, and the results are described in the subsections which follow.
6.2.1 RELATIONS BETWEEN VARIABLES

Before conducting efficiency analysis, a correlation coefficients analysis is applied to
determine the relations between the input and output variables. Table 6.1 presents all of
the relations between each input and output variable. The results show that all of the
variables can satisfy the isotonicity test properly, which means that an output should
not decrease with an increased input. All of the correlation coefficients are positive in
134
Ch 6
VS I. Therefore, all of the different resources and facilities are generally dimensioned
jointly to avoid conflict.
Table 6.1: Correlation coefficients among input and output variables: VS I

Input Variables
Size of
Output Number of Number of Number of Length of Operational
Terminal
Variables Employees Gates Runways Runway Expenditures
Area
Number of
0.340 0.632** 0.280 0.312 0.281 0.662**
passengers
Amount of
freight and 0.230 0.190 0.070 0.532** 0.313 0.297
mail
Aircraft
0.452* 0.584** 0.480* 0.135 0.187 0.640**
movements
Total
0.439* 0.725** 0.145 0.181 0.233 0.932**
revenue
**Correlation is significant at the 0.01 level (2-tailed)
* Correlation is significant at the 0.05 level (2-tailed)
6.2.2. RELATIVE EFFICIENCY ANALYSIS

The panel data for VS I are shown in Table 6.2, while the columns for variables in VS
II are highlighted in shadow. Table 6.3 shows the five evaluation items that were used
for airport efficiency and sample character. The first column shows the ownership
category of each airport, the second column presents the CRS technical efficiency
measure, the third and fourth columns are the VRS pure technical efficiency and the
scale efficiency, respectively. The fifth column indicates whether the DMU is operating
in an area of increasing or decreasing returns to scale. In addition, in order to discern
the influence of airport ownership on airport efficiency, the separation of the sample
airports into private operation and public operation is based on their property status.
In this research, 24 sample airports are classified into seven different categories, which
are based on their airport ownership and governance. To compare the results easily,
these sample airports are separated into two groups: publicly operated and privately
operated. There are 15 airports that are run by the public sector (most of them are in the
Asia-Pacific region), and nine airports are run by private companies (see Table 6.4).
135
Table 6.2: Panel data of sample airports
Input Variables Output Variables
Size of Length of Operational Number of Amount of Aircraft
Number of Number of Number of Total Revenues
Terminal Area Runway Expenditure Passengers Freight and Mails Movements
Employees Gates Runways ($million)
(𝑚! ) (ft) ($million) (000’s) (tons) (times)
Amsterdam (AMS) 2579 94 6 591885 3244 985.64 47430 1567712 446693 1602.42
Barcelona (BCN) 569 101 3 155200 2850 290.36 30208 104239 321491 324.03
Frankfurt (FRA) 17996 174 3 800000 4000 1851.29 53467 2111116 485783 2104.52
Istanbul (IST) 1750 32 3 318500 2767 172.04 28533 766221 254531 283.98
London (LGW) 2186 107 1 202519 3316 584.43 34214 112366 263653 854.14
London (LHR) 5516 264 2 364800 3780 1820.23 67056 1486262 478693 2891.66
Madrid (MAD) 797 76 4 300000 3863 580.71 50846 328985 469740 647.92
Munich (MUC) 7400 200 2 458000 4000 1063.23 34552 274464 432296 1368.4
Paris (CDG) 3858 124 4 542595 3454 1084.6 60875 2040000 559812 1768.92
Paris (ORY) 3304 102 3 371500 3123 497.65 26210 140000 230167 732.32
Rome (FCO) 3278 86 4 285000 3677 477.66 35227 137424 346654 948.93
Zurich (ZRH) 1254 67 3 138614 3167 401.63 22099 387671 274991 789.76
Bangkok (BKK) 3245 120 2 563000 3850 432.04 46932 1291931 311435 733.69
Beijing (PEK) 1965 120 3 1382000 3600 517.26 55938 1367710 429646 561.6
Guangzhou (CAN) 3482 74 2 320000 3700 235.65 33435 685868 280392 378.69
Hong Kong (HKG) 1131 106 2 710000 3800 425.23 47700 3400000 296000 1120.93
Incheon (ICN) 933 90 3 600000 3833 396.64 29973 2423717 211102 973.92
Kuala Lumpur (KUL) 1578 106 2 479404 4090 256.8 27529 667495 209681 292.24
Osaka (KIX) 388 52 2 330000 3750 458.45 16014 846522 133502 959.15
Tokyo (HRT) 720 87 2 789700 3250 1109.51 32654 2100448 193321 1830.69
Shanghai (PVG) 6440 98 3 824000 3733 212.15 28236 2603027 265735 482.15
Singapore (SIN) 1396 102 3 1043020 3583 402.63 22877 415726 185304 921.99
Shenzhen (SZX) 3998 55 1 152000 3400 95.52 21401 598036 187942 217.22
Sydney (SYD) 306 65 3 387487 2978 136.47 32900 470000 298964 773.69
Source: ATRS (2010).
136
Ch6
An examination of Table 6.3 reveals that 17 of 24 airports are operated relatively

efficiently when using the CCR model to calculate airport efficiency, and 19 of 24
airports are relatively efficient when employing the BCC model to calculate airport
efficiency. A number of points emerge by means of the basic DEA models:
(1) Too many airports are on the efficient frontier; hence it is difficult to help airport
authorities to improve their operations (i.e. obtaining scores on the CCR and BCC
of unity).
(2) Best practice calculations indicate that almost all European airports are operated at
a high level of relative efficiency, with the exception of Paris (ORY).
(3) All efficient airports determined by the CCR model are also efficient in the BCC
model, signifying that the dominant source of efficiency is scale; that is:
𝑆 𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 𝑇𝑒𝑐ℎ𝑛𝑖𝑐𝑎𝑙 𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦: 𝑇𝐸 =
𝑉𝑅𝑆 𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 𝑃𝑢𝑟𝑒 𝑡𝑒𝑐ℎ𝑛𝑖𝑐𝑎𝑙 𝑒𝑓𝑓𝑖𝑐𝑛𝑒𝑛𝑐𝑦: 𝑃𝑇𝐸 ×𝑆𝑐𝑎𝑙𝑒 𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 (𝑆𝐸) .
(4) The rationale for interpreting the BCC model as management skills is based on the
contrast between the CCR and BCC models. The CCR model identifies the overall
inefficiency while the BCC differentiates between technical efficiency and scale
efficiency (Golany and Roll, 1989). Based on this differentiation, the ratio between
the CCR and BCC models enables the estimation of scale efficiency, and,
assuming that efficiency is due to managerial skills and scale effects, the BCC
scores are interpreted as managerial skills. Therefore, according to the BCC scores,
five airports are found to be inefficient.
(5) According to the scale efficiency, there are only seven airports that are operated
relatively inefficiently.
(6) Three situations are listed in the last column. If output increases by the same
proportional change, then there are CRS, by less than the proportional change, then
there are DRS, and by more than that proportional change, then there are IRS
(Coelli et. al. 2005). Consequently, the returns to scale faced by an airport are
purely technologically imposed and are not influenced by economic decisions or by
market conditions. Therefore, from Table 6.3, four of the airports are found to need
to decrease their scale, and three of the airports are found to need to increase their
scale.
137
Ch6
Table 6.3: Efficiency scores obtained by basic DEA model: VS I

Ownership CCR model BCC model Scale*
No. DMU RTS
category (CRS) (VRS) efficiency
European
1 Amsterdam (AMS) B 1 1 1 -
2 Barcelona (BCN) F 1 1 1 -
3 Frankfurt (FRA) CE 0.9899 1 0.9899 DRS
4 Istanbul (IST) D 1 1 1 -
5 London (LGW) G 1 1 1 -
6 London (LHR) G 1 1 1 -
7 Madrid (MAD) F 1 1 1 -
8 Munich (MUC) E 1 1 1 -
9 Paris (CDG) B 1 1 1 -
10 Paris (ORY) B 0.6500 0.6551 0.9924 IRS
11 Rome (FCO) B 0.9711 0.9948 0.9728 DRS
12 Zurich (ZRH) B 1 1 1 -
Mean 0.9676 0.9708 0.9963
Asia-Pacific
13 Bangkok (BKK) C 1 1 1 -
14 Beijing (PEK) C 1 1 1 -
15 Guangzhou (CAN) C 0.9983 1 0.9983 DRS
16 Hong Kong (HKG) F 1 1 1 -
17 Incheon (ICN) F 0.9391 0.9418 0.9971 IRS
Kuala Lumpur
18 F 0.7890 0.8844 0.8791 IRS
(KUL)
19 Osaka (KIX) CE 1 1 1 -
20 Tokyo (HRT) F 1 1 1 -
21 Shanghai (PVG) F 1 1 1 -
22 Singapore (SIN) A 0.7878 0.8318 0.9471 DRS
23 Shenzhen (SZX) C 1 1 1 -
24 Sydney (SYD) B 1 1 1 -
Mean 0.9595 0.9715 0.9851
Mean of all samples 0.9635 0.9712 0.9907
S.D 0.0896 0.0786 0.0266
(7) Among these inefficient airports in CCR and BCC model, only Paris (ORY) is
relatively smaller than other airports and it is far away from the other results
(0.6500 in CRS efficiency and 0.6551 in VRS efficiency).
*
The scale efficiency is the quotient obtained by the division of the technical efficiency with constant returns to scale and variable
returns to scale (Cooper et al 2006). If this scale efficiency is near one, it expresses that the airport is near to the optimal scale of
operations. The area of operation has been obtained by running a DEA problem with non-increasing returns to scale.
138
Ch6
(8) From Table 6.4, the BCC model identifies five airports as being inefficient.
According to ownership category, two of them are privately operated (i.e. Paris:
ORY and Rome: FCO), and three of them are publicly operated (i.e. Singapore:
SIN, Incheon: ICN, and Kuala Lumpur: KUL).
(9) Table 6.4 also reveals that the efficiency of publicly operated airports is better than
that of those that are privately operated. However, in general, only ORY has a
relatively lower efficiency score and the efficiency scores for other private airports
were all higher.
Table 6.4: Airport efficiency with different ownership: VS I

Ownership CCR model BCC model Scale
DMU
category* (CRS) (VRS) efficiency
Amsterdam (AMS) (A) 1 1 1
Istanbul (IST) (A) 1 1 1
London (LGW) (A) 1 1 1
London (LHR) (A) 1 1 1
Paris (CDG) (A) 1 1 1
Paris (ORY) (A) 0.6500 0.6551 0.9924
Rome (FCO) (A) 0.9711 0.9983 0.9728
Zurich (ZRH) (A) 1 1 1
Sydney (SYD) (A) 1 1 1
Mean 0.9579 0.9615 0.9961
S.D 0.1159 0.1149 0.0091
Barcelona (BCN) (B) 1 1 1
Frankfurt (FRA) (B) 0.9899 1 0.9899
Madrid (MAD) (B) 1 1 1
Munich (MUC) (B) 1 1 1
Bangkok (BKK) (B) 1 1 1
Beijing (PEK) (B) 1 1 1
Guangzhou (CAN) (B) 0.9983 1 0.9983
Hong Kong (HKG) (B) 1 1 1
Incheon (ICN) (B) 0.9391 0.9418 0.9971
Kuala Lumpur (KUL) (B) 0.7890 0.8975 0.8791
Osaka (KIX) (B) 1 1 1
Tokyo (HRT) (B) 1 1 1
Shanghai (PVG) (B) 1 1 1
Singapore (SIN) (B) 0.7878 0.8318 0.9471
Shenzhen (SZX) (B) 1 1 1
Mean 0.9669 0.9780 0.9874
S.D. 0.0741 0.05 0.0329
*(A) represents privately operated airports; (B) represents publicly operated airports.
139
Ch6
6.2.3 THE CLUSTERING ANALYSIS REFERRAL

Table 6.5 shows the referral frequency for every airport in the BCC model. The results
of the analysis show that Sydney airport has the highest referral frequency (i.e. 5). The
airports that referred to Sydney are Paris (ORY), Rome, Incheon, Kuala Lumpur, and
Singapore. London (LHR) has the second highest referral frequency (i.e. 3). The
airports that referred to LHR are Paris (ORY), Rome, and Singapore. A further nine
relatively efficient airports were also referred. The result of this analysis indicates that,
among these 17 efficient airports, Sydney and London (Heathrow) are the most
relatively efficient.
Table 6.5: The referral of clustering analysis: VS I

Referral Referral Referral Referral
No. DMU No. DMU
Clustering Frequency Clustering Frequency
Amsterdam Bangkok
1 1 1 13 13 0
(AMS) (BKK)
Barcelona Beijing
2 2 1 14 14 0
(BCN) (PEK)
Frankfurt Guangzhou
3 3 0 15 15 0
(FRA) (CAN)
Istanbul Hong Kong
4 4 2 16 16 2
(IST) (HKG)
London Incheon 4, 16,
5 5 0 17 0
(LGW) (ICN) 19, 24
Kuala
London 16, 19, 23,
6 6 3 18 Lumpur 0
(LHR) 24
(KUL)
Madrid Osaka
7 7 1 19 19 2
(MAD) (KIX)
Munich Tokyo
8 8 0 20 20 0
(MUC) (HRT)
Paris Shanghai
9 9 2 21 21 0
(CDG) (PVG)
Paris 1, 2, 4, 6, 9, Singapore
10 0 22 6, 24 0
(ORY) 12, 24 (SIN)
Rome 6, 7, 9, 12, Shenzhen
11 0 23 23 1
(FCO) 24 (SZX)
Zurich Sydney
12 12 2 24 24 5
(ZRH) (SYD)
6.2.4 THE SLACK VARIABLE ANALYSIS

As mentioned in Section 6.1.2, the results of the slack analysis can only provide
guidance by which inefficient airports can improve their efficiency. Sometime this
guidance is difficult to achieve. The results of slack analysis by individual airport are
listed from Table 6.6 to Table 6.9. In the analysis of difference variables of CRS
efficiency (from Table 6.6 and 6.7), 7 of the 24 airports are shown to be inefficient.
Therefore, from the input variables, it is recommended that these seven airports over
140
Ch6
the long-term should reach an efficient output by reducing (on average) the number of
employees by 2673.58, the number of gates by 4.02, the number of runways by 0.2, the
size of terminal area by 115,397 𝑚! , the length of runways by 188.73 m, and their
operational expenditures by $82.07 million. Among the output variables, they are
recommended (on average) to raise the number of passengers by 8694.04, the amount
of freight and mail by 422680.6 tonnes, the aircraft movements by 60053.39, and total
revenue by $182.02 million. Among the output variables (on average), they are
recommended to raise the number of passengers by 8694.051, the amount of freight and
mail by 422680.6 tonnes, the aircraft movements 60053.38, and the total revenue by
$182.018 million to become efficient.
In the analysis of different variables of VRS efficiency (as shown in Tables 6.8 and 6.9),
five of the twenty-four airports are shown to be inefficient. Therefore, from the input
variables, in the short-term, these five airports are recommended (on average) to reduce
their employees by 801.22, the number of gates by 8.95, the number of runways by
0.37, the size of terminal area by 148,810 𝑚! , and their length of runways by 335.70𝑚
to become efficient. Among the output variables (on average), they are recommended
to raise the number of passengers by 8466.59, the amount of freight and mail by
424317.2 tonnes, the aircraft movements 69303.78, and the total revenue by $216.48
million to become efficient.
141
Ch6
Table 6.6: Long-term projection of individual airports (1): VS I

Operational
Number of Number of Number of Size of terminal Length of
expenditure
employees gates runways area (m2) runway
(million $)
Frankfurt (FRA) 0.9899
Actual data 17996 174 3 800000 4000 1851.29
Projection 4006.68 174 3 533283.74 3731.47 1276.81
Difference -13989.32 0 0 -266716.26 -268.53 -574.48
% -77.74% 0 0 -33.34% -6.71% -31.03%
Paris (ORY) 0.6500
Actual data 3304 102 3.00 371500 3123 497.65
Projection 1454.88 102 2.96 371500 3123 497.65
Difference -1849.12 0 -0.04 0 0 0
% -55.97% 0 -1.39% 0 0 0
Rome (FCO) 0.9711
Actual data 3278 86 4 285000 3677 477.66
Projection 1293.38 86 3.56 285000 3677 477.66
Difference -1984.62 0 -0.44 0 0 0
% -60.54% 0 -10.89% 0 0 0
Guangzhou (CAN) 0.9983
Actual data 3482 74 2 320000 3700 235.65
Projection 2967.31 74 2 320000 3700 235.65
Difference -514.69 0 0 0 0 0
% -14.78% 0 0 0 0 0
Incheon (ICN) 0.9391
Actual data 933 90 3 600000 3833 396.64
Projection 892.61 90 2.06 594904.02 3643.09 396.64
Difference -40.39 0 -0.94 -5095.98 -189.91 0
% -4.33% 0 -31.46% -0.85% -4.95% 0
Kuala Lumpur (KUL) 0.7890
Actual data 1578 106 2 479404 4090 256.80
Projection 1578 77.84 2 479404 3227.34 256.80
Difference 0 -28.16 0 0 -862.66 0
% 0 -26.57% 0 0 -21.09% 0
Singapore (SIN) 0.7878
Actual data 1396 102 3 1043020 3583 402.63
Projection 1059.07 102 3 507056.42 3583 402.63
Difference -336.93 0 0 -535963.58 0 0
% -24.14% 0 0 -51.39% 0 0
-2673.58 -4.02286 -0.20286 -115397 -188.729 -82.0686

Average
-33.93% -3.80% -6.25% -12.23% -4.68% -4.43%
142
Ch6
Table 6.7: Long-term projection of individual airports (2): VSI

Number of Amount of freight Aircraft Total revenue
passengers (000's) and mails (tonnes) movements (million $)
Actual data 53467 2111116 485783 2104.52
Projection 61778.97 2132565.9 490718.78 2125.90
Difference 8311.97 21449.90 4935.78 21.38
% 15.55% 1.02% 1.02% 1.02%
Paris (ORY) 0.6500
Actual data 26210 140000 230167 732.32
Projection 40325.93 737637.08 354128.09 1126.73
Difference 14115.93 597637.08 123961.09 394.41
% 53.86% 426.88% 53.86% 53.86%
Rome (FCO) 0.9711
Actual data 35227 137424 346654 948.93
Projection 36277.09 553791.61 369256.44 977.22
Difference 1050.09 416367.61 22602.44 28.29
% 2.98% 302.98% 6.52% 2.98%
Actual data 33435 685868 280392 378.69
Projection 33493.58 817069.73 280883.25 488.33
Difference 58.58 131201.73 491.25 109.64
% 0.18% 19.13% 0.18% 28.95%
Actual data 29973 2423717 211102 973.92
Projection 39445.47 2580934.56 260070.49 1037.095
Difference 9472.47 157217.56 48968.49 63.18
% 31.60% 6.49% 23.20% 6.49%
Actual data 27529 667495 209681 292.24
Projection 34888.84 1415738.2 265738.87 701.11
Difference 7359.84 748243.2 56057.87 408.87
% 26.73% 112.10% 26.73% 139.91%
Actual data 22877 415726 185304 921.99
Projection 43366.48 1302373.14 348660.75 1170.35
Difference 20489.48 886647.14 163356.75 248.356
% 89.56% 213.28% 88.16% 26.94%
8694.051 422680.6 60053.38 182.018

Average
31.49% 154.55% 28.52% 37.16%
143
Ch6
Table 6.8: Short-term projection of individual airports (1): VSI

Operational
Number of Number of Number of Size of terminal Length of
expenditure
employees gates runways area (m2) runway
(million $)
Paris (ORY) 0.6551
Actual data 3304 102 3.00 371500 3123 497.65
Projection 1502.33 102 2.99 371500 3123 497.65
Difference -1801.66 0 -0.01 0 0 0
% -54.53% 0 -0.44% 0 0 0
Rome (FCO) 0.9983
Actual data 3278 86 4 285000 3677 477.66
Projection 1340.50 81.98 3.20 285000 3269.50 477.66
Difference -1937.5 -4.02 -0.80 0 -407.51 0
% -59.11% -4.67% -20.03% 0 -11.08% 0
Actual data 933 90 3 600000 3833 396.64
Projection 932.50 90 2.12 597838.4 3682.98 396.64
Difference -0.50 0 -0.88 -2161.56 -150.02 0
% -0.05% 0 -29.21% -0.36% -3.91% 0
Actual data 1578 106 2 479404 4090 256.80
Projection 1578 70.80 2 396636.6 3447.26 256.80
Difference 0 -35.20 0 -82767.4 -642.74 0
% 0 -33.21% 0 -17.26% -15.71% 0
Actual data 1396 102 3 1043020 3583 402.63
Projection 1129.57 96.46 2.84 383900.8 3104.78 402.63
Difference -266.43 -5.54 -0.16 -659119 -478.22 0
% -19.09% -5.43% -5.27% -63.19% -13.35% 0
-801.218 -8.952 -0.37 -148810 -335.698 0

Average
-26.56% -8.66% -10.99% -16.16% -8.81% 0.00%
144
Ch6
Table 6.9: Short-term projection of individual airports (2): VSI

Number of Amount of freight Aircraft Total revenue
passengers (000's) and mails (tonnes) movements (million $)
Paris (ORY) 0.6551
Actual data 26210 140000 230167 732.32
Projection 40008.76 747131 351342.8 1117.86
Difference 13798.76 607131 121175.8 385.54
% 52.65% 433.66% 52.65% 52.65%
Rome (FCO) 0.9983
Actual data 35227 137424 346654 948.93
Projection 35285.8 608342.9 347232.7 950.51
Difference 58.80 470918.9 578.66 1.58
% 0.17% 342.68% 0.17% 0.17%
Actual data 29973 2423717 211102 973.92
Projection 39883.68 2573498 265049.3 1034.11
Difference 9910.68 14978.5 53947.28 60.19
% 33.07% 6.18% 25.56% 6.18%
Actual data 27529 667495 209681 292.24
Projection 30671.52 1346330 238427.6 740.83
Difference 3142.52 678835.4 28746.58 448.59
% 11.42% 101.70% 13.71% 153.50%
Actual data 22877 415726 185304 921.99
Projection 38299.2 630645.4 327374.6 1108.49
Difference 15422.2 214919.4 142070.6 186.50
% 67.41% 51.70% 76.67% 20.23%
8466.592 397356.6 69303.78 216.48

Average
32.94% 187.18% 33.75% 46.55%
145
Ch6
6.2.5 HYPOTHESIS TESTING

The separation of sample airports into private operation and public operation is based
on their property status. In this research, 24 sample airports are classified into seven
different categories that are based on airport ownership and governance. In order to
answer research questions, these airports are separated into two groups: public sector
operation and private company operation. There are 15 airports under the public sector
(most of them are in Asia-Pacific) and nine airports under private companies. This
section puts forward the hypothesis to be tested as follows:
Hypothesis: Airports under private management are more efficient than those under
public management.
Table 6.10 presents the Mann–Whitney test results. The minus sign of the Z-score
indicates that privately managed airports are found to have higher efficiency scores
than publicly managed facilities, thus validating the hypothesis that private airports are
more efficient than their public counterparts (Parker 1999). It also shows that the z-
value is −0.126, with a significance level p of p = 0.900. The probability value p is
not less than or equal to 0.05, so the result is not significant. Therefore, there is no
statistically significant difference in airport efficiency between privately operated and
publicly operated airports. However, this result is quite different from the common
opinion that private companies can be operated more efficiently than those in the public
sector. Therefore, sensitivity analysis is conducted in the following section in order to
determine if this result is reliable.
Table 6.10: Mann-Whitney test of differences in efficiency: VS I

Asymptotic
Mann-Whitney Mann-Whitney
Reference Significance
U-test Z-test
(two-tailed)
Privately managed airports
vs. 66.00 −0.126 0.900
publicly managed airports
.
6.3 AIRPORT EFFICIENCY ANALYSIS: VS II

In this section, VS II (which includes six variables) is used to conduct one kind of
sensitivity analysis (i.e. another one using another analysis tool is undertaken in the
146
Ch6
next section). The five steps that are used to analyse airport efficiency are described in
the subsections which follow.
6.3.1 RELATIONS BETWEEN VARIABLES

Before conducting efficiency analysis it is also necessary to apply a correlation
coefficients analysis to understand the relations between the input and output variables
in this variable set. Table 6.11 shows all of the correlation coefficients of all variables
in VS II to be positive. Therefore, all of the different resources and facilities are
dimensioned jointly to avoid conflict.
Table 6.11: Correlation coefficients among inputs and outputs variables: VS II

Input Variables
Output Number of Size of
Number of Employees
Variables Runways Terminal Area
Number of
0.340 0.280 0.312
passengers
Amount of freight
0.230 0.070 0.532**
and mail
Aircraft movements 0.452* 0.480* 0.135
**Correlation is significant at the 0.01 level (2-tailed)

*Correlation is significant at the 0.05 level (2-tailed)
6.3.2 RELATIVE EFFICIENCY ANALYSIS

The panel data for VS II are shown in Table 6.2. It indicates that 6 of the 24 airports are
found to be operated relatively efficiently when analysed using a CCR model to
calculate airport efficiency while 12 of the 24 airports are found to be operated
relatively efficiently when a BCC model is used to calculate airport efficiency. A
number of points emerge from Table 6.12 and Table 6.13 including:
(1) There are relatively fewer efficient airports as compared to inefficient airports
when comparison is conducted using Table 6.3.
(2) Best practice calculations indicate that European airports are operated relatively
more efficiently than Asian airports. This is shown from the efficiency average.
(3) All of the efficient CRS airports are also efficient in the VRS model, signifying
that the dominant source of efficiency is scale.
(4) According to the scale efficiency, there are seven airports that are operated
relatively efficiently.
147
Ch6
(5) Twelve of the inefficient airports need to decrease their scale, and four of them
need to increase their scale.
(6) Among the inefficient airports, Paris (ORY) (0.4305 in the CRS model; 0.4810 in
the BCC model) and Singapore (0.4423 in the CRS model; 0.4579 in the BCC
model) are relatively smaller than other airports.
(7) The results shown in Table 6.13 reveal that when using the CCR model, the
efficiency of privately operated airports is better than that of those that are publicly
operated. However, when using BCC model, public airports are shown to achieve
better efficiency, as mentioned in the previous section. The results from the BCC
model are closer to the real world than those of the CCR model. Therefore, by
means of VS II, publicly operated airports can get higher efficiency as compared to
those that are operated privately. However, in general, only ORY had a relatively
lower efficiency score, and the efficiency of other private airports all was higher.
148
Ch6
Table 6.12: Efficiency scores obtained using basic DEA model: VS II

Ownership CCR Model BCC Model Scale
No. DMU RTS
Category (CRS) (VRS) Efficiency
European
Amsterdam
1 B 0.7120 0.9010 0.7902 DRS
(AMS)
Barcelona
2 F 1 1 1 -
(BCN)
3 Frankfurt (FRA) CE 0.7384 0.9996 0.7387 DRS
4 Istanbul (IST) D 0.6930 0.7046 0.9835 DRS
5 London (LGW) G 1 1 1 -
6 London (LHR) G 1 1 1 -
7 Madrid (MAD) F 1 1 1 -
8 Munich (MUC) E 0.8263 0.9030 0.9151 DRS
9 Paris (CDG) B 0.9725 1 0.9725 DRS
10 Paris (ORY) B 0.4305 0.4810 0.8950 DRS
11 Rome (FCO) B 0.6803 0.7629 0.8917 DRS
12 Zurich (ZRH) B 1 1 1 -
Mean 0.8378 0.8960 0.9322
Asia-Pacific
13 Bangkok (BKK) C 0.7817 0.8229 0.9499 DRS
14 Beijing (PEK) C 0.9276 1 0.9276 DRS
Guangzhou
15 C 0.7036 0.7039 0.9996 -
(CAN)
Hong Kong
16 F 1 1 1 -
(HKG)
17 Incheon (ICN) F 0.8642 0.9061 0.9538 IRS
Kuala Lumpur
18 F 0.6447 0.6493 0.9929 DRS
(KUL)
19 Osaka (KIX) CE 0.8151 1 0.8151 IRS
20 Tokyo (HRT) F 0.9956 1 0.9956 IRS
21 Shanghai (PVG) F 0.6671 0.8065 0.8272 DRS
22 Singapore (SIN) A 0.4423 0.4579 0.9659 DRS
23 Shenzhen (SZX) C 0.9609 1 0.9609 IRS
24 Sydney (SYD) B 1 1 1 -
Mean 0.8169 0.8622 0.9490
Mean of all
0.8273 0.8791 0.9406
samples
149
Ch6
Table 6.13: Airport efficiency in different ownership: VS II

Ownership CCR model BCC model Scale
DMU
category (CRS) (VRS) efficiency
Amsterdam (AMS) (A) 0.7120 0.9010 0.7902
Istanbul (IST) (A) 0.6930 0.7046 0.9835
London (LGW) (A) 1 1 1
London (LHR) (A) 1 1 1
Paris (CDG) (A) 0.9725 1 0.9725
Paris (ORY) (A) 0.4305 0.4810 0.8950
Rome (FCO) (A) 0.6803 0.7629 0.8917
Zurich (ZRH) (A) 1 1 1
Sydney (SYD) (A) 1 1 1
Mean 0.8320 0.8722 0.9481
S.D 0.2093 0.1856 0.0738
Barcelona (BCN) (B) 1 1 1
Frankfurt (FRA) (B) 0.7384 0.9996 0.7387
Madrid (MAD) (B) 1 1 1
Munich (MUC) (B) 0.8263 0.9030 0.9151
Bangkok (BKK) (B) 0.7817 0.8229 0.9499
Beijing (PEK) (B) 0.9276 1 0.9276
Guangzhou (CAN) (B) 0.7036 0.7039 0.9996
Hong Kong (HKG) (B) 1 1 1
Incheon (ICN) (B) 0.8642 0.9061 0.9538
Kuala Lumpur (KUL) (B) 0.6447 0.6493 0.9929
Osaka (KIX) (B) 0.8151 1 0.8151
Tokyo (HRT) (B) 0.9956 1 0.9956
Shanghai (PVG) (B) 0.6671 0.8065 0.8272
Singapore (SIN) (B) 0.4423 0.4579 0.9659
Shenzhen (SZX) (B) 0.9609 1 0.9609
Mean 0.8245 0.8833 0.9362
S.D. 0.1647 0.1663 0.0806
(A) represents privately operated airports; (B) represents publicly operated airports.
6.3.3 THE REFERRAL OF CLUSTERING ANALYSIS

Table 6.14 shows the referral frequency for the airports in the BCC model. The results
show that Hong Kong Airport has the highest referral frequency (i.e. 9); it can be seen
that the airports that referred to Hong Kong are Amsterdam, Frankfurt, Istanbul,
Bangkok, Guangzhou, Incheon, Kuala Lumpur, Shanghai, and Singapore. Madrid has
the second highest referral frequency (i.e. 7). This result indicates that Hong Kong and
Madrid are relatively efficient among these airports that have an efficiency score of 1.
150
Ch6
Table 6.14: The referrals in clustering analysis: VS II

Referral Referral Referral Referral
No. DMU No. DMU
Clustering Frequency Clustering Frequency
Amsterdam Bangkok
1 16, 7, 9 0 13 16, 6 0
(AMS) (BKK)
Barcelona
2 2 3 14 Beijing (PEK) 14 2
(BCN)
Frankfurt Guangzhou 6, 16, 5, 7,
3 16, 9, 6 0 15 0
(FRA) (CAN) 2
Istanbul 6, 16, 12, 9, Hong Kong
4 0 16 16 9
(IST) 7 (HKG)
London Incheon 16, 2
5 5 3 17 0
(LGW) (ICN) 19
Kuala
London 16, 7,
6 6 6 18 Lumpur 0
(LHR) 14, 5
(KUL)
Madrid
7 7 7 19 Osaka (KIX) 19 1
(MAD)
Munich
8 6 0 20 Tokyo (HRT) 20 0
(MUC)
Shanghai
9 Paris (CDG) 9 5 21 16, 9 0
(PVG)
Singapore 14, 16,
10 Paris (ORY) 6, 7, 9 0 22 0
(SIN) 7, 5
Shenzhen
11 Rome (FCO) 7, 2 0 23 23 0
(SZX)
Sydney
12 Zurich (ZRH) 12 1 24 24 0
(SYD)
6.3.4 THE SLACK VARIABLE ANALYSIS

As mentioned in Section 6.1.2, the results from slack analysis can only provide
guidance by which inefficient airports can improve their efficiency. However, some of
these suggestions are quite difficult to achieve in the case of airports. The results of
slack analysis by individual airports are listed in Tables 6.15 to 6.18. In the analysis of
difference variables of CRS efficiency (as shown in Table 6.15 and Table 6.16) 17 of
the 24 airports were found to be inefficient. Therefore, for these 17 airports to reach an
efficient output in the long-term, on average, they have to reduce their employees by
1271.09, the number of runways by 0.2, and the size of their terminal areas by
67,493.5𝑚! (from the output variables). Among the output variables, on average, these
airports are recommended to raise the number of passengers by 19,717.59, the amount
of freight and mail by 739,865.2 tonnes, and aircraft movements by 110,799.8. The
analysis of the different variables of VRS efficiency (as shown in Table 6.12) shows 12
of 24 airports to be inefficient. Therefore, in the short-term, these airports are
recommended, on average, to reduce their employees by 1039.06, the number of
runways by 0.20, and the size of their terminal areas by 76442.4 𝑚! in order to reach an
efficient output. Among the output variables, on average, they are recommended to
151
Ch6
raise their number of passengers by 20,145.81, their amount of freight and mail by
638,989.5 tonnes and their aircraft movements by 115,100.9 in order to obtain
efficiency.
Table 6.15: Long-term projections for individual airports (1): VS II

Input Output
Size of Number of Amount of
Number of Number of Aircraft
Terminal Area Passengers Freight and
Employees Runways Movements
(m2) (000's) Mails (tonnes)
Amsterdam (AMS) 0.7120
Actual data 2579 6 591885 47430 1567712 446693
Projection 2579 5.71 591885 66610.85 2201700 627337.1
Difference 0 -0.29 0 19180.85 633988 180644.1
% 0.00% -4.79% 0.00% 40.44% 40.44% 40.44%
Actual data 17996 3 800000 53467 2111116 485783
Projection 6830.96 3 660800.8 94214.2 2859179 657917.7
Difference -11165 0 -139199 40747.2 748062.9 172134.7
% -62.04% 0.00% -17.40% 76.21% 35.43% 35.43%
Istanbul (IST) 0.6930
Actual data 1750 3 318500 28533 766221 254531
Projection 1750 2.99 318500 41174.36 1105690 367299.3
Difference 0 0.01 0 12641.36 339469.3 112768.3
% 0.00% -0.30% 0.00% 44.30% 44.30% 44.30%
Munich (MUC) 0.8263
Actual data 7400 2 458000 34552 274464 432296
Projection 4469.42 2 401611.5 68311.17 332157.4 523166.4
Difference -2930.58 0 -56388.5 33759.17 57693.44 90870.37
% -39.60% 0.00% -12.31% 97.71% 21.02% 21.02%
Paris (CDG) 0.9725
Actual data 3858 4 542595 60875 2040000 559812
Projection 3858 4 542595 70901.24 2097753 575660.5
Difference 0 0 0 10026.24 57753.37 15848.54
% 0.00% 0.00% 0.00% 16.47% 2.83% 2.83%
Paris (ORY) 0.4305
Actual data 3304 3 371500 26210 140000 230167
Projection 3304 3 371500 64311.35 325168.9 534594
Difference 0 0 0 38101.35 185168.9 304427
% 0.00% 0.00% 0.00% 145.37% 132.26% 132.26%
Rome (FCO) 0.6803
Actual data 3278 4 285000 35227 137424 346654
Projection 1822.09 4 285000 52868.11 202001.7 509552.3
Difference -1455.91 0 0 17641.11 64577.75 162898.3
% -44.41% 0.00% 0.00% 50.08% 46.99% 46.99%
Bangkok (BKK) 0.7817
Actual data 3245 2 563000 46932 1291931 311435
Projection 3245 2 520246.6 60039.05 1652738 429165.5
Difference 0 0 -42753.4 13107.05 360807.3 117730.5
% 0.00% 0.00% -7.59% 27.93% 27.93% 37.80%
152
Ch6
Table 6.16: Long-term projections for individual airports (2): VS II

Input Output
Beijing (PEK) 0.9276
Actual data 1965 3 1382000 55938 1367710 429646
Projection 1965 3 1039735 73267.21 4836946 463162.5
Difference 0 0 -342265 17329.21 3469236 33516.5
% 0.00% 0.00% -24.77% 30.98% 253.65% 7.80%
Actual data 3482 2 320000 33435 685868 280392
Projection 3482 2 320000 52945.4 974794 398508.8
Difference 0 0 0 19510.4 288926 118116.8
% 0.00% 0.00% 0.00% 58.35% 42.13% 42.13%
Actual data 933 3 600000 29973 2423717 211102
Projection 933 1.650926 585787.1 39357.95 2804581 244274.6
Difference 0 -1.35 -14212.9 9384.95 380863.8 33172.64
% 0.00% -44.97% -2.37% 31.31% 15.71% 15.71%
Actual data 1578 2 479404 27529 667495 209681
Projection 1578 2 479404 45285.11 1777012 325248.1
Difference 0 0 0 17756.11 1109517 115567.1
% 0.00% 0.00% 0.00% 64.50% 166.22% 55.12%
Osaka (KIX) 0.8151
Actual data 388 2 330000 16014 846522 133502
Projection 388 1.38522 299125.1 22048.67 1038489 163776.5
Difference 0 -0.61 -30874.9 6034.67 191967.2 30274.47
% 0.00% -30.74% -9.36% 37.68% 22.68% 22.68%
Tokyo (HRT) 0.9956
Actual data 720 2 789700 32654 2100448 193321
Projection 720 1.57 475757.9 32798.28 2109729 215060.7
Difference 0 -0.43 -313942 144.28 9280.68 21739.67
% 0.00% -21.38% -39.75% 0.44% 0.44% 11.25%
Shanghai (PVG) 0.6671
Actual data 6440 3 824000 28236 2603027 265735
Projection 2141.09 2.48 824000 62501.93 3902154 398358.8
Difference -4298.92 -0.56 0 34265.93 1299127 132623.8
% -66.75% -17.19% 0.00% 121.36% 49.91% 49.91%
Actual data 1396 3 1043020 22877 415726 185304
Projection 1396 3 835266.2 62411.11 3772669 418922.1
Difference 0 0 -207754 39534.11 3356943 233618.1
% 0.00% 0.00% -19.92% 172.81% 807.49% 126.07%
Shenzhen (SZX) 0.9609
Actual data 3998 1 152000 21401 598036 187942
Projection 2239.89 0.82 152000 27436.08 622363.1 195587.2
Difference -1758.11 -0.18 0 6035.08 24327.11 7645.17
% -43.97% -17.82% 0.00% 28.20% 4.07% 4.07%
-1271.09 -0.20059 -67493.5 +19717.59 +739865.2 +110799.8
Average
-15.10% -8.07% -7.85% +61.42% +100.79% +40.93%
153
Ch6
Table 6.17: Short-term projection of individual airports (1): VS II

Input Output
Amsterdam (AMS) 0.9010
Actual data 2579 6 591885 47430 1567712 446693
Projection 2579 3.712317 496397.2 56074.58 1739959 495771.8
Difference 0 -2.28768 -95487.8 8644.58 172246.9 49078.83
% 0.00% -38.13% -16.13% 18.23% 10.99% 10.99%
Actual data 17996 3 800000 53467 2111116 485783
Projection 3887.90 3 516605.2 60438.15 2111882 485959.3
Difference -14108.1 0 -283395 6971.15 766.27 176.32
% -78.40% 0.00% -35.42% 13.04% 0.04% 0.04%
Istanbul (IST) 0.7046
Actual data 1750 3 318500 28533 766221 254531
Projection 1750 2.98 318500 40494.95 1087445 361238.6
Difference 0 -0.02 0 11961.95 321224.5 106707.6
% 0.00% -0.79% 0.00% 41.92% 41.92% 41.92%
Munich (MUC) 0.9030
Actual data 7400 2 458000 34552 274464 432296
Projection 5516 2 364800 67056 1486262 478693
Difference -1884 0 -93200 32504 1211798 46397
% -25.46% 0.00% -20.35% 94.07% 441.51% 10.73%
Paris (ORY) 0.4810
Actual data 3304 3 371500 26210 140000 230167
Projection 3304 3 344089.9 59434.27 990072 478556.8
Difference 0 0 -27410.1 33224.27 850072 248389.8
% 0.00% 0.00% -7.38% 126.76% 607.19% 107.92%
Rome (FCO) 0.7629
Actual data 3278 4 285000 35227 137424 346654
Projection 773.38 3.90 285000 48708.09 305703.3 454382.7
Difference -2504.62 -0.10 0 13481.09 168279.3 107728.7
% -76.41% -2.59% 0.00% 38.27% 122.45% 31.08%
Bangkok (BKK) 0.8229
Actual data 3245 2 563000 46932 1291931 311435
Projection 3245 2 543579.7 57031.49 2477391 384075.9
Difference 0 0 -19420.3 10099.49 1185460 72640.94
% 0.00% 0.00% -3.45% 21.52% 91.76% 23.32%
Actual data 3482 2 320000 33435 685868 280392
Projection 3482 2 320000 52857.72 974403.3 398349.1
Difference 0 0 0 19422.72 288535.3 117957.1
% 0.00% 0.00% 0.00% 58.09% 42.07% 42.07%
Actual data 933 3 600000 29973 2423717 211102
Projection 933 2.03 600000 39467.6 2674970 257543.4
Difference 0 -0.97 0 9494.60 251253.5 46441.4
% 0.00% -32.25% 0.00% 31.68% 10.37% 22.00%
154
Ch6
Table 6.18: Short-term projection of individual airports (2): VS II

Input Output
Actual data 1578 2 479404 27529 667495 209681
Projection 1578 2 409481.2 42398.45 1143050 322937.6
Difference 0 0 -69922.8 14869.45 475555.2 113256.6
% 0.00% 0.00% -14.59% 54.01% 71.24% 54.01%
Shanghai (PVG) 0.8065
Actual data 6440 3 824000 28236 2603027 265735
Projection 1477.05 2.25 688756.6 49371.88 3227419 329477.2
Difference -4962.95 -0.75 -135243 21135.88 624391.7 63742.22
% -77.06% -24.87% -16.41% 74.85% 23.99% 23.99%
Actual data 1396 3 1043020 22877 415726 185304
Projection 1396 3 700234.7 49957.79 1111286 404658.7
Difference 0 0 -342785 27080.79 695560.3 219354.7
% 0.00% 0.00% -32.86% 118.38% 167.31% 118.38%
-1039.06 -0.20 -76442.4 20145.81 638989.5 115100.9

Average
-19.88% -6.63% -10.56% 68.63% 175.32% 48.17%
6.3.5 Hypothesis testing

Having established the efficiency rankings of the sample airports, there is the need to
test again the following hypothesis:
public management.
To compare these results easily, the 24 sample airports are classified into two groups:
those operated by the public sector and those operated by private companies. There are
15 airports in this study which were run by the public sector at the time of the study
(most of them in the Asia-Pacific region) and nine airports that were run by private
companies. From Table 6.19 it can be seen that the z-value is −0.127, with a
significance level p of p = 0.899. The probability value p is not less than or equal
to 0.05, so the result is not significant. Therefore, there is no statistically significant
difference found in airport efficiency between privately operated and publicly operated
airports.
155
Ch6
Table 6.19: Mann-Whitney test of differences in efficiency: VS II

Mann-Whitney Mann-Whitney Asymptotic Significance
Reference
U-test Z-test (two-tailed)
vs. 65.5 −0.127 0.899
6.4 AIRPORT EFFICIENCY ANALYSIS: INTEGRATED AHP/DEA MODEL

Another sensitivity analysis (applying a different analysis model), is conducted in this
section. This section also aims to answer Research Question 1 and to provide further
information about Research Question 3, which are:
Research Question 1: Does the result of airport efficiency vary by conducting
different evaluation methods?
Research Question 3: Does the influence of alternative scales on the results of the
AHP analysis cause a different weight for each variable?
An integrated AHP/DEA model (and alternative scales used when calculating weight of
each variable) is used to assess airport efficiency as sensitivity analysis. Several
AHP/DEA models have been used to assess efficiency in several different fields (see
Section 4.6). In this research, the method that was proposed by Jyoti et al. (2008) was
used in which they used an AHP survey to acquire the weight of each output variable
by evaluating the judgements of 20 senior scientists and research and development
managers. After calculations, the data were transformed into dimensionless values by
computing the respective relative scores of each output variable. The relative weighted
scores were computed by multiplying the relative measure. This research follows this
process to assess airport efficiency by means of an integrated AHP/DEA model. In
addition, it expands the integrated AHP/DEA model to include input variables, which is
described in the following section.
6.4.1 THE PROCESS DESCRIPTION OF AIRPORT EFFICIENCY EVALUATION

To demonstrate the developed model, an illustration is taken to evaluate the efficiency
of the 24 sample airports. The weights of each variable (which were determined as
described in Chapter 5) are listed in Table 6.20 while the panel data for the sample
airports is shown in Table 6.2. The panel data in Table 6.2 are transformed into
156
Ch6
dimensionless values by computing the respective relative scores of each variable for
all airports, as shown in Table 6.21 (standardisation of the panel data). In Tables 6.22
the relative weighted scores are computed by multiplying the relative scores with the
respective weight of the respective variables by 1-9 scale. Take Amsterdam airport as
an example. The number of employees in this column is 0.3267, which is calculated as
(14.33 × 0.0228). (Other results which are calculated with alternative scale are attached
in Appendix III).
In addition, Coelli et al. (2005) revealed the concept that an assumption in a DEA-CCR
model is that all DMUs are operating at an optimal scale. However, imperfect
competition, government regulations, and constraints on finance may all cause a DMU
to not operate at an optimal scale. When not all DMUs are operating at the optimal
scale, the use of the DEA-BCC model specification is more suitable to assess airport
efficiency. Therefore, in the following section, a DEA-BCC (i.e. output-oriented)
model is adopted to compute relative efficiency scores. The results from the AHP/DEA
model are shown in the following sections.
157
Ch6
Table 6.20: Weights of variables in alternative scales

Input Variables
(1) (2) (3) (4) (5) (6)
10/10- ∅
Sub-Criteria 1-9 𝑒 !!! 2!!! 9/9-9/1 Mean S.D.
18/2 mapping
(A1)
Number of 0.0228 0.0124 0.0051 0.0522 0.0417 0.0613 0.0326 0.0226
employees
(A2)
Number of 0.1173 0.1153 0.1108 0.1103 0.1128 0.1087 0.1125 0.0033
gates
(A3)
Number of 0.0864 0.0825 0.0695 0.1013 0.0986 0.1009 0.0899 0.0127
runways
(A4)
Size of 0.0946 0.0971 0.0869 0.1066 0.1058 0.1050 0.0993 0.0079
terminal area
(A5)
Length of 0.0547 0.0490 0.0336 0.0860 0.0786 0.0886 0.0651 0.0225
runway
(B1)
Operational 0.6241 0.6436 0.6941 0.5436 0.5626 0.5355 0.6006 0.0634
cost
Output Variables
(5) (6)
(1) (2) (3) (4)
Sub-Criteria 10/10- ∅ Mean S.D.
1-9 𝑒 !!! 2!!! 9/9-9/1
18/2 mapping
(C1)
Number of 0.2605 0.2670 0.2991 0.2213 0.2111 0.1919 0.2418 0.0403
passengers
(C2)
Amount of
0.1649 0.1644 0.1527 0.1803 0.1722 0.1711 0.1676 0.0093
freight and
mail
(C3)
Aircraft 0.0750 0.0628 0.0402 0.1171 0.1140 0.1354 0.0908 0.0369
movement
(D1)
Total 0.4995 0.5057 0.5079 0.5024 0.5027 0.5016 0.5033 0.0030
revenue
158
Table 6.21: Relative input and output scores
Relative Input Score of DMUs Relative Output Score of DMUs
Number of Number of Number of Size of Length of Operational Number of Amount of Aircraft Total
Employees Gates Runways Terminal Area Runway Expenditure Passengers Freight and Mails Movements Revenues
Amsterdam (AMS) 14.33 35.61 100 42.83 79.32 53.24 70.73 46.11 79.79 55.42
Barcelona (BCN) 3.16 38.26 50 11.23 69.68 15.68 45.05 3.07 57.43 11.21
Frankfurt (FRA) 100 65.91 50 57.89 97.80 100 79.73 62.09 86.78 72.78
Istanbul (IST) 9.72 12.12 50 23.05 67.65 09.29 42.55 22.54 45.47 9.82
London (LGW) 12.15 40.53 16.67 14.65 81.08 31.57 51.02 3.30 47.10 29.54
London (LHR) 30.65 100 33.33 26.40 92.42 98.32 100 43.71 85.51 100
Madrid (MAD) 4.43 28.79 66.67 21.71 94.45 31.37 75.83 9.68 83.91 22.41
Munich (MUC) 41.12 75.76 33.33 33.14 97.80 57.43 51.53 8.07 77.22 47.32
Paris (CDG) 21.44 46.97 66.67 39.26 84.45 58.59 90.78 60.00 100 61.17
Paris (ORY) 18.36 38.64 50 26.88 76.36 26.88 39.09 4.12 41.12 25.33
Rome (FCO) 18.22 32.58 66.67 20.62 89.90 25.80 52.53 4.04 61.92 32.82
Zurich (ZRH) 6.97 25.38 50 10.03 77.43 21.69 41.01 11.40 49.12 27.31
Bangkok (BKK) 18.03 45.46 33.33 40.74 94.13 23.34 69.99 38.00 55.63 25.37
Beijing (PEK) 10.92 45.46 33.33 100 88.02 27.94 83.42 40.23 76.75 19.42
Guangzhou (CAN) 19.35 28.03 33.33 23.15 90.46 12.73 49.86 28.30 37.71 13.10
Hong Kong (HKG) 6.28 40.15 33.33 51.37 92.91 22.97 71.13 100 52.87 38.76
Incheon (ICN) 5.18 34.09 50 43.42 93.72 21.43 44.70 71.29 37.71 33.68
Kuala Lumpur (KUL) 8.77 40.15 33.33 34.69 100 13.87 41.05 19.63 37.46 10.11
Osaka (KIX) 2.16 19.70 33.33 23.88 91.69 24.76 23.88 24.90 23.85 33.17
Tokyo (HRT) 4 32.96 33.33 57.14 79.46 59.93 48.70 61.78 34.53 63.31
Shanghai (PVG) 35.79 37.12 50 59.62 91.27 11.46 42.11 76.56 47.47 16.67
Singapore (SIN) 7.76 38.64 50 75.47 87.60 21.75 34.12 12.23 33.10 31.88
Shenzhen (SZX) 22.22 20.83 16.67 11 83.13 5.16 31.92 17.59 33.57 07.51
Sydney (SYD) 1.70 24.62 50 28.04 72.81 73.7 49.06 13.82 53.40 26.76
159
Table 6.22: Relative weighted input and output scores (scale 1-9)
Relative Weighted Input Score Obtained by the DMUs Relative Weighted Output Score Obtained by the DMUs
1-9 weights 0.0228 0.1173 0.0864 0.0946 0.0547 0.6241 0.2227 0.1335 0.0570 0.5867
Employees Gates Runways Terminal Area Runway Expenditure Passengers Freight and Mails Movements Revenues
Amsterdam (AMS) 0.3267 4.1766 8.64 4.0515 4.3385 33.2275 15.7516 6.1557 4.5480 32.5149
Barcelona (BCN) 0.0721 4.4877 4.32 1.0624 3.8116 9.7885 10.0326 0.4098 3.2735 6.5769
Frankfurt (FRA) 2.2800 7.7311 4.32 5.4761 5.3497 62.4100 17.7559 8.2890 4.9465 42.7000
Istanbul (IST) 0.2217 1.4218 4.32 2.1802 3.7006 5.7997 9.4759 3.0091 2.5918 5.7614
London (LGW) 0.2770 4.7542 1.44 1.3863 4.4349 19.7021 11.3622 0.4406 2.6847 17.3311
London (LHR) 0.6988 11.7300 2.88 2.4971 5.0554 61.3629 22.2700 5.8353 4.8741 58.6700
Madrid (MAD) 0.1010 3.3768 5.76 2.0535 5.1664 19.5767 16.8873 1.2923 4.7829 13.1479
Munich (MUC) 0.9375 8.8864 2.88 3.1351 5.3497 35.8432 11.4757 1.0773 4.4015 27.7626
Paris (CDG) 0.4888 5.5096 5.76 3.7141 4.6194 36.5636 20.2167 8.0100 5.7000 35.8884
Paris (ORY) 0.4186 4.5320 4.32 2.5430 4.1767 16.7766 8.7053 0.5500 2.3438 14.8611
Rome (FCO) 0.4153 3.8212 5.76 1.9509 4.9176 16.1027 11.6984 0.5393 3.5294 19.2555
Zurich (ZRH) 0.1589 2.9770 4.32 0.9488 4.2356 13.5396 9.1329 1.5219 2.7998 16.0228
Bangkok (BKK) 0.4111 5.3319 2.88 3.8538 5.1490 14.5648 15.5868 5.0730 3.1709 14.8846
Beijing (PEK) 0.2490 5.3319 2.88 9.4600 4.8147 17.4377 18.5776 5.3707 4.3748 11.3937
Guangzhou (CAN) 0.4412 3.2880 2.88 2.1904 4.9484 7.9441 11.1038 3.7781 2.1495 7.6858
Hong Kong (HKG) 0.1433 4.7098 2.88 4.8601 5.0822 14.3352 15.8407 13.3500 3.0136 22.7405
Incheon (ICN) 0.1182 3.9989 4.32 4.1071 5.1263 13.3714 9.9547 9.5172 2.1495 19.7601
Kuala Lumpur (KUL) 0.1999 4.7098 2.88 3.2816 5.4700 8.6571 9.1418 2.6206 2.1352 5.9315
Osaka (KIX) 0.0492 2.3105 2.88 2.2589 5.0153 15.4551 5.3181 3.3242 1.3595 19.4608
Tokyo (HRT) 0.0912 3.8656 2.88 5.4056 4.3466 37.4034 10.8455 8.2476 1.9682 37.1440
Shanghai (PVG) 0.8159 4.3543 4.32 5.6404 4.9925 7.1519 9.3779 10.2208 2.7058 9.7803
Singapore (SIN) 0.1769 4.5320 4.32 7.1396 4.7919 13.5733 7.5985 1.6327 1.8867 18.7040
Shenzhen (SZX) 0.5065 2.4437 1.44 1.0405 4.5472 3.2201 7.1086 2.3483 1.9135 4.4061
Sydney (SYD) 0.0388 2.8880 4.32 2.6524 3.9828 4.6006 10.9257 1.8450 3.0438 15.7001
160
Ch6
6.4.2 RELATIVE EFFICIENCY ANALYSIS: VS I

Table 6.23 shows that six evaluation items used to determine airport efficiency. The
first column shows relative efficiency scores that are calculated using the original
DEA-BCC model; the second shows the relative efficiency scores that are calculated
using the integrated AHP/DEA model (the weights of each variable are considered
using a 1-9 scale). Columns third to seven reveal the relative efficiency scores that are
computed using the integrated AHP/DEA model (the weights of each variable are
considered in alternative scales).
An examination of VS I reveals that 5 of the 24 airports are operated relatively

inefficiently when using the BCC model and AHP/DEA model to calculate airport
efficiency. A number of points emerge from Table 6.23 and Table 6.24, including:
(1) The number of relatively efficient airports is greater than the number of inefficient
airports (by means of different scale). This is similar to the results which were
calculated using the basic DEA model.
(2) Best practice calculations indicate that European airports are operated relatively
more efficiently than Asian airports, which happens in all scales.
(3) The situation occurring in the AHP analysis (Chapter 6) is not discovered in Table
6.23, which is that when conducting alternative scales, the weights in scales 1 to 3
are close, and the weights in scales 4-6 are also close.
(4) Among these inefficient airports, Paris (ORY) has the lowest relative efficiency
(0.6551) in the BCC model, but Singapore has the lowest relative efficiency
(0.6672) in all AHP/DEA models, and both of them are far away from the others
under consideration in this study.
(5) The discriminatory power of the results does not improve significantly when
compared with the relative efficient scores between the BCC model and the
AHP/DEA model (1-9 scale). In addition, the scores that are calculated by
alternative scales do not noticeably improve either.
(6) Among the efficiency scores, the integrated AHP/DEA model provides most of the
airports relative efficiency scores, with the exception of five airports (i.e. FRA,
ORY, ICN, KUL, and SIN). Meanwhile, in the BCC model, FRA is found to
operate efficiently, but FCO is not. However, the efficiency score reveals that there
are only slight differences when conducting alternative judgement scales in the
AHP/DEA model.
161
Ch6
(7) The results shown in Table 6.24 indicate that when using the BCC model, public
airports can achieve better efficiency. However, when conducting the integrated
AHP/DEA model the results show that private operated airports can get higher
efficiency than public ones by conducting six different scales. However, in general,
only Singapore gets relative lower efficiency scores, and the efficiency of other
private airports is all higher.
Table: 6.23: Efficiency scores obtained by AHP/DEA model: VS I

DMU BCC 1-9 𝑒 !!! 2!!! 9/9-9/1 10/10-18/2
∅
mapping
Europe
Amsterdam (AMS) 1 1 1 1 1 1 1
Barcelona (BCN) 1 1 1 1 1 1 1
Frankfurt (FRA) 1 0.9996 0.9996 0.9996 0.9996 0.9996 0.9996
Istanbul (IST) 1 1 1 1 1 1 1
London (LGW) 1 1 1 1 1 1 1
London (LHR) 1 1 1 1 1 1 1
Madrid (MAD) 1 1 1 1 1 1 1
Munich (MUC) 1 1 1 1 1 1 1
Paris (CDG) 1 1 1 1 1 1 1
Paris (ORY) 0.6551 0.8205 0.8205 0.8205 0.8205 0.8205 0.8205
Rome (FCO) 0.9948 1 1 1 1 1 1
Zurich (ZRH) 1 1 1 1 1 1 1
Mean 0.9708 0.9850 0.9850 0.9850 0.9850 0.9850 0.9850
Asia-Pacific
Bangkok (BKK) 1 1 1 1 1 1 1
Beijing (PEK) 1 1 1 1 1 1 1
Guangzhou (CAN) 1 1 1 1 1 1 1
Hong Kong (HKG) 1 1 1 1 1 1 1
Incheon (ICN) 0.9418 0.9944 0.9944 0.9944 0.9944 0.9944 0.9944
Kuala Lumpur
0.8845 0.8975 0.8975 0.8975 0.8975 0.8975 0.8975
(KUL)
Osaka (KIX) 1 1 1 1 1 1 1
Tokyo (HRT) 1 1 1 1 1 1 1
Shanghai (PVG) 1 1 1 1 1 1 1
Singapore (SIN) 0.8318 0.6672 0.6672 0.6672 0.6672 0.6672 0.6672
Shenzhen (SZX) 1 1 1 1 1 1 1
Sydney (SYD) 1 1 1 1 1 1 1
Mean 0.9715 0.9633 0.9633 0.9633 0.9633 0.9633 0.9633
Mean of all samples 0.9712 0.9737 0.9737 0.9737 0.9737 0.9737 0.9737
162
Ch6
(8) The results for referral frequency for the airports in the AHP/DEA model shows
Hong Kong and Sydney to have the highest referral frequency (i.e. 4). The results
indicate that Hong Kong and Sydney are relatively efficient among these airports
that have an efficiency score of 1.
Table 6.24: Airport efficiency under different ownership: VS I

Ownership 10/10- ∅
DMU BCC 1-9 𝑒 !!! 2!!! 9/9-9/1
category* 18/2 mapping
Amsterdam
(A) 1 1 1 1 1 1 1
(AMS)
Istanbul (IST) (A) 1 1 1 1 1 1 1
London (LGW) (A) 1 1 1 1 1 1 1
London (LHR) (A) 1 1 1 1 1 1 1
Paris (CDG) (A) 1 1 1 1 1 1 1
Paris (ORY) (A) 0.6551 0.8205 0.8205 0.8205 0.8205 0.8205 0.8205
Rome (FCO) (A) 0.9983 1 1 1 1 1 1
Zurich (ZRH) (A) 1 1 1 1 1 1 1
Sydney (SYD) (A) 1 1 1 1 1 1 1
Mean 0.9615 0.9801 0.9801 0.9801 0.9801 0.9801 0.9801
S.D 0.1149 0.0598 0.0598 0.0598 0.0598 0.0598 0.0598
Barcelona (BCN) (B) 1 1 1 1 1 1 1
Frankfurt (FRA) (B) 1 0.9996 0.9996 0.9996 0.9996 0.9996 0.9996
Madrid (MAD) (B) 1 1 1 1 1 1 1
Munich (MUC) (B) 1 1 1 1 1 1 1
Bangkok (BKK) (B) 1 1 1 1 1 1 1
Beijing (PEK) (B) 1 1 1 1 1 1 1
Guangzhou
(B) 1 1 1 1 1 1 1
(CAN)
Hong Kong
(B) 1 1 1 1 1 1 1
(HKG)
Incheon (ICN) (B) 0.9418 0.9944 0.9944 0.9944 0.9944 0.9944 0.9944
Kuala Lumpur
(B) 0.8975 0.8975 0.8975 0.8975 0.8975 0.8975 0.8975
(KUL)
Osaka (KIX) (B) 1 1 1 1 1 1 1
Tokyo (HRT) (B) 1 1 1 1 1 1 1
Shanghai (PVG) (B) 1 1 1 1 1 1 1
Singapore (SIN) (B) 0.8318 0.6672 0.6672 0.6672 0.6672 0.6672 0.6672
Shenzhen (SZX) (B) 1 1 1 1 1 1 1
Mean 0.9781 0.9706 0.9706 0.9706 0.9706 0.9706 0.9706
S.D. 0.0500 0.0880 0.0880 0.0880 0.0880 0.0880 0.0880
163
Ch 6
6.4.3 RELATIVE EFFICIENCY ANALYSIS: VS II

VS II is conducted and discussed in this section in order to determine the sensitivity
analysis. An examination of VS II (as shown in Table 6.25) reveals that 12 of the 24
airports are operated relatively inefficiently when using the BCC model and the
AHP/DEA models to calculate airport efficiency. A number of points emerge from
Table 6.25 and Table 6.26, including:
(1) The number of efficient airports is the same as that of inefficient airports (not only
in the BCC model but also in all AHP/DEA models).
(2) The efficiency scores between European airports and Asian airports are very close.
The average efficiency of Asian airports in the BCC model is 0.8791, and the
average efficiency of European airports is 0.8960. In the AHP/DEA model, the
average efficiency of Asian airports is 0.8560, and the average efficiency of
European airports is 0.8385.
(3) The situation occurring in the AHP analysis is not discovered in Table 6.25, which
is that when conducting alternative scales, the weights in scales 1 to 3 are close,
and weights in scales 4-6 are also close.
(4) Among these inefficient airports, the efficiency scores of Paris (ORY) (0.4810 in
the BCC model, 0.4405 in the AHP/DEA model) and Singapore (0.4580 in the
BCC model, 0.4294 in the AHP/DEA model) are relatively smaller than those of
other airports.
(5) The efficiency scores between the BCC model and the AHP/DEA model (1-9 scale)
are significantly different; for example, the MUC (from 0.9031 to 0.5153) and
FCO (from 0.7629 to 0.6711). However, the scores that are calculated using
different judgement scales do not obviously change.
(6) The results in VS II can obviously provide a better discriminatory power than VS I.
(7) The results shown in Table 6.26 indicate that when using the BCC model, public
airports can achieve better efficiency. However, when conducting the integrated
AHP/DEA model, the results show that privately operated airports can get higher
efficiency as compared to those that are public by conducting six different scales.
However, the difference is only slight. In general, only Singapore (SIN) and Paris
(ORY) get relative lower efficiency scores, and the efficiency of the other airports
is are higher.
164
Ch 6
Table: 6.25: Efficiency scores obtained using the AHP/DEA model: VS II

DMU BCC 1-9 𝑒 !!! 2!!! 9/9-9/1 10/10-18/2 ∅ mapping
Europe
Amsterdam
0.9010 0.8460 0.8460 0.8460 0.8460 0.8460 0.8460
(AMS)
Barcelona (BCN) 1 1 1 1 1 1 1
Frankfurt (FRA) 0.9996 0.9115 0.9115 0.9115 0.9115 0.9115 0.9115
Istanbul (IST) 0.7046 0.6776 0.6773 0.6773 0.6773 0.6773 0.6773
London (LGW) 1 1 1 1 1 1 1
London (LHR) 1 1 1 1 1 1 1
Madrid (MAD) 1 1 1 1 1 1 1
Munich (MUC) 0.9031 0.5153 0.5153 0.5153 0.5153 0.5153 0.5153
Paris (CDG) 1 1 1 1 1 1 1
Paris (ORY) 0.4810 0.4405 0.4405 0.4405 0.4405 0.4405 0.4405
Rome (FCO) 0.7629 0.6711 0.6711 0.6711 0.6711 0.6711 0.6711
Zurich (ZRH) 1 1 1 1 1 1 1
Mean 0.8960 0.8385 0.8385 0.8385 0.8385 0.8385 0.8385
Asia-Pacific
Bangkok (BKK) 0.8229 0.8212 0.8212 0.8212 0.8212 0.8212 0.8212

Beijing (PEK) 1 1 1 1 1 1 1
Guangzhou
0.7039 0.7355 0.7355 0.7355 0.7355 0.7355 0.7355
(CAN)
Hong Kong
1 1 1 1 1 1 1
(HKG)
Incheon (ICN) 0.9061 0.9061 0.9061 0.9061 0.9061 0.9061 0.9061
Kuala Lumpur
0.6493 0.6137 0.6137 0.6137 0.6137 0.6137 0.6137
(KUL)
Osaka (KIX) 1 1 1 1 1 1 1
Tokyo (HRT) 1 1 1 1 1 1 1
Shanghai (PVG) 0.8065 0.7656 0.7656 0.7656 0.7656 0.7656 0.7656
Singapore (SIN) 0.4579 0.4294 0.4294 0.4294 0.4294 0.4294 0.4294
Shenzhen (SZX) 1 1 1 1 1 1 1
Sydney (SYD) 1 1 1 1 1 1 1
Mean 0.8622 0.8560 0.8560 0.8560 0.8560 0.8560 0.8560
Mean of all
0.8791 0.8472 0.8472 0.8472 0.8472 0.8472 0.8472
samples
165
Ch 6
Table 6.26: Airport efficiency under different ownership: VS II

DMU BCC 1-9 𝑒 !!! 2!!! 9/9-9/1
category* 18/2 mapping
Amsterdam
(A) 0.9010 0.8460 0.8460 0.8460 0.8460 0.8460 0.8460
(AMS)
Istanbul (IST) (A) 0.7046 0.6776 0.6773 0.6773 0.6773 0.6773 0.6773
London (LGW) (A) 1 1 1 1 1 1 1
London (LHR) (A) 1 1 1 1 1 1 1
Paris (CDG) (A) 1 1 1 1 1 1 1
Paris (ORY) (A) 0.4810 0.4405 0.4405 0.4405 0.4405 0.4405 0.4405
Rome (FCO) (A) 0.7629 0.6711 0.6711 0.6711 0.6711 0.6711 0.6711
Zurich (ZRH) (A) 1 1 1 1 1 1 1
Sydney (SYD) (A) 1 1 1 1 1 1 1
Mean 0.8722 0.8484 0.8483 0.8483 0.8483 0.8483 0.8483
S.D 0.1856 0.2067 0.2068 0.2068 0.2068 0.2068 0.2068
Barcelona
(B) 1 1 1 1 1 1 1
(BCN)
Frankfurt (FRA) (B) 0.9996 0.9115 0.9115 0.9115 0.9115 0.9115 0.9115
Madrid (MAD) (B) 1 1 1 1 1 1 1
Munich (MUC) (B) 0.9031 0.5153 0.5153 0.5153 0.5153 0.5153 0.5153
Bangkok (BKK) (B) 0.8229 0.8212 0.8212 0.8212 0.8212 0.8212 0.8212
Beijing (PEK) (B) 1 1 1 1 1 1 1
Guangzhou
(B) 0.7039 0.7355 0.7355 0.7355 0.7355 0.7355 0.7355
(CAN)
Hong Kong
(B) 1 1 1 1 1 1 1
(HKG)
Incheon (ICN) (B) 0.9061 0.9061 0.9061 0.9061 0.9061 0.9061 0.9061
Kuala Lumpur
(B) 0.6493 0.6137 0.6137 0.6137 0.6137 0.6137 0.6137
(KUL)
Osaka (KIX) (B) 1 1 1 1 1 1 1
Tokyo (HRT) (B) 1 1 1 1 1 1 1
Shanghai (PVG) (B) 0.8065 0.7656 0.7656 0.7656 0.7656 0.7656 0.7656
Singapore (SIN) (B) 0.4579 0.4294 0.4294 0.4294 0.4294 0.4294 0.4294
Shenzhen (SZX) (B) 1 1 1 1 1 1 1
Mean 0.8833 0.8466 0.8466 0.8466 0.8466 0.8466 0.8466
S.D. 0.1663 0.1946 0.1946 0.1946 0.1946 0.1946 0.1946
166
Ch 6
6.4.4 RELATIVE EFFICIENCY ANALYSIS IN DIFFERENT GROUPS: VS I

As mentioned in Chapter 5 (see Section 5.3), there were two different groups of experts
interviewed in this research. In this section, the results of the relative efficient scores
are divided into two different groups, those viewed from an academic perspective and
those viewed from a practical perspective. An examination of Table 6.27 reveals that,
from an academic perspective, 5 of the 24 airports are operated relatively inefficiently,
and from a practical perspective, 6 of the 24 airports are operated relatively
inefficiently when using an AHP/DEA integrated model that is based on alternative
judgement scales to calculate airport efficiency. A number of points emerge from Table
6.27, including:
(1) From both standpoints, there are relatively more efficient airports than inefficient
airports.
(2) From both standpoints, in regard to the relative inefficient scores (which are
calculated using different scales) it is difficult to determine the differences until the
fifth or sixth digit after the decimal point.
(4) The relative efficiency scores between European airports and Asian airports are
very close.
(5) From an academic viewpoint, Frankfurt is not an efficient airport but is only
slightly lower than Paris (ORY) and Singapore, which have lower efficiency as
compared to the results shown from experts from practice.
(6) The efficiency scores and ranks in these two groups are similar to the results shown
in Table 6.23.
(7) According to ownership category, in general, the results show that the experts from
practice felt that private airports can achieve higher efficiency as compared to
those that are public (i.e. 0.9839 vs. 0.9694).
6.4.5 RELATIVE EFFICIENCY ANALYSIS IN A DIFFERENT GROUP: VS II

This section discusses the relative efficiency scores on the basis of VS II in two groups.
Table 6.28 reveals that, from both standpoints, 12 of the 24 airports are operated
relative inefficiency as shown when conducting the AHP/DEA integrated models that
167
Ch 6
are based on alternative scales to calculate airport efficiency. A number of points

emerge from Table 6.28, including:
(1) For both standpoints, there are as many relative efficient airports as there are those
that are inefficient.
(2) From both standpoints, in regard to the relative efficiency scores (which are
calculated using different scales) it is difficult to determine the differences until the
fifth digit after the decimal point.
(4) The relative efficiency scores between European airports and Asian airports are
very close.
(5) From an academic viewpoint, Munich, Guangzhou, Shanghai and Singapore have
lower efficiency than was expressed by experts from practice. Among these
airports, only Singapore has significantly lower efficiency.
(6) The efficiency scores and rank in these two groups are similar to the results shown
in Table 6.25.
(7) According to ownership category, in general, the results on both sides show that
private airports can achieve higher efficiency than those that are public (i.e. 0.8483
vs. 0.8421 from academia and 0.8849 vs. 0.8532 from practice). The efficiency
scores from academia are very close.
168
Table 6.27: Relative efficiency scores obtained by AHP/DEA model by groups: VS I
Academia Practice
DMU 1-9 𝑒 !!! 2!!! 9/9-9/1 10/10-18/2 ∅ mapping 1-9 𝑒 !!! 2!!! 9/9-9/1 10/10-18/2 ∅ mapping
Amsterdam (AMS) 1 1 1 1 1 1 1 1 1 1 1 1
Barcelona (BCN) 1 1 1 1 1 1 1 1 1 1 1 1
Frankfurt (FRA) 0.9992 0.9992 0.9992 0.9992 0.9992 0.9992 1 1 1 1 1 1
Istanbul (IST) 1 1 1 1 1 1 1 1 1 1 1 1
London (LGW) 1 1 1 1 1 1 1 1 1 1 1 1
London (LHR) 1 1 1 1 1 1 1 1 1 1 1 1
Madrid (MAD) 1 1 1 1 1 1 1 1 1 1 1 1
Munich (MUC) 1 1 1 1 1 1 1 1 1 1 1 1
Paris (CDG) 1 1 1 1 1 1 1 1 1 1 1 1
Paris (ORY) 0.655078 0.655073 0.655083 0.655075 0.655073 0.655080 0.855066 0.855096 0.855076 0.855070 0.855084 0.855092
Rome (FCO) 1 1 1 1 1 1 1 1 1 1 1 1
Zurich (ZRH) 1 1 1 1 1 1 1 1 1 1 1 1
Bangkok (BKK) 1 1 1 1 1 1 1 1 1 1 1 1
Beijing (PEK) 1 1 1 1 1 1 1 1 1 1 1 1
Guangzhou (CAN) 1 1 1 1 1 1 1 1 1 1 1 1
Hong Kong (HKG) 1 1 1 1 1 1 1 1 1 1 1 1
Incheon (ICN) 0.941803 0.941805 0.941804 0.941782 0.941805 0.941791 0.941794 0.941804 0.941809 0.941813 0.941812 0.941787
Kuala Lumpur
0.897542 0.897542 0.897544 0.897542 0.897542 0.897542 0.897542 0.897542 0.897542 0.897542 0.897541 0.897542
(KUL)
Osaka (KIX) 1 1 1 1 1 1 1 1 1 1 1 1
Tokyo (HRT) 1 1 1 1 1 1 1 1 1 1 1 1
Shanghai (PVG) 1 1 1 1 1 1 1 1 1 1 1 1
Singapore (SIN) 0.631753 0.631762 0.631758 0.631725 0.631762 0.631748 0.701763 0.701790 0.701755 0.701735 0.701736 0.701751
Shenzhen (SZX) 1 1 1 1 1 1 1 1 1 1 1 1
Sydney (SYD) 1 1 1 1 1 1 1 1 1 1 1 1
Table 6.28: Relative efficiency scores obtained by AHP/DEA model by groups: VS II
169
Academia Practice
DMU 1-9 𝑒 !!!
2!!!
9/9-9/1 10/10-18/2 ∅ mapping 1-9 𝑒 !!!
2!!!
9/9-9/1 10/10-18/2 ∅ mapping
Amsterdam (AMS) 0.845978 0.845978 0.845978 0.845978 0.845978 0.845978 0.845978 0.845978 0.845978 0.845978 0.845978 0.845978
Barcelona (BCN) 1 1 1 1 1 1 1 1 1 1 1 1
Frankfurt (FRA) 0.911448 0.911448 0.911448 0.911448 0.911448 0.911448 0.911448 0.911448 0.911448 0.911448 0.911448 0.911448
Istanbul (IST) 0.677262 0.677262 0.677262 0.677262 0.677262 0.677262 0.677262 0.677262 0.677262 0.677262 0.677262 0.677262
London (LGW) 1 1 1 1 1 1 1 1 1 1 1 1
London (LHR) 1 1 1 1 1 1 1 1 1 1 1 1
Madrid (MAD) 1 1 1 1 1 1 1 1 1 1 1 1
Munich (MUC) 0.504758 0.504771 0.504734 0.504779 0.504766 0.504771 0.515271 0.515271 0.515271 0.515271 0.515271 0.515271
Paris (CDG) 1 1 1 1 1 1 1 1 1 1 1 1
Paris (ORY) 0.440539 0.440539 0.440539 0.440539 0.440539 0.440539 0.440539 0.440539 0.440539 0.440539 0.440539 0.440539
Rome (FCO) 0.6711 0.6711 0.6711 0.6711 0.6711 0.6711 0.6711 0.6711 0.6711 0.6711 0.6711 0.6711
Zurich (ZRH) 1 1 1 1 1 1 1 1 1 1 1 1
Bangkok (BKK) 0.821187 0.821187 0.821187 0.821187 0.821187 0.821187 0.821187 0.821187 0.821187 0.821187 0.821187 0.821187
Beijing (PEK) 1 1 1 1 1 1 1 1 1 1 1 1
Guangzhou (CAN) 0.724465 0.724465 0.724465 0.724465 0.724465 0.724465 0.735465 0.735465 0.735465 0.735465 0.735465 0.735465
Hong Kong (HKG) 1 1 1 1 1 1 1 1 1 1 1 1
Incheon (ICN) 0.906072 0.906072 0.906072 0.906072 0.906072 0.906072 0.906072 0.906072 0.906072 0.906072 0.906072 0.906072
Kuala Lumpur
0.613665 0.613688 0.613645 0.613633 0.613671 0.613669 0.613687 0.613687 0.613687 0.613687 0.613687 0.613687
(KUL)
Osaka (KIX) 1 1 1 1 1 1 1 1 1 1 1 1
Tokyo (HRT) 1 1 1 1 1 1 1 1 1 1 1 1
Shanghai (PVG) 0.755667 0.755477 0.755762 0.755961 0.755772 0.755884 0.765596 0.765596 0.765596 0.765596 0.765596 0.765596
Singapore (SIN) 0.339368 0.339355 0.339367 0.339352 0.339324 0.393345 0.529336 0.529399 0.529357 0.529368 0.529336 0.529347
Shenzhen (SZX) 1 1 1 1 1 1 1 1 1 1 1 1
Sydney (SYD) 1 1 1 1 1 1 1 1 1 1 1 1
Mean 0.847207 0.847207 0.847207 0.847207 0.847207 0.847207 0.847207 0.847207 0.847207 0.847207 0.847207 0.847207
170
Ch 6
6.4.6 HYPOTHESIS TESTING: VS I

In this section, after conducting integrated AHP/DEA models, having established the
efficiency rankings for the sample airports, a hypothesis related to the rankings
obtained needs to be examined. After calculating, the results that were computed using
alternative scales are found to be very similar; therefore, only the results calculated
using the 1-9 scale are used to test the hypothesis.
public management.
To compare the results easily, the 24 sample airports are classified into two groups:
those operate by the public sector and those operated by private companies. There are
15 airports under the public sector (most of them in the Asia-Pacific) and nine airports
under private ownership. From Table 6.29, it can be seen that the z-value is −0.798 with
a significance level p of p = 0.599. The probability value p is not less than or equal
to 0.05, so the result is not significant. There is no statistically significant difference in
airport efficiency between privately operated and publicly operated airports as
calculated using the integrated AHP/DEA model.
Table 6.29: Mann-Whitney test of differences in efficiency: VS I

Asymptotic
Reference significance
U-test Z-test
(two-tailed)
vs. 58 −0.798 0.599
6.4.7 Hypothesis testing: VS II

From Table 6.30, the z-value is -0.191 with a significance level p of p = 0.861. The
probability value p is not less than or equal to 0.05, so the result is not significant.
There is no statistically significant difference in airport efficiency between privately
operated and publicly operated airports.
171
Ch 6

Asymptotic
Reference significance
U-test Z-test
(two-tailed)
vs. 64.5 −0.191 0.861

Many different analysis procedures and results are described. In this section, cross
discussions about efficiency in two variables sets, two different models (basic DEA
models and AHP/DEA models), two different groups (academia and practice), and
alternative scales in AHP analysis are presented.
6.5.1 DIFFERENT VARIABLE SETS

The most common sensitivity analysis technique in the basic DEA model is to change
the numbers of DMUs or the numbers of variables (Cooper et al., 2006). In this
research, reducing the numbers of variables is more suitable because the numbers of
similar DMUs (airports) are limited in these two regions. Among the sensitivity
analysis processes, in the first part, an output-oriented, variable return-to-scale analysis
(DEA-BCC model) is used in stepwise modelling for selecting the core DEA variables;
however, because of research limitations, only two variables sets are discussed.
Table 6.31 shows the evidence to support the assumption of the DEA method (which
states that the number of variables will influence the relative efficiency). Firstly, when
the variables were reduced from six inputs and four outputs to three inputs and three
outputs, the numbers of inefficient DMUs are shown to increase from five to twelve.
Secondly, the average efficiency score decreased from 0.9719 to 0.8791, and the
average change is −0.0928. The results from Table 6.31 reveal that a lower number of
variables can help to increase the discriminatory power of the DEA model.
Furthermore, even when using integrated AHP/DEA models to evaluated airport
efficiency, the results are also influenced by the number of variables (see Tables 6.22
and 6.24).
172
Ch 6
Table: 6.31: A comparison of efficiency scores between the two variable sets
VS I VS II
No. DMUs Ownership Category BCC Model (VRS) BCC Model (VRS)
European
Amsterdam
1 B 1 0.9010
(AMS)
Barcelona
2 F 1 1
(BCN)
3 Frankfurt (FRA) CE 1 0.9996
4 Istanbul (IST) D 1 0.7046
5 London (LGW) G 1 1
6 London (LHR) G 1 1
7 Madrid (MAD) F 1 1
8 Munich (MUC) E 1 0.9030
9 Paris (CDG) B 1 1
10 Paris (ORY) B 0.6551 0.4810
11 Rome (FCO) B 0.9983 0.7629
12 Zurich (ZRH) B 1 1
Asia-Pacific
13 Bangkok (BKK) C 1 0.8229
14 Beijing (PEK) C 1 1
Guangzhou
15 C 1 0.7039
(CAN)
Hong Kong
16 F 1 1
(HKG)
17 Incheon (ICN) F 0.9418 0.9061
Kuala Lumpur
18 F 0.8975 0.6493
(KUL)
19 Osaka (KIX) CE 1 1
20 Tokyo (HRT) F 1 1
21 Shanghai (PVG) F 1 0.8065
22 Singapore (SIN) A 0.8318 0.4579
23 Shenzhen (SZX) C 1 1
24 Sydney (SYD) B 1 1
Mean 0.9719 0.8791
S.D 0.0786 0.1698
Average change in efficiencies -0.0928
No. of efficient DMUs 19 12
The reason why the discriminatory power of the result calculated using VS II is higher
than that of VS I can be discovered from the weight distribution of variables in Table
6.32.
In the DEA model, the input and output weights are automatically calculated; hence,
calculating the relative efficiency using the DEA model will help the DMUs to select
these relatively better variables (Kong and Fu, 2012). From Table 6.32, it can be clearly
173
Ch 6
seen that there are many zeros in the weights of inefficient airports for the selected
variables in VS I (which is unreasonable when evaluating an airport’s peer
performance). In VS II, the situation is improved, which means first of all, the number
of variables will influence the result when evaluating airport efficiency. Secondly, these
variables can place more influence on airport efficiency. With regard to the
discriminatory power and the distribution of variable weights, VS II makes the DEA
results more accurate and reflects the real decision-making situation of an airport
performance evaluation. Consequently, there are still some zero weights in VS II,
which means there are still some weaknesses that can be improved upon.
6.5.2 DIFFERENT GROUPS

In contrast to the results shown in Chapter 5, the weights of each variable can be
recognised easily when divided into different groups. As shown in Table 5.8, even if
the rank of each variable is the same, the weights of each variable vary according to the
different groups. In addition, this situation still occurs when using different scales to
conduct AHP analysis. As can be seen in Table 5.22, even if the rank of each variable is
still the same, the weights of each variable vary according to the different groups.
However, after combining the weights with the DEA model, the results do not show the
variations obviously, such as is the case with the results shown in Tables 6.27 and 6.28.
174
Table 6.32: The weight distribution of variables in VS I and VS II
VS I VS II
Terminal Area
Terminal Area
Expenditure
Operational
Movements
Movements
Freight and
Freight and
Employees
Employees
Passengers
Passengers
Number of
Number of
Number of
Number of
Amount of
Number of
Number of
Number of
Amount of
Length of
Revenues
Runways
Runways
Runway
Aircraft
Aircraft
Size of
Size of
Gates
Mails
Mails
Total
DMU
1 Amsterdam (AMS) 0.1512 0.5355 0 0 0.3134 0 0.0498 0.2374 0.7129 0 0.3068 0 0 0.3068 0 0
2 Barcelona (BCN) 0.4693 0 0 0.5307 0 0 0 0 0 0 0.4693 0 0.5307 0.4693 0 0.5307
3 Frankfurt (FRA) 0 0.3922 0.2917 0 0 0 0 0.1733 0.3918 0.4349 0 0.2924 0 0 0.2924 0
4 Istanbul (IST) 0.3089 0.6911 0 0 0 0 0 0 0 0 0.1984 0 0.0215 0.1984 0 0.0215
5 London (LGW) 0.4086 0 0.5914 0 0 0 0 0 0 0 0.4086 0.5914 0 0.4086 0.5914 0
6 London (LHR) 0 0 0.8145 0.1855 0 0 0 0 0 0 0 0.8145 0.1855 0 0.8145 0.1855
7 Madrid (MAD) 0.4418 0.5582 0 0 0 0 0 0 0 0 0.3636 0.5807 0.0557 0.3636 0.5807 0.0557
8 Munich (MUC) 0 0.2064 0.4127 0 0.2998 0.08116 0 0 0 0 0 0.9949 0 0 0.9949 0
9 Paris (CDG) 0 0.4300 0.5700 0 0 0 0 0 0.9883 0.0117 0.1781 0.1227 0.6532 0.1781 0.1227 0.6532
10 Paris (ORY) 0 0.3477 0 0.53250 0.1228 0.1135 0.0409 0 0.3498 0.6093 0.4224 0.8466 0 0.4224 0.8466 0
11 Rome (FCO) 0 0 0 0.29434 0 0.3600 0.0944 0 0.5061 0.3995 0 0 0.8417 0 0 0.8417
12 Zurich (ZRH) 0.0783 0 0 0.7447 0 0.1769 0.5510 0.3119 0.1371 0 0.2336 0 0.7664 0.2336 0 0.7664
13 Bangkok (BKK) 0 0.2271 0.2269 0.1829 0.2927 0.0703 0 0 0 0 0.3052 0.7732 0 0.3052 0.7732 0
14 Beijing (PEK) 0 0 0.2253 0 0.2813 0.4934 0 0 0 0 0.2155 0.6161 0 0.2155 0.6161 0
15 Guangzhou (CAN) 0 0.1304 0.3136 0.0521 0.3018 0.1767 0.5782 0 0.4218 0 0.4992 0.4762 0.4977 0.4992 0.4762 0.4977
16 Hong Kong (HKG) 0 0.7280 0.2720 0 0 0 0 0 0 0 0.6591 0.3409 0 0.6591 0.3409 0
17 Incheon (ICN) 0 0.8553 0 0 0 0.3714 0 0.3560 0 0.6440 0.3183 0 0.2632 0.3183 0 0.2632
18 Kuala Lumpur (KUL) 0.0677 0 7.2334 0 0 2.2682 0 0 0 0 0.3543 0.8410 0 0.3543 0.8410 0
19 Osaka (KIX) 0.9111 0 0 0.0889 0 0 0 0 0.0114 0.9886 0.4050 0 0.1256 1.4050 0 0.1256
20 Tokyo (HRT) 0.1744 0 0.8256 0 0 0 0 0 0.5303 0.4697 0.0161 0 0 1.0161 0 0
21 Shanghai (PVG) 0 0 0 0 0 0 0.4175 0.5825 0 0 0 0 0 0 0 0
22 Singapore (SIN) 0 0 0 0 0 0.5493 0 0 0 0 0.3565 0.4350 0 0.3565 0.4350 0
23 Shenzhen (SZX) 0 0.5550 0.3435 0.0931 0 0.0084 0 0 0 0 0 0 0.0691 0 0 0.0691
24 Sydney (SYD) 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
175
Ch 6
6.5.3 ALTERNATIVE JUDGEMENT SCALES IN THE AHP/DEA MODELS

Another sensitivity analysis in this research is carried out by using different analysis
models. In this chapter, an integrated AHP/DEA model is applied and also conducted
with an alternative scale in AHP analysis.
Tables 6.23 and 6.25 show the efficiency scores calculated using the AHP/DEA-BCC
model using alternative scales in two variables sets. The scores reveal that the
differences between the BCC model and the AHP/DEA model can be identified easily.
However, differences among the alternative scales did not appear. These results are in
agreement with those of Jyoti et al. (2008). However, when comparing the scores with
those of other scales, the variations are difficult to recognise, even when the number of
inefficient DMUs are all the same. In this case, the answer to Research Question 3 is
not significant.
There are some reasons why the relative scores are so close when using alternative
scales. Firstly, the weight distribution of variables for the AHP/DEA model may cause
the results. One could easily find that there are many zero weights for selected variables
in these six different scale data sets, which is unreasonable when evaluating an airport’s
peer performance. This happens because the input and output weights are automatically
calculated, which can cause the relative scores to be very close (see the tables attached
as Appendix IV). Secondly, it can be seen in Table 6.20 that the weights among the
input and output variables are very close in the last column of the value of “Standard
Deviation”, which brings the weighted data closer (see Appendix III). Therefore, the
results for the AHP/DEA models do not accurately reflect the real decision-making
situation for airport efficiency evaluation. It is necessary to find another method to
assess airport efficiency.
176
Ch 6
6.6 SUMMARY
This chapter conducted and presented the results of an empirical analysis. The
empirical analysis was evaluated by means of different models (from a traditional DEA
model (CCR and BCC model) to an integrated AHP/DEA model), different variable
sets, (i.e. VS I and VS II), and by using alternative scales in an AHP analysis.
The correlation coefficients analysis shows the variables that were selected in both
variable sets to be robust. In addition, the Mann-Whitney test shows that the hypothesis
that airports under private management are more efficient than those under public
management was not significant in both variables sets. In addition, this research used an
integrated AHP/DEA model to evaluate airport efficiency, which was proposed by Jyoti
et al. (2008). Using Saaty’s 1-9 scale, it can be seen that this AHP/DEA model has
provided a fair and useful technique by which to evaluate the performance of airports in
terms of their relative efficiencies, not only on the basis of the quality of variables but
also based on the integration of diverse viewpoints. However, when introducing
alternative scales to the AHP model, the results were not reliable, and the differences
were not obvious. Therefore, this kind of calculation method cannot provide realistic
results when benchmarking airports.
An AHP/DEA-AR model is conducted in the next chapter in order to prevent

impractical variable weights and to verify Research Question 1. The weights of the
variables obtained from an AHP survey are passed into the assurance region of the
DEA-AR model for empirical analysis. The empirical results can help to increase
discriminatory power, and they can also help to overcome the shortcomings of the
AHP/DEA model (such as illogical local weights).
177
Ch7
Airport Efficiency Analysis II
CHAPTER 7
AIRPORT EFFICIENCY ANALYSIS II:

AHP/DEA-AR MODEL
Chapter 6 focused on presenting the airport efficiency analysis by means of basic DEA
models and integrated AHP/DEA models. However, the discriminatory powers of these
models are not good enough. Therefore, in this chapter, by conducting sensitivity
analysis, another AHP/DEA-AR model is applied to assess airport efficiency. The
concepts and analysis process for the efficiency scores are described in first section.
The second and third sections describe the efficiency score computed on VS I and VS II.
The fourth section, the results are compared with the DEA-BCC model and integrated
AHP/DEA models. Some additional thoughts on airport efficiency analysis are
presented in the last section.
178
Ch7
7.1 AHP/DEA-AR MODEL

7.1.1 CONCEPTS OF DEA-AR MODEL
Chapter 6 described how the DEA was combined with the AHP to form an AHP/DEA
model for weight derivation and aggregation in the AHP, which was used to evaluate
airport efficiency; however, these results were not robust. In this chapter, a DEA model
with Assurance Region (AR) for priority derivation in the AHP is proposed, which is
referred to as the DEA-AR model. This DEA-AR model can help to solve the
shortcomings of the AHP/DEA model, such as illogical local weights, over insensitivity
to some comparisons, information loss, and overestimation of some local weights. It
can also provide a better priority estimate and better decision conclusions than the
AHP/DEA model (Wang et al. 2008; Liang and Fang 2011; Kong and Fu 2012). In
addition, it can provide the answers to the research questions.
In practice, when conducting efficiency evaluation, DMUs do not always allow the
choice of the best variable or place higher weights on particular variables because the
best weights choice could result in extreme weight distributions (Doyle and Green
1994). In addition, this course of action is not usually employed in actual airport
efficiency assessment practices. Therefore, this study employs the DEA-AR approach
that was developed by Thompson et al. (1986) to avoid this unreasonable distribution of
weights. The DEA-AR model can allow weights to vary within a range by imposing
constraints on the relative magnitudes of the weights for special items. The AR model
is used to impose restrictions on the upper bound (𝑈!" ) and lower bound (𝐿!" ) of a ratio
of the weights of two variables (𝑢! 𝑢! ), as follows:
!
𝐿!" ≤ ! ! ≤ 𝑈!" . (1)
!
This research employs the DEA-AR model to reflect the relative importance of input
and output variables. Therefore, by adding the constraints in Equation (1) into the BCC
model, the DEA-AR-BCC model (DEA-AR model in short) can be obtained. However,
the question of how the lower and upper bounds are determined needs to be addressed.
Some studies determine the lower and upper bounds based on the weight analysis of a
DEA model (Thompson et al. 1986), and some studies determine them based on expert
179
Ch7
opinions (such as, Zhu 1996; Seifert and Zhu 1998; Takamura and Tone 2003) a
notable example of this being the AHP.
7.1.2 LITERATURE FOR THE DEA-AR MODEL AND THE AHP/DEA-AR MODEL
In an early study, the DEA-AR model was used to evaluate the efficiency of 83 farms
in Kansas (Thompson et al. 1990). After conducting the DEA-AR model, Thompson et
al. (1990) were able to reduce the number of efficient DMUs from 23 to 8. Their results
revealed that the DEA-AR model could provide more accurate efficiency scores than
traditional DEA models. Taylor et al. (1997) used DEA and Linked-Cone Assurance
Region (LC-AR) models to investigate the efficiency and profitability potential of
Mexican banks as they engaged in activities that incurred interest and non-interest
expenses and produced income. In addition, Lee et al. (2009) employed three DEA
models (i.e. DEA-BCC model, the DEA-AR model, and output integration) to measure
and compare the performance of national research and development programs. The
results provided policy implications for effectively formulating and implementing
national research and development programs. Traditional DEA models and Additive-
AR models were used by Liang and Fang (2011) to evaluate the productivity and
quality performance of TFT-LCD suppliers. Their results were valuable in terms of
optimising the selection of an appropriate supplier for the TFT-LCD company. These
previous studies reveal that the DEA-AR model can help improve the accuracy of
efficiency evaluation although they were not conducted in the context of air transport.
Similarly, the model combining the AHP and DEA-AR model has been previously
employed in other fields of studies. For example, the AHP/DEA-AR model was first
used in a study that evaluated the performance of the Nanjing Textiles Corporation
(Zhu 1996). Seifert and Zhu (1998) used this model to investigate excess and deficits in
regard to Chinese industrial productivity for the years 1953-1990 and found that the
weights of ARs could be obtained through expert opinions using the Delphi and AHP
approaches.
Several applications of the AHP/DEA-AR model can also be found for the public sector.
For example, Takamura and Tone (2003) used the AHP/DEA-AR model to provide two
possible locations to relocate Japanese government agencies out of Tokyo. Meanwhile,
Meng et al. (2008) combined AHP, AR, and a two-level DEA to evaluate the research
180
Ch7
performance of research institutes in the Chinese Academy of Sciences. Finally, Wang

et al. (2008) proposed a DEA model with AR for priority derivation in the AHP, which
is referred to as the DEA-AR model.
7.1.3 AHP/DEA-AR MODEL IN THIS RESEARCH

In this research, the AHP/DEA-AR model is applied to evaluate the efficiency of
airports in both Europe and the Asia-Pacific regions. This approach can make
contributions to the literature in the following two aspects: Firstly, an AHP survey is
conducted with airport managers and researchers to derive the upper and lower bounds
of weight restrictions of the ARs. All the variables are pre-selected by a representative
literature review of airport efficiency studies over the last 20 years. The reliability is
also increased by means of interviews with a number of industry experts. Secondly, and
perhaps the most important contribution of this research, is the fact that this is the first
study that has proposed to use the AHP/DEA-AR model to assess airport efficiency.
That is, instead of assessing airport efficiency from the perspective of airport operators,
the weights of the variables are considered from two different vewpoints (i.e. practice
and academia). The process for conducting the AHP/DEA-AR model in this research is
described below.
As mentioned in previous chapters, a total of 35 questionnaires were sent out, and a

total of 25 completed questionnaires were collected. The weight of each variable was
calculated from the answers of the respondents. The calculations were conducted using
Super Decision, which is a dedicated AHP software program. The results were then
cross checked using the Maple 14 software program. The reliability of the weights was
considered acceptable if the consistency in the respondent’s questionnaire, or the
Consistency Index (CI), < 0.1. Where the CI was not considered acceptable, the
questionnaires were excluded from the effective responses. In this research, an effective
rate of 62.86% was achieved for 22 acceptable questionnaires. The main difference
between the AHP/DEA model and the AHP/DEA-AR model is the method by which
the AHP result is derived. When conducting the AHP/DEA model, the geometric mean
method was used to solve the eigenvalue of pair-wise comparison matrix, which was in
turn used to calculate the local weight of each level on a spread-sheet (Saaty 2003).
When conducting AHP/DEA-AR model, the preference of each respondent was needed
to calculate the individual results (Kong and Fu 2012).
181
Ch7
7.2 AIRPORT EFFICIENCY ANALYSIS: VS I

As mentioned in Chapter 6, two different variable sets are adopted with the AHP/DEA-
AR model. VS I (which includes ten variables) is conducted to evaluate airport
performance in this section.
7.2.1 BOUNDS CALCULATION

When conducting the AHP/DEA-AR model, the weights of each variable are displayed
by each respondent. The results derived from AHP analysis are then used as a guideline
for setting the upper and lower bounds. The respondents’ derived weights attached to
the input and output variables of performance are shown in Table 7.1. Respondents 1 to
15 are drawn from practice while respondents 16 to 22 come from academia. From the
weight averages of these two groups, the respondents from practice are found to
relatively place more emphasis on the financial variables than on other variables (WI6:
Operational expenditure and WO4: Total revenue). This result is similar to common
opinion because, generally speaking, an airport authority should concentrate more on
financial variables.
The results derived from the AHP analysis then served as a guideline for setting the
upper and lower bounds in the AHP/DEA-AR model. To incorporate these weights in
DEA-AR model, pair-wise divisions between the weights were made. The largest and
smallest values of each weight ratio for all respondents were then found, and the upper
and lower bounds values of this weight ratio were then constructed. For example, for
Respondent 11 (as shown in Table 7.1) the ratio 𝑊𝐼! 𝑊𝐼! takes on a value of
0.0646/0.0238. The ratio 𝑊𝐼! 𝑊𝐼! for the other 21 respondents can be calculated by
this order. Therefore, the highest 𝑊𝐼! 𝑊𝐼! =2.7143 from Respondent 11 is used as the
upper bound of the ratio 𝑊𝐼! 𝑊𝐼! , and the smallest 𝑊𝐼! 𝑊𝐼! is 0.0765 from
Respondent 2 is used as the lower bound. Therefore, the range of 𝑊𝐼! 𝑊𝐼! is 0.0765
≤ 𝑊𝐼! 𝑊𝐼! ≤ 2.7143. This ratio weight inequality constraint is then incorporated into
the AHP/DEA-AR model. Other ranges (or upper and lower bounds) of ratio weights
can be found in Table 7.2. The upper and lower bounds that are addressed by the AHP
alternative scales are listed in Appendix IV.
182
Ch7
Table 7.1: AHP weights of input and output variables of respondents (1-9 scale)
Input Variables Output Variables
Number Size of Amount
Number of Number Length of Operational Number of Aircraft Total
Variables Employees of Gates
of Terminal
Runway Expenditure Passengers
of Freight
Movements Revenues
Runways Area and Mails
Respondent WI1 WI2 WI3 WI4 WI5 WI6 WO1 WO2 WO3 WO4
1 0.0113 0.0697 0.0661 0.0755 0.0275 0.7500 0.0632 0.0505 0.0113 0.875
2 0.0046 0.0601 0.0252 0.0382 0.0148 0.8571 0.1118 0.0704 0.0177 0.8000
3 0.0205 0.2157 0.099 0.1368 0.0281 0.5000 0.0667 0.0667 0.0095 0.8571
4 0.0103 0.0952 0.0389 0.0899 0.0158 0.7500 0.2222 0.2222 0.0556 0.5000
5 0.0056 0.0627 0.0302 0.0543 0.0139 0.8333 0.2273 0.2273 0.0455 0.5000
6 0.0063 0.0641 0.0288 0.0518 0.0157 0.8333 0.0857 0.0857 0.0286 0.8000
7 0.0058 0.0669 0.0242 0.0148 0.0551 0.8333 0.1062 0.043 0.0175 0.8333
8 0.0308 0.1803 0.2993 0.2303 0.0925 0.1667 0.3815 0.3468 0.105 0.1667
9 0.0318 0.205 0.0744 0.3522 0.1699 0.1667 0.0912 0.0574 0.0181 0.8333
10 0.005 0.0385 0.0329 0.0531 0.0133 0.8571 0.6086 0.2004 0.066 0.1250
11 0.0646 0.0238 0.0155 0.0201 0.0189 0.8571 0.0416 0.0262 0.1322 0.8000
12 0.0332 0.0189 0.1016 0.0125 0.0337 0.8000 0.3304 0.1041 0.0656 0.5000
13 0.0257 0.3525 0.1152 0.0769 0.2297 0.2000 0.3185 0.1291 0.0524 0.5000
14 0.0269 0.2047 0.4294 0.0558 0.1164 0.1667 0.3694 0.141 0.3229 0.1667
15 0.0129 0.0785 0.0371 0.0345 0.0371 0.8000 0.4405 0.1388 0.0874 0.3333
Average 0.0197 0.1158 0.0945 0.0864 0.0588 0.6248 0.2310 0.1273 0.0690 0.5727
16 0.1898 0.1421 0.2201 0.1604 0.1209 0.1667 0.0781 0.3293 0.0926 0.5
17 0.0067 0.0538 0.0226 0.0592 0.0244 0.8333 0.3704 0.3704 0.0926 0.1667
18 0.0179 0.1004 0.0823 0.2445 0.0549 0.5000 0.5269 0.2102 0.0629 0.200
19 0.0099 0.0577 0.0757 0.079 0.0277 0.7500 0.3030 0.3030 0.0606 0.3333
20 0.0134 0.1041 0.0284 0.0839 0.0202 0.7500 0.4700 0.0702 0.2098 0.2500
21 0.031 0.1641 0.314 0.1525 0.1717 0.1667 0.3467 0.3815 0.1050 0.1667
22 0.0207 0.0392 0.0675 0.0198 0.0195 0.8333 0.1071 0.0357 0.1071 0.7500
Average 0.0413 0.0945 0.1158 0.1142 0.0628 0.5714 0.3146 0.2429 0.1044 0.3381
Average
0.0266 0.1090 0.1013 0.0953 0.0601 0.6078 0.2576 0.1641 0.0803 0.4981
in Total
183
Ch7
Table 7.2: Upper and lower bounds of variable weight ratios (1-9 scale)
Input Weight Output Weight
Upper Lower Upper Lower
Ratio Ratio
WI1/WI2 2.7143 0.0765 WO1/WO2 6.6952 0.2372
WI1/WI3 4.1677 0.0626 WO1/WO3 9.2212 0.3147
WI1/WI4 3.2139 0.0732 WO1/WO4 4.8688 0.0722
WI1/WI5 3.4180 0.1053 WO2/WO3 7.0211 0.3333
WI1/WI6 1.1386 0.0054 WO2/WO4 2.2885 0.0476
WI2/WI3 3.6655 0.4767 WO3/WO4 1.9370 0.0111
WI2/WI4 4.5839 0.4106
WI2/WI5 7.6762 0.5608
WI2/WI6 1.7625 0.0236
WI3/WI4 8.1280 0.2112
WI3/WI5 3.6890 0.4392
WI3/WI6 2.5759 0.0181
WI4/WI5 5.6899 0.2686
WI4/WI6 2.1128 0.0156
WI5/WI6 1.1485 0.0173

The results of the DEA-BCC model, the AHP/DEA model, and the proposed
AHP/DEA-AR are summarised in Table 7.3. It can be seen there are eight evaluation
items for airport efficiency. The first column presents the efficiency scores, which are
calculated using the original DEA-BCC model. The second column shows the
efficiency scores, which are calculated using the AHP/DEA model, and the weights of
each variable are considered using a 1-9 scale. The third column shows the efficiency
scores calculated using the DEA-AR model, and the weights of each variable are
considered using a 1-9 scale. Columns four to eight reveal the efficiency scores, which
are computed by the weights of each variable considered in the alternative scales.
An examination of VS I reveals that 5 of the 24 airports are operated relatively

inefficiently when using the BCC model and the integrated AHP/DEA model to
calculate airport efficiency, and the number of efficient airports are quite varied in the
AHP/DEA-AR model. The following points have been developed from this table:
(1) The number of relatively efficient airports is found to be less than that of those that
are inefficient by means of the AHP/DEA-AR model.
(2) The reason for the use of alternative scales is to transform the AHP questionnaire
into numerical scales more accurately. The result shows that in the proposed
184
Ch7
AHP/DEA-AR model, different scales provide different numbers of efficient

airports.
(3) The Standard Deviation (S.D.) of the efficiency score for the entire sample in the
DEA-BCC model is 0.0783, and in the AHP/DEA model, it is 0.0773. Whereas,
the S.D. is from 0.1862, which is calculated by scale 2!!! to 0.2445, which is
calculated from scale ∅ mapping in the proposed AHP/DEA-AR model. Therefore,
these relative efficiency scores and S.D. indicate that the proposed AHP/DEA-AR
model possesses better discriminatory power than the DEA-BCC model and the
AHP/DEA model.
(4) The results from the proposed model show that the mean of efficiency scores for
the whole sample in the 1-9 scale is 0.7231 (see Table 7.3), whereas the average
efficiency scores for European airports is 0.7669, and for Asia-Pacific airports is
0.6794. Generally speaking, the efficiency of European airports is better than those
of the Asia-Pacific region. Furthermore, it is worth noting that some Asian airports
are performing well; for example, Hong Kong and Shenzhen are performing better
than many European airports. However, Singapore Airport only achieved an
efficiency score of 0.2471 among the 24 airports under consideration. The most
reasonable explanation for this is that Singapore Airport has the second largest
terminal area size (1,043,020 𝑚! ) (see Table 6.2), but its output performance is
much lower, being ranked 21st for number of passengers, 17th for amount of freight,
23rd for aircraft movements, and 11th for total revenue. This explains why
Singapore Airport did not achieve higher efficiency scores through the DEA
analysis. Potential root causes for the efficiency score include management
performance objectives that place a greater emphasis on passenger experience and
the introduction of additional terminal capacity. Consequently, although in the
short term, Singapore Airport is under-utilised, it allows for passenger number
growth in the long term.
(5) Among those efficiency scores which were calculated using the AHP/DEA-AR
model, the S.D. value shows that ø mapping scale can provide more discriminatory
power (i.e. 0.2445). It also shows that the AHP/DEA-AR model can also provide
higher discriminatory power than either the DEA-BCC model or the AHP/DEA
model.
185
Ch7
(6) The results shown in Table 7.4 reveal that when applying the BCC model, public
airports can achieve better efficiency, but using the AHP/DEA model, private
airports can achieve higher efficiency than those that are public. In addition, when
applying the AHP/DEA-AR model, the results show that privately operated
airports can get higher efficiency than public ones by conducting six different
scales.
(7) Another benefit of using the AHP/DEA-AR model is to avoid extreme weight
distribution. The weight distribution of output variables for the DEA model is
shown in Table 7.5. Because the output oriented DEA-BCC model is applied in
this research, only output variable distribution needs to be discussed herein.
(Cooper et al., 2006). From Table 7.5, it can be seen that there are many zero
weights for the selected output variables, which is unreasonable when evaluating
airport peer performance. Such an unreasonable situation is not found to exist in
the proposed DEA-AR model, in which all of the output weights are larger than
zero. That implies when using AHP/DEA-AR model to assess airport efficiency,
all the output variables are considered. If comparing the weight distribution of
variables with the integrated AHP/DEA model, the result is once again much better
than that obtained when using the integrated AHP/DEA model.
In regard to the discriminatory power and the distribution of variables weights, the
proposed AHP/DEA-AR model makes the DEA results more accurate and able to
reflect the real decision-making situation for airport performance evaluation.
Consequently, in this research, the AHP/DEA-AR model is found to achieve better
results than either the traditional DEA model or the integrated AHP/DEA model.
186
Ch7
Table 7.3: Efficiency scores obtained using the AHP/DEA-AR model: VS I

AHP/DEA DEA-AR Model
DEA model
DMU BCC 1 2 3 4 5 6
Model 10/10- ∅
1-9 1-9 𝑒 !!! 2!!! 9/9-9/1
18/2 mapping
European
Amsterdam
1 1 0.7589 0.7876 0.7933 0.6924 0.7180 0.6862
(AMS)
Barcelona (BCN) 1 1 0.9088 1 1 0.7027 0.7633 0.6856
Frankfurt (FRA) 1 0.9996 0.8574 0.9412 0.9509 0.7538 0.7842 0.7462
Istanbul (IST) 1 1 0.7360 0.7715 0.8484 0.6356 0.6495 0.6296
London (LGW) 1 1 0.6167 0.6879 0.7341 0.4660 0.5092 0.4539
London (LHR) 1 1 1 1 1 1 1 1
Madrid (MAD) 1 1 0.8645 0.9761 1 0.6276 0.6871 0.6109
Munich (MUC) 1 1 0.4965 0.5901 0.6798 0.3666 0.4095 0.3555
Paris (CDG) 1 1 1 1 1 0.9530 0.9789 0.9467
Paris (ORY) 0.6551 0.8205 0.3541 0.4184 0.4734 0.2408 0.2676 0.2332
Rome (FCO) 0.9983 1 0.6094 0.7208 0.7822 0.4224 0.4714 0.4088
Zurich (ZRH) 1 1 1 1 1 1 1 1
Mean 0.9711 0.9850 0.7669 0.8245 0.8552 0.6551 0.6866 0.6464
Asia
Bangkok (BKK) 1 1 0.6436 0.6689 0.7590 0.5756 0.5923 0.5709

Beijing (PEK) 1 1 0.6579 0.7460 0.7660 0.5420 0.5746 0.5343
Guangzhou
1 1 0.7168 0.7620 0.8603 0.6048 0.6204 0.5986
(CAN)
Hong Kong
1 1 1 1 1 1 1 1
(HKG)
Incheon (ICN) 0.9418 0.9944 0.8531 0.8548 0.8555 0.8488 0.8500 0.8477
Kuala Lumpur
0.8975 0.8975 0.4379 0.4630 0.5418 0.3723 0.3842 0.3685
(KUL)
Osaka (KIX) 1 1 0.5923 0.6036 0.6138 0.5777 0.5800 0.5766
Tokyo (HRT) 1 1 0.6343 0.6417 0.6650 0.6256 0.6274 0.6247
Shanghai (PVG) 1 1 0.7907 0.8032 0.8046 0.7787 0.7822 0.7778
Singapore (SIN) 0.8318 0.6672 0.2471 0.2916 0.3135 0.1905 0.2063 0.1865
Shenzhen (SZX) 1 1 1 1 1 1 1 1
Sydney (SYD) 1 1 0.5793 0.6518 0.7774 0.4296 0.4572 0.4208
Mean 0.9726 0.9633 0.6794 0.7072 0.7464 0.6288 0.6396 0.6255

Mean of all
0.9719 0.9741 0.7231 0.7658 0.8008 0.6419 0.6631 0.6360
samples
S.D. 0.0786 0.0773 0.2141 0.2037 0.1862 0.2421 0.2344 0.2445
187
Ch7
Table 7.4: Airport efficiency with different ownership: VS I

AHP/DEA-AR
Ownership AHP/DEA
DMU BCC 10/10- ∅
category* model 1-9 𝑒 !!! 2!!! 9/9-9/1
18/2 mapping
Amsterdam
(A) 1 1 0.7589 0.7876 0.7933 0.6924 0.718 0.6862
(AMS)
Istanbul (IST) (A) 1 1 0.736 0.7715 0.8484 0.6356 0.6495 0.6296
London
(A) 1 1 0.6167 0.6879 0.7341 0.466 0.5092 0.4539
(LGW)
London
(A) 1 1 1 1 1 1 1 1
(LHR)
Paris (CDG) (A) 1 1 1 1 1 0.953 0.9789 0.9467
Paris (ORY) (A) 0.6551 0.8205 0.3541 0.4184 0.4734 0.2408 0.2676 0.2332
Rome (FCO) (A) 0.9983 1 0.6094 0.7208 0.7822 0.4224 0.4714 0.4088
Zurich (ZRH) (A) 1 1 1 1 1 1 1 1
Sydney (SYD) (A) 1 1 0.5793 0.6518 0.7774 0.4296 0.4572 0.4208
Mean 0.9615 0.9801 0.7394 0.7820 0.8232 0.6489 0.6724 0.6421
S.D 0.1149 0.0598 0.2264 0.1951 0.1693 0.2829 0.2712 0.2866
Barcelona
(B) 1 1 0.9088 1 1 0.7027 0.7633 0.6856
(BCN)
Frankfurt
(B) 1 0.9996 0.8574 0.9412 0.9509 0.7538 0.7842 0.7462
(FRA)
Madrid
(B) 1 1 0.8645 0.9761 1 0.6276 0.6871 0.6109
(MAD)
Munich
(B) 1 1 0.4965 0.5901 0.6798 0.3666 0.4095 0.3555
(MUC)
Bangkok
(B) 1 1 0.6436 0.6689 0.759 0.5756 0.5923 0.5709
(BKK)
Beijing (PEK) (B) 1 1 0.6579 0.746 0.766 0.542 0.5746 0.5343
Guangzhou
(B) 1 1 0.7168 0.762 0.8603 0.6048 0.6204 0.5986
(CAN)
Hong Kong
(B) 1 1 1 1 1 1 1 1
(HKG)
Incheon (ICN) (B) 0.9418 0.9944 0.8531 0.8548 0.8555 0.8488 0.85 0.8477
Kuala Lumpur
(B) 0.8975 0.8975 0.4379 0.463 0.5418 0.3723 0.3842 0.3685
(KUL)
Osaka (KIX) (B) 1 1 0.5923 0.6036 0.6138 0.5777 0.58 0.5766
Tokyo (HRT) (B) 1 1 0.6343 0.6417 0.665 0.6256 0.6274 0.6247
Shanghai
(B) 1 1 0.7907 0.8032 0.8046 0.7787 0.7822 0.7778
(PVG)
Singapore
(B) 0.8318 0.6672 0.2471 0.2916 0.3135 0.1905 0.2063 0.1865
(SIN)
Shenzhen
(B) 1 1 1 1 1 1 1 1
(SZX)
Mean 0.9781 0.9706 0.7134 0.7561 0.7873 0.6378 0.6574 0.6323
S.D. 0.0500 0.0880 0.2139 0.2149 0.2001 0.2246 0.2195 0.2263
188
Ch7
Table 7.5: The weight distribution of output variables: VSI

BCC model AHP/DEA-AR model
DMUs Wo1 Wo2 Wo3 Wo4 Wo1 Wo2 Wo3 Wo4
1 0.0498 0.2374 0.7129 0 0.0974 0.4810 0.4112 0.0103

2 1 0 0 0 0.1591 0.0820 0.7586 0.0004
3 0 0.1733 0.3918 0.4349 0.0902 0.5317 0.3671 0.0111
4 1 0 0 0 0.1236 0.4956 0.3806 0.0003
5 1 0 0 0 0.1991 0.0977 0.6876 0.0156
6 1 0 0 0 0.1253 0.7766 0.0970 0.0011
7 1 0 0 0 0.1637 0.1582 0.6777 0.0005
8 0 0 1 0 0.1263 0.1499 0.7081 0.0157
9 0 0 0.9883 0.0117 0.1029 0.5503 0.3463 0.0006
10 0.0409 0 0.3498 0.6093 0.1736 0.1385 0.6831 0.0048
11 0.0944 0 0.5061 0.3995 0.1667 0.0971 0.7351 0.0010
12 0.5510 0.3119 0.1371 0 0.1166 0.3054 0.5771 0.0009
13 1 0 0 0 0.1347 0.5539 0.3109 0.0004
14 1 0 0 0 0.1235 0.4511 0.4251 0.0003
15 0.5782 0 0.4218 0 0.1278 0.3916 0.4803 0.0003
16 1 0 0 0 0.0284 0.9523 0.0191 0.0001
17 0 0.3560 0 0.6440 0.0029 0.9836 0.0122 0.0013
18 1 0 0 0 0.1358 0.4988 0.3651 0.0003
19 0 0 0.0114 0.9886 0.0250 0.8708 0.0835 0.0207
20 0 0 0.5303 0.4697 0.0744 0.7150 0.1975 0.0131
21 0.4175 0.5825 0 0 0.0078 0.7594 0.2326 0.0002
22 0 0 0 1 0.1339 0.3633 0.4859 0.0169
23 1 0 0 0 0.0590 0.7348 0.2061 0.0001
24 1 0 0 0 0.1383 0.2952 0.5633 0.0032
189
Ch7
7.2.3 RELATIVE EFFICIENCY ANALYSIS IN DIFFERENT GROUPS

This section discusses the efficiency scores on the basis of VS I in two groups, from an
academic standpoint and a practical standpoint. Table 7.6 reveals that, from both
standpoints, relatively efficient airports are pointed out when conducting an AHP/DEA-
AR model based on alternative scales. A number of points emerge from Table 7.6,
including:
(1) From both standpoints, in the relatively inefficient airports (which are calculated
using different scales) it is easy to recognise the differences.
(2) The situation occurring in AHP analysis is not discovered in Table 6.27, which is
that when conducting alternative scales, the weights in scales 1 to 3 are close, and
weights in scales 4-6 are also close.
(3) The efficient airports are slightly different from academic viewpoints and practical
viewpoints. Among inefficient airports, only Singapore and Paris (ORY) have
significant lower efficiency scores.
(4) The efficiency scores and rankings of these two groups are similar to the results
shown in Table 7.6.
190
Table: 7.6: Relative efficiency scores obtained using the AHP/DEA-AR model by groups: VS I
Academia Practice
Amsterdam (AMS) 0.6204 0.6226 0.6342 0.6402 0.6335 0.6411 0.7552 0.7491 0.7926 0.6921 0.6879 0.6715
Barcelona (BCN) 0.4217 0.4218 0.4433 0.5056 0.4769 0.5107 0.8799 0.8733 1 0.7024 0.6944 0.6401
Frankfurt (FRA) 0.6780 0.6850 0.7144 0.6958 0.6892 0.6963 0.8470 0.8300 0.9474 0.7531 0.7477 0.7290
Istanbul (IST) 0.5981 0.5999 0.6138 0.6140 0.6076 0.6150 0.7282 0.7436 0.8600 0.6353 0.6466 0.6281
London (LGW) 0.3178 0.3297 0.3832 0.3506 0.3368 0.3521 0.5866 0.5712 0.7320 0.4656 0.4575 0.4225
London (LHR) 1 1 1 1 1 1 1 1 1 1 1 1
Madrid (MAD) 0.4336 0.4333 0.4463 0.4908 0.4705 0.4941 0.8331 0.8132 1 0.6276 0.6254 0.5766
Munich (MUC) 0.2376 0.2435 0.2699 0.2749 0.2623 0.2762 0.4940 0.4789 0.6788 0.3661 0.3602 0.3304
Paris (CDG) 0.8739 0.8727 0.8715 0.8985 0.8907 0.8997 1 1 1 0.9529 0.9489 0.9319
Paris (ORY) 0.1652 0.1711 0.1980 0.1856 0.1776 0.1863 0.3418 0.3341 0.4754 0.2406 0.2400 0.2188
Rome (FCO) 0.2667 0.2753 0.3164 0.3102 0.2942 0.3120 0.5936 0.5768 0.7806 0.4223 0.4187 0.3793
Zurich (ZRH) 1 1 1 1 1 1 1 1 1 1 1 1
Bangkok (BKK) 0.5354 0.5370 0.5464 0.5453 0.5412 0.5456 0.6314 0.6400 0.7643 0.5756 0.5749 0.5622
Beijing (PEK) 0.4622 0.4616 0.4650 0.4809 0.4740 0.4819 0.6373 0.6200 0.7660 0.5420 0.5353 0.5159
Guangzhou (CAN) 0.5510 0.5510 0.5570 0.5734 0.5653 0.5744 0.7116 0.7192 0.8664 0.6048 0.6147 0.5940
Hong Kong (HKG) 1 1 1 1 1 1 1 1 1 1 1 1
Incheon (ICN) 0.8478 0.8507 0.8555 0.8470 0.8476 0.8463 0.8500 0.8502 0.8577 0.8471 0.8460 0.8447
Kuala Lumpur
0.3385 0.3395 0.3483 0.3491 0.3448 0.3497 0.4311 0.4374 0.5479 0.3723 0.3749 0.3635
(KUL)
Osaka (KIX) 0.5749 0.5848 0.6136 0.5729 0.5730 0.5727 0.5868 0.5884 0.6295 0.5768 0.5775 0.5754
Tokyo (HRT) 0.6271 0.6344 0.6612 0.6235 0.6240 0.6229 0.6323 0.6309 0.6577 0.6250 0.6247 0.6235
Shanghai (PVG) 0.7633 0.7669 0.7656 0.7713 0.7701 0.7715 0.7860 0.7847 0.8046 0.7787 0.7771 0.7754
Singapore (SIN) 0.1545 0.1589 0.1770 0.1613 0.1584 0.1615 0.2379 0.2292 0.3122 0.1901 0.1870 0.1775
Shenzhen (SZX) 1 1 1 1 1 1 1 1 1 1 1 1
Sydney (SYD) 0.3577 0.3677 0.4218 0.3776 0.3688 0.3786 0.5666 0.5731 0.7967 0.4294 0.4365 0.4101
191
Ch7

This section describes how, after conducting the AHP/DEA-AR model and having
established the efficiency rankings of the sample airports, the hypothesis related to the
rankings obtained needs to be tested. However, unlike in Chapter Seven, in this section,
all of the results are used to test the hypothesis due to the varying outcomes for the
efficiency scores.
Table 7.7 shows that all the significance levels 𝑝 are not less than or equal to 0.05;
therefore, the result is not significant. In other words, there is no statistically significant
difference in airport efficiency between privately operated and publicly operated
airports as calculated using the AHP/DEA-AR model. The main objective of
privatisation is to improve efficiency and reduce government involvement in industry
(Humphreys, 1999); however, the results in this research are found to be quite different
from this main objective. In addition, the results also contrast with those of Barros and
Dieke (2007), who found a significantly higher difference in efficiency scores for
privately managed airports as compared to those that were publicly managed.
Table 7.7: Mann-Whitney test of differences in efficiency: Variable Set I

Asymptotic
Reference Scale Mann-Whitney Mann-Whitney
Significance
U-test Z-test
(two-tailed)
1-9 60 -0.449 0.653
Privately managed
𝑒 !!!
60.5 -0.421 0.682
airports 2 !!!
61 -0.392 0.726
vs. 9/9-9/1 66 -0.090 0.953
publicly managed 10/10-
airports 18/2
67 -0.030 1.000
∅
mapping
65 -0.149 0.907
7.3 AIRPORT EFFICIENCY ANALYSIS: VS II

In order to conduct sensitivity analysis, VS II (six variables) is undertaken in this
section in order to evaluate airport performance.
7.3.1 BOUNDS CALCULATION

The respondents’ derived weights attached to the input and output variables of
performance are shown in Table 7.8. The results derived from the AHP analysis serve
as a guideline for setting the upper and lower bounds in the AHP/DEA-AR model. This
ratio weight inequality constraint is incorporated in the AHP/DEA-AR model. Other
192
Ch7
ranges (or upper and lower bounds) of ratio weights can be found in Table 7.9. The
upper and lower bounds addressed by the AHP alternative scales are listed in Appendix
V.
Table 7.8: AHP weights of input and output variables of respondents (1-9 scale)
Amount of
Number of Number of Size of Number of Aircraft
Freight and
Employees Runways Terminal Passengers Movements
Mails
Respondent WI1 WI2 WI3 WO1 WO2 WO3
1 0.0909 0.4545 0.4545 0.4545 0.4545 0.0909

2 0.0667 0.4667 0.4667 0.5591 0.3522 0.0887
3 0.0667 0.4667 0.4667 0.4667 0.4667 0.0667
4 0.0574 0.3643 0.5783 0.4444 0.4444 0.1111
5 0.0608 0.5861 0.3531 0.4545 0.4545 0.0909
6 0.0811 0.3420 0.5769 0.4286 0.4286 0.1429
7 0.0887 0.5591 0.3522 0.6370 0.2583 0.1047
8 0.0667 0.4667 0.4667 0.4579 0.4161 0.1260
9 0.0811 0.3420 0.5769 0.5469 0.3445 0.1085
10 0.0887 0.3522 0.5591 0.6955 0.2290 0.0754
11 0.6337 0.1744 0.1919 0.2081 0.1311 0.6608
12 0.1564 0.7450 0.0986 0.6608 0.2081 0.1311
13 0.1085 0.5469 0.3445 0.6370 0.2583 0.1047
14 0.0633 0.7429 0.1939 0.4434 0.1692 0.3874
15 0.1429 0.4286 0.4286 0.6608 0.2081 0.1311
Average 0.1236 0.4692 0.4072 0.5170 0.3216 0.1614
16 0.2599 0.4126 0.3275 0.2402 0.5499 0.2098

17 0.0751 0.3575 0.5675 0.4444 0.4444 0.1111
18 0.0754 0.2290 0.6955 0.6586 0.2628 0.0786
19 0.0836 0.4443 0.4721 0.4545 0.4545 0.0909
20 0.0810 0.1884 0.7306 0.6267 0.0936 0.2797
21 0.0650 0.5736 0.3614 0.4161 0.4579 0.1260
22 0.2098 0.5499 0.2402 0.4286 0.1429 0.4286
Average 0.1214 0.3936 0.4850 0.4670 0.3437 0.1892

Average
0.1229 0.4452 0.4320 0.5011 0.3286 0.1703
in total
Table 7.9: Upper and lower bounds of variables weight ratios (scale 1-9)
Input weight ratio Upper Lower Output Weight Ratio Upper Lower
WI1/WI2 3.6343 0.0851 WO1/WO2 6.6974 0.4368
WI1/WI3 3.3019 0.0992 WO1/WO3 9.2220 0.3150
WI2/WI3 7.5596 0.2578 WO3/WO4 7.0000 0.1984
193
Ch7

The results of the DEA-BCC model, the AHP/DEA model, and the proposed
AHP/DEA-AR are summarised in Table 7.10. It can be seen there are eight evaluation
items for airport efficiency. The first column presents the efficiency scores, which are
calculated using the original DEA-BCC model. The second column shows the
efficiency scores, which are calculated using the AHP/DEA model. The weights for
each variable are considered using a 1-9 scale. The third column shows the efficiency
scores calculated using the DEA-AR model, and the weights of each variable are
considered using a 1-9 scale. Columns four to eight reveal the efficiency scores, which
are computed by the weights of each variable is considered in alternative scales.
An examination of VS II reveals that the number of efficient airports vary when using
different analysis models, even in the case of the AHP/DEA-AR model, and by means
of different AHP scales, the number of efficient airports vary as well (see Table 8.10).
(1) There are less efficient airports than inefficient airports (by means of the
AHP/DEA-AR model).
(2) The result shows that in the proposed AHP/DEA-AR model, different scales can
produce different numbers of efficient airports.
(3) The Standard Deviation (S.D.) of the efficiency score for the whole sample in the
DEA-BCC model is 0.1698, and in the AHP/DEA model is 0.1947. However, the
S.D. is from 0.2137, which is calculated by scale 21-9 to 0.2471 that is calculated
from scale ∅ mapping in the proposed AHP/DEA-AR model. Therefore, these
relative efficiency scores and S.D. indicate that the proposed AHP/DEA-AR model
possesses better discriminatory power than either the DEA-BCC model or the
AHP/DEA model.
(4) The results from the proposed model show that the mean of the efficiency scores
for the whole sample in the 1-9 scale is 0.7206 (see Table 7.10), whereas the
average efficiency scores for European airports is 0.7735, and for Asia-Pacific
airports is 0.6677. In general, the efficiency of European airports is better than that
of the Asia-Pacific region. Furthermore, it is worth noting that some Asian airports
are performing well; for example, Hong Kong and Shenzhen are performing better
than many European airports. However, Singapore Airport only achieved an
efficiency score of 0.2852 among these 24 airports.
194
Ch7
(5) Among those efficiency scores which were calculated using the AHP/DEA-AR
model, the value of standard deviation shows that a ø mapping scale can provide
more discriminatory power (S.D. is 0.2471).
(6) Table 7.11 reveals that when applying the AHP/DEA-AR model, the results show
that privately operated airports can get higher efficiency than those that are
publicly operated by conducting six different scales. When using the BCC model,
public airports can achieve better efficiency, and when using the AHP/DEA model,
private airports can achieve higher efficiency as compared to those that are public.
(7) Another benefit of using the AHP/DEA-AR model is to avoid extreme weight
distribution. The weight distribution of variables for the DEA model is shown in
Table 7.12. In regard to the discriminatory power and the distribution of variable
weights, the proposed AHP/DEA-AR model makes the DEA and AHP/DEA
results more accurate and also more able to reflect the actual decision-making
situation for airport performance evaluation.
195
Ch7
Table 7.10: Efficiency scores obtained using the AHP/DEA-AR model: VS II

AHP/DEA DEA-AR Model
DEA model
DMU BCC 1 2 3 4 5 6
Model !!! !!! 10/10- ∅
1-9 1-9 𝑒 2 9/9-9/1
18/2 mapping
European
Amsterdam
0.9010 0.8460 0.7854 0.7858 0.7900 0.6909 0.7162 0.6849
(AMS)
Barcelona (BCN) 1 1 0.9325 0.9450 1 0.6900 0.7460 0.6752
Frankfurt (FRA) 0.9996 0.9115 0.9362 0.9365 0.9367 0.7513 0.7813 0.7443
Istanbul (IST) 0.7046 0.6776 0.6183 0.6175 0.6279 0.6048 0.6066 0.6043
London (LGW) 1 1 0.6429 0.6564 0.7221 0.4589 0.4999 0.4480
London (LHR) 1 1 1 1 1 1 1 1
Madrid (MAD) 1 1 0.8588 0.8694 0.9752 0.6093 0.6604 0.5964
Munich (MUC) 0.9030 0.5153 0.5566 0.5646 0.6572 0.3633 0.4047 0.3526
Paris (CDG) 1 1 1 1 1 0.9523 0.9778 0.9459
Paris (ORY) 0.4810 0.4405 0.3399 0.3450 0.3936 0.2306 0.2530 0.2249
Rome (FCO) 0.7629 0.6711 0.6115 0.6206 0.7033 0.4080 0.4508 0.3971
Zurich (ZRH) 1 1 1 1 1 1 1 1
Mean 0.8960 0.8385 0.7735 0.7784 0.8172 0.6466 0.6747 0.6395
Asia
Bangkok (BKK) 0.8229 0.8212 0.6090 0.6147 0.6422 0.5699 0.5847 0.5662
Beijing (PEK) 1 1 0.7416 0.7463 0.7661 0.5421 0.5746 0.5344
Guangzhou
0.7039 0.7355 0.6089 0.6099 0.6246 0.5782 0.5841 0.5768
(CAN)
Hong Kong
1 1 1 1 1 1 1 1
(HKG)
Incheon (ICN) 0.9061 0.9061 0.8529 0.8534 0.8561 0.8483 0.8494 0.8474
Kuala Lumpur
0.6493 0.6137 0.3902 0.3914 0.3990 0.3643 0.3735 0.3620
(KUL)
Osaka (KIX) 1 1 0.5872 0.5866 0.5963 0.5765 0.5780 0.5763
Tokyo (HRT) 1 1 0.6317 0.6326 0.6370 0.6234 0.6248 0.6231
Shanghai (PVG) 0.8065 0.7656 0.8026 0.8034 0.8047 0.7789 0.7825 0.7781
Singapore (SIN) 0.4579 0.4294 0.2852 0.2884 0.3081 0.1891 0.2047 0.1854
Shenzhen (SZX) 1 1 1 1 1 1 1 1
Sydney (SYD) 1 1 0.5028 0.5060 0.5549 0.4103 0.4295 0.4055
Mean 0.8622 0.8560 0.6677 0.6694 0.6824 0.6234 0.6321 0.6213

Mean of all
0.8791 0.8472 0.7206 0.7239 0.7498 0.6350 0.6534 0.6304
samples
S.D. 0.1698 0.1947 0.2213 0.2206 0.2137 0.2452 0.2387 0.2471
196
Ch7
Table 7.11: Airport efficiency with different ownership: VS II

AHP/DEA-AR
Ownership AHP/DEA
DMU BCC 10/10- ∅
category* model 1-9 𝑒 !!! 2!!! 9/9-9/1
18/2 mapping
Amsterdam
(A) 0.901 0.846 0.7854 0.7858 0.79 0.6909 0.7162 0.6849
(AMS)
Istanbul (IST) (A) 0.7046 0.6776 0.6183 0.6175 0.6279 0.6048 0.6066 0.6043
London
(A) 1 1 0.6429 0.6564 0.7221 0.4589 0.4999 0.448
(LGW)
London
(A) 1 1 1 1 1 1 1 1
(LHR)
Paris (CDG) (A) 1 1 1 1 1 0.9523 0.9778 0.9459
Paris (ORY) (A) 0.481 0.4405 0.3399 0.345 0.3936 0.2306 0.253 0.2249
Rome (FCO) (A) 0.7629 0.6711 0.6115 0.6206 0.7033 0.408 0.4508 0.3971
Zurich (ZRH) (A) 1 1 1 1 1 1 1 1
Sydney (SYD) (A) 1 1 0.5028 0.506 0.5549 0.4103 0.4295 0.4055
Mean 0.8722 0.8484 0.7223 0.7257 0.7546 0.6395 0.6593 0.6345
S.D 0.1856 0.2067 0.2396 0.2372 0.2156 0.2889 0.2798 0.2913
Barcelona
(B) 1 1 0.9325 0.945 1 0.69 0.746 0.6752
(BCN)
Frankfurt
(B) 0.9996 0.9115 0.9362 0.9365 0.9367 0.7513 0.7813 0.7443
(FRA)
Madrid
(B) 1 1 0.8588 0.8694 0.9752 0.6093 0.6604 0.5964
(MAD)
Munich
(B) 0.903 0.5153 0.5566 0.5646 0.6572 0.3633 0.4047 0.3526
(MUC)
Bangkok
(B) 0.8229 0.8212 0.609 0.6147 0.6422 0.5699 0.5847 0.5662
(BKK)
Beijing (PEK) (B) 1 1 0.7416 0.7463 0.7661 0.5421 0.5746 0.5344
Guangzhou
(B) 0.7039 0.7355 0.6089 0.6099 0.6246 0.5782 0.5841 0.5768
(CAN)
Hong Kong
(B) 1 1 1 1 1 1 1 1
(HKG)
Incheon (ICN) (B) 0.9061 0.9061 0.8529 0.8534 0.8561 0.8483 0.8494 0.8474
Kuala Lumpur
(B) 0.6493 0.6137 0.3902 0.3914 0.399 0.3643 0.3735 0.362
(KUL)
Osaka (KIX) (B) 1 1 0.5872 0.5866 0.5963 0.5765 0.578 0.5763
Tokyo (HRT) (B) 1 1 0.6317 0.6326 0.637 0.6234 0.6248 0.6231
Shanghai
(B) 0.8065 0.7656 0.8026 0.8034 0.8047 0.7789 0.7825 0.7781
(PVG)
Singapore
(B) 0.4579 0.4294 0.2852 0.2884 0.3081 0.1891 0.2047 0.1854
(SIN)
Shenzhen
(B) 1 1 1 1 1 1 1 1
(SZX)
Mean 0.8833 0.8466 0.7196 0.7228 0.7469 0.6323 0.6499 0.6279
S.D. 0.1663 0.1946 0.2184 0.2186 0.2201 0.2260 0.2210 0.2275
197
Ch7
Table 7.12: The weight distribution of variables in the AHP/DEA-AR model: VS II

DEA-BCC model AHP/DEA model AHP/DEA-AR model
Wo1 Wo2 Wo3 Wo1 Wo2 Wo3 Wo1 Wo2 Wo3
0.3068 0 0 0 0.0120 0.0685 0.02777 0.39908 0.57315

0.4693 0 0.5307 0.0426 0 0 0.10495 0.05410 0.84095
0 0.2924 0 0 0.0143 0.0544 0.02626 0.45084 0.52289
0.1984 0 1.0215 0.0047 0.0501 0.0769 0.01208 0.74242 0.24550
0.4086 0.5914 0 0.0376 0 0 0.13713 0.06727 0.79560
0 0.8145 0.1855 0.0192 0 0 0.11991 0.78727 0.09282
0.3636 0.5807 0.0557 0.0253 0 0 0.11208 0.10832 0.77959
0 0.9949 0 0 0 0.0866 0.02187 0.10943 0.86870
0.1781 0.1227 0.6532 0 0.0216 0.0383 0.01374 0.58519 0.40107
0.4224 0.8466 0 0 0 0.1626 0.11891 0.09488 0.78621
0 0 0.8417 0 0 0.1079 0.11122 0.06481 0.82398
0.2336 0 0.7664 0.0468 0 0 0.13169 0.34507 0.52325
0.3052 0.7732 0 0.0274 0 0 0.13408 0.55133 0.31459
0.2155 0.6161 0 0.0230 0 0 0.09583 0.35000 0.55417
0.4992 0.4762 0.4977 0.0197 0.0524 0 0.02467 0.31868 0.65666
0.6591 0.3409 0 0.0270 0 0 0.01793 0.97001 0.01206
0.3183 0 0.2632 0 0.0426 0 0.00531 0.98247 0.01222
0.3543 0.8410 0 0.0016 0 0.1725 0.09656 0.34972 0.55372
0.4050 0 0.1256 0 0.1096 0.0282 0.00724 0.87623 0.11653
0.0161 0 0 0.0097 0.0371 0 0.06724 0.64609 0.28667
0 0 0 0 0.0260 0.0486 0.01124 0.65284 0.33592
0.3565 0.4350 0 0.0547 0.0073 0 0.10191 0.27662 0.62147
0 0 0.0691 0 0.1406 0.0369 0.01276 0.81644 0.17080
0 0 0 0.0392 0 0 0.02574 0.23163 0.74263
198

Ch8 Airport Efficiency Analysis II
7.3.3 RELATIVE EFFICIENCY ANALYSIS IN DIFFERENT GROUPS

This section discusses the efficiency scores on the basis of VS II in two groups, an
academic standpoint and a practical standpoint. Table 7.13 reveals that, from both
standpoints, relatively efficient airports are pointed out when conducting an AHP/DEA-
AR model based on alternative scales. A number of points emerge from Table 7.13,
including:
(1) From both standpoints, in the relatively inefficient airports (which are calculated
by different scales) it is easy to recognise the differences.
(2) The situation occurring in the AHP analysis is not found in Table 6.27, which is
that when conducting alternative scales, the weights in scales 1 to 3 are close, and
weights in scales 4-6 are also close.
(3) The efficient airports are slightly different from the academic viewpoint as
compared to the practical viewpoint. Among inefficient airports, only Singapore
and Paris (ORY) are found to have significantly lower efficiency.
(4) The efficiency scores and rankings of these two groups are similar to the results
shown in Table 7.10.
199
Table: 7.13: Relative efficiency scores obtained using the AHP/DEA-AR model by groups: VS II
Academia Practice
Amsterdam (AMS) 0.6169 0.6153 0.6124 0.6389 0.6318 0.6401 0.7854 0.7858 0.7900 0.6909 0.7162 0.6849
Barcelona (BCN) 0.4181 0.4120 0.4092 0.5059 0.4766 0.5111 0.9325 0.9343 0.5119 0.6900 0.7460 0.6752
Frankfurt (FRA) 0.6707 0.6693 0.6675 0.6931 0.6856 0.6942 0.9362 0.9365 0.9367 0.7513 0.7813 0.7443
Istanbul (IST) 0.5886 0.5871 0.5832 0.5971 0.5958 0.5977 0.6183 0.6027 0.6279 0.6048 0.6066 0.6043
London (LGW) 0.3068 0.3056 0.3148 0.3472 0.3322 0.3495 0.6429 0.6547 0.7221 0.4589 0.4999 0.4480
London (LHR) 1 1 1 1 1 1 1 1 1 1 1 1
Madrid (MAD) 0.4300 0.4266 0.4263 0.4832 0.4653 0.4863 0.8588 0.8496 0.9752 0.6093 0.6604 0.5964
Munich (MUC) 0.2295 0.2264 0.2209 0.2714 0.2578 0.2730 0.5566 0.5645 0.6572 0.3627 0.4045 0.3524
Paris (CDG) 0.8729 0.8710 0.8666 0.8979 0.8899 0.8989 1 1 1 0.9520 0.9777 0.9458
Paris (ORY) 0.1587 0.1576 0.1585 0.1791 0.1721 0.1802 0.3399 0.3427 0.3936 0.2306 0.2530 0.2249
Rome (FCO) 0.2562 0.2534 0.2527 0.3021 0.2868 0.3046 0.6115 0.6175 0.7033 0.4080 0.4508 0.3971
Zurich (ZRH) 1 1 1 1 1 1 1 1 1 1 1 1
Bangkok (BKK) 0.5336 0.5337 0.5367 0.5411 0.5381 0.5415 0.6090 0.6144 0.6422 0.5698 0.5847 0.5662
Beijing (PEK) 0.4618 0.4611 0.4636 0.4808 0.4739 0.4819 0.7416 0.7463 0.7661 0.5421 0.5746 0.5344
Guangzhou (CAN) 0.5485 0.5474 0.5449 0.5596 0.5562 0.5603 0.6089 0.6056 0.6246 0.5782 0.5841 0.5768
Hong Kong (HKG) 1 1 1 1 1 1 1 1 1 1 1 1
Incheon (ICN) 0.8482 0.8488 0.8489 0.8472 0.8476 0.8464 0.8524 0.8499 0.8557 0.8467 0.8481 0.8462
Kuala Lumpur
0.3366 0.3361 0.3364 0.3441 0.3415 0.3446 0.3902 0.3886 0.3990 0.3643 0.3735 0.3620
(KUL)
Osaka (KIX) 0.5698 0.5693 0.5684 0.5724 0.5718 0.5727 0.5872 0.5743 0.5963 0.5765 0.5780 0.5763
Tokyo (HRT) 0.6211 0.6212 0.6220 0.6211 0.6210 0.6211 0.6317 0.6326 0.6370 0.6234 0.6248 0.6231
Shanghai (PVG) 0.7672 0.7673 0.7660 0.7716 0.7705 0.7718 0.8026 0.8034 0.8047 0.7789 0.7825 0.7781
Singapore (SIN) 0.1504 0.1501 0.1510 0.1598 0.1564 0.1603 0.2852 0.2884 0.3081 0.1891 0.2047 0.1854
Shenzhen (SZX) 1 1 1 1 1 1 1 1 1 1 1 1
Sydney (SYD) 0.3436 0.3418 0.3399 0.3644 0.3576 0.3657 0.5028 0.4979 0.5549 0.4103 0.4295 0.4055
200
Ch7

This section describes, after conducting the AHP/DEA-AR model and having
established the efficiency rankings of the sample airports, a hypothesis related to the
rankings obtained needs to be tested. However, unlike Chapter 6, in this section, all of
the results are used to test the hypothesis due to the varying outcomes for the efficiency
scores.
Table 7.14 shows that all the significance levels 𝑝 are not less than or equal to 0.05;
therefore, the result is not significant. In other words, there is no statistically significant
difference in airport efficiency between privately operated and publicly operated
airports as calculated using the AHP/DEA-AR model with VS II.

Asymptotic
Reference Scale Mann-Whitney Mann-Whitney
Significance
U-test Z-test
(two-tailed)
1-9 63 −0.270 0.815
Privately managed 𝑒 !!!
63 −0.270 0.815
airports 2 !!!
65.5 −0.120 0.907
vs.
publicly managed 9/9-9/1 67 −0.030 1.000
airports 10/10-18/2 67 −0.030 1.000
∅ mapping 66 −0.090 0.953

Research Question 1 states that:
evaluation methods?
The results of airport efficiency, which are evaluated using different methods and in
different variables sets, are compared in this section.
Tables 7.15 and 7.16 show several evaluation items for airport efficiency in VS I and
VS II. The first column presents the ownership category for each airport. The second
column shows relative efficiency scores calculated using the original DEA-BCC model.
The third column shows the relative efficiency scores calculated using the integrated
AHP/DEA model (the weight of each variable is considered using a 1-9 scale).
201
Ch7
Columns four to eight reveal the relative efficiency scores computed using the
integrated AHP/DEA model (the weights of each variable are considered in alternative
judgement scales). On the right hand side, columns one to six reveal the relative
efficiency scores calculated using the AHP/DEA-AR model.
There is some evidence to support the supposition that the discriminatory power of the
AHP/DEA-AR model is better than that of either the AHP/DEA model or the
traditional DEA model. Firstly, when using AHP/DEA-AR model, the number of
inefficient airports are increased not only in Table 7.15 but also in Table 7.16.
Secondly, when using the DEA-BCC and AHP/DEA models, the efficiency scores of
publicly operated airports are higher than those of privately operated airports in both
variable sets. However, by conducting the AHP/DEA-AR model, the results reverse in
both variable sets. Consequently, Research Question 1 can be answered: the outcomes
of airport efficiency evaluation vary as a result of combining evaluation techniques.
Thirdly, in both models, Paris (ORY) and Singapore (SIN) are the least two inefficient
airports among the 24 sample airports. However, in Table 7.15 the scores do not help to
display these differences. However, in Table 7.16, the scores can help to represent how
far from others. Therefore, from this result, there is an indication of some evidence to
support that when using DEA analysis, fewer numbers of variables can help to increase
discriminatory power. However, this advantage cannot be addressed easily when only
using the AHP/DEA-AR model.
Fourthly, a comparison of the AHP/DEA model and the AHP/DEA-AR model, as

shown in the results in both tables, suggests that the latter model can help to raise the
discriminatory power of the results. Finally, the efficiency scores in Table 7.15 and
Table 7.16 explain that when using the AHP/DEA-AR model, a more robust result can
be provided even when using different numbers of variables. To sum up the above
discussion, the AHP/DEA-AR model can make the results more robust as compared to
the other two models (the BCC model and the AHP/DEA model).
202
Table: 7.15: Relative efficiency scores obtained using the AHP/DEA model and the AHP/DEA-AR model: VS I
AHP/DEA Model AHP/DEA-AR Model
DMU
category*
BCC 1-9 𝑒 !!! 2!!! 9/9-9/1 10/10-18/2 ∅ mapping 1-9 𝑒 !!! 2!!! 9/9-9/1
18/2 mapping
Amsterdam (AMS) (A) 1 1 1 1 1 1 1 0.7589 0.7876 0.7933 0.6924 0.718 0.6862
Istanbul (IST) (A) 1 1 1 1 1 1 1 0.7360 0.7715 0.8484 0.6356 0.6495 0.6296
London (LGW) (A) 1 1 1 1 1 1 1 0.6167 0.6879 0.7341 0.4660 0.5092 0.4539
London (LHR) (A) 1 1 1 1 1 1 1 1 1 1 1 1 1
Paris (CDG) (A) 1 1 1 1 1 1 1 1 1 1 0.9530 0.9789 0.9467
Paris (ORY) (A) 0.6551 0.8205 0.8205 0.8205 0.8205 0.8205 0.8205 0.3541 0.4184 0.4734 0.2408 0.2676 0.2332
Rome (FCO) (A) 0.9983 1 1 1 1 1 1 0.6094 0.7208 0.7822 0.4224 0.4714 0.4088
Zurich (ZRH) (A) 1 1 1 1 1 1 1 1 1 1 1 1 1
Sydney (SYD) (A) 1 1 1 1 1 1 1 0.5793 0.6518 0.7774 0.4296 0.4572 0.4208
Mean 0.9615 0.9801 0.9801 0.9801 0.9801 0.9801 0.9801 0.7394 0.7820 0.8232 0.6489 0.6724 0.6421
S.D 0.1149 0.0598 0.0598 0.0598 0.0598 0.0598 0.0598 0.2264 0.1951 0.1693 0.2829 0.2712 0.2866
Barcelona (BCN) (B) 1 1 1 1 1 1 1 0.9088 1 1 0.7027 0.7633 0.6856
Frankfurt (FRA) (B) 1 0.9996 0.9996 0.9996 0.9996 0.9996 0.9996 0.8574 0.9412 0.9509 0.7538 0.7842 0.7462
Madrid (MAD) (B) 1 1 1 1 1 1 1 0.8645 0.9761 1 0.6276 0.6871 0.6109
Munich (MUC) (B) 1 1 1 1 1 1 1 0.4965 0.5901 0.6798 0.3666 0.4095 0.3555
Bangkok (BKK) (B) 1 1 1 1 1 1 1 0.6436 0.6689 0.7590 0.5756 0.5923 0.5709
Beijing (PEK) (B) 1 1 1 1 1 1 1 0.6579 0.746 0.7660 0.542 0.5746 0.5343
Guangzhou (CAN) (B) 1 1 1 1 1 1 1 0.7168 0.762 0.8603 0.6048 0.6204 0.5986
Hong Kong (HKG) (B) 1 1 1 1 1 1 1 1 1 1 1 1 1
Incheon (ICN) (B) 0.9418 0.9944 0.9944 0.9944 0.9944 0.9944 0.9944 0.8531 0.8548 0.8555 0.8488 0.85 0.8477
Kuala Lumpur (KUL) (B) 0.8975 0.8975 0.8975 0.8975 0.8975 0.8975 0.8975 0.4379 0.463 0.5418 0.3723 0.3842 0.3685
Osaka (KIX) (B) 1 1 1 1 1 1 1 0.5923 0.6036 0.6138 0.5777 0.58 0.5766
Tokyo (HRT) (B) 1 1 1 1 1 1 1 0.6343 0.6417 0.665 0.6256 0.6274 0.6247
Shanghai (PVG) (B) 1 1 1 1 1 1 1 0.7907 0.8032 0.8046 0.7787 0.7822 0.7778
Singapore (SIN) (B) 0.8318 0.6672 0.6672 0.6672 0.6672 0.6672 0.6672 0.2471 0.2916 0.3135 0.1905 0.2063 0.1865
Shenzhen (SZX) (B) 1 1 1 1 1 1 1 1 1 1 1 1 1
Mean 0.9781 0.9706 0.9706 0.9706 0.9706 0.9706 0.9706 0.7134 0.7561 0.7873 0.6378 0.6574 0.6323
S.D. 0.0500 0.0880 0.0880 0.0880 0.0880 0.0880 0.0880 0.2139 0.2149 0.2001 0.2246 0.2195 0.2263
203
Table: 7.16: Relative efficiency scores obtained using the AHP/DEA model and the AHP/DEA-AR model: VS II
AHP/DEA Model AHP/DEA-AR Model
Ownership !!! 10/10- ∅
DMU
category*
BCC 1-9 𝑒 !!!
2!!!
9/9-9/1 10/10-18/2 ∅ mapping 1-9 𝑒 2!!! 9/9-9/1
18/2 mapping
Amsterdam (AMS) (A) 0.9010 0.8460 0.8460 0.8460 0.8460 0.8460 0.8460 0.7854 0.7858 0.7900 0.6909 0.7162 0.6849
Istanbul (IST) (A) 0.7046 0.6776 0.6773 0.6773 0.6773 0.6773 0.6773 0.6183 0.6175 0.6279 0.6048 0.6066 0.6043
London (LGW) (A) 1 1 1 1 1 1 1 0.6429 0.6564 0.7221 0.4589 0.4999 0.448
London (LHR) (A) 1 1 1 1 1 1 1 1 1 1 1 1 1
Paris (CDG) (A) 1 1 1 1 1 1 1 1 1 1 0.9523 0.9778 0.9459
Paris (ORY) (A) 0.4810 0.4405 0.4405 0.4405 0.4405 0.4405 0.4405 0.3399 0.345 0.3936 0.2306 0.253 0.2249
Rome (FCO) (A) 0.7629 0.6711 0.6711 0.6711 0.6711 0.6711 0.6711 0.6115 0.6206 0.7033 0.408 0.4508 0.3971
Zurich (ZRH) (A) 1 1 1 1 1 1 1 1 1 1 1 1 1
Sydney (SYD) (A) 1 1 1 1 1 1 1 0.5028 0.506 0.5549 0.4103 0.4295 0.4055
Mean 0.8722 0.8484 0.8483 0.8483 0.8483 0.8483 0.8483 0.7223 0.7257 0.7546 0.6395 0.6593 0.6345
S.D 0.1856 0.2067 0.2068 0.2068 0.2068 0.2068 0.2068 0.2396 0.2372 0.2156 0.2889 0.2798 0.2913
Barcelona (BCN) (B) 1 1 1 1 1 1 1 0.9325 0.945 1 0.69 0.746 0.6752
Frankfurt (FRA) (B) 0.9996 0.9115 0.9115 0.9115 0.9115 0.9115 0.9115 0.9362 0.9365 0.9367 0.7513 0.7813 0.7443
Madrid (MAD) (B) 1 1 1 1 1 1 1 0.8588 0.8694 0.9752 0.6093 0.6604 0.5964
Munich (MUC) (B) 0.9031 0.5153 0.5153 0.5153 0.5153 0.5153 0.5153 0.5566 0.5646 0.6572 0.3633 0.4047 0.3526
Bangkok (BKK) (B) 0.8229 0.8212 0.8212 0.8212 0.8212 0.8212 0.8212 0.609 0.6147 0.6422 0.5699 0.5847 0.5662
Beijing (PEK) (B) 1 1 1 1 1 1 1 0.7416 0.7463 0.7661 0.5421 0.5746 0.5344
Guangzhou (CAN) (B) 0.7039 0.7355 0.7355 0.7355 0.7355 0.7355 0.7355 0.6089 0.6099 0.6246 0.5782 0.5841 0.5768
Hong Kong (HKG) (B) 1 1 1 1 1 1 1 1 1 1 1 1 1
Incheon (ICN) (B) 0.9061 0.9061 0.9061 0.9061 0.9061 0.9061 0.9061 0.8529 0.8534 0.8561 0.8483 0.8494 0.8474
Kuala Lumpur (KUL) (B) 0.6493 0.6137 0.6137 0.6137 0.6137 0.6137 0.6137 0.3902 0.3914 0.399 0.3643 0.3735 0.362
Osaka (KIX) (B) 1 1 1 1 1 1 1 0.5872 0.5866 0.5963 0.5765 0.578 0.5763
Tokyo (HRT) (B) 1 1 1 1 1 1 1 0.6317 0.6326 0.637 0.6234 0.6248 0.6231
Shanghai (PVG) (B) 0.8065 0.7656 0.7656 0.7656 0.7656 0.7656 0.7656 0.8026 0.8034 0.8047 0.7789 0.7825 0.7781
Singapore (SIN) (B) 0.4579 0.4294 0.4294 0.4294 0.4294 0.4294 0.4294 0.2852 0.2884 0.3081 0.1891 0.2047 0.1854
Shenzhen (SZX) (B) 1 1 1 1 1 1 1 1 1 1 1 1 1
Mean 0.8833 0.8466 0.8466 0.8466 0.8466 0.8466 0.8466 0.7196 0.7228 0.7469 0.6323 0.6499 0.6279
S.D. 0.1663 0.1946 0.1946 0.1946 0.1946 0.1946 0.1946 0.2184 0.2186 0.2201 0.2260 0.2210 0.2275
204
Ch7
7.5 ALTERNATIVE THOUGHTS ON AIRPORT EFFICIENCY ANALYSIS

7.5.1 STRAIGHTFORWARD AHP APPROACH
In this research, AHP is used to acquire the weights for each variable; another thought
on how to apply AHP is offered in this section, which is to calculate efficiency score
from AHP weights, and it is called the straightforward AHP approach. The first step of
this approach is to standardise each variable and then to multiply the value with the
weights of each variable. The final step is to add the value of all input and output
variables individually and then to apply efficiency equation to show the final efficiency
scores. The left side of Table 7.17 shows the raw data for VS I. The right side of Table
7.17 shows the value after standardisation. The amount in Table 7.18 reveals the result
which is multiplied with the AHP weight and efficiency equation. The first column of
this table is calculated from Table 7.17. The ratio column is calculated from the
efficiency equation, which is the sum of actual outputs/ sum of actual inputs. Then, the
final column is the efficiency computed using the straightforward AHP approach.
Take AMS airport as an example. Among these 24 sample airports, the maximum value
in the number of employees is 17,996, which is selected from Frankfurt Airport (see
Table 7.17). The value 14.33 is calculated from 2,579 17,996 ×100 = 14.33. The
result from Table 7.18 in the first column is 0.33 = 14.33×0.0228.
The amount in ratio column =1.08
= 0.33 + 4.18 + 8.64 + 4.05 + 4.34 + 33.23) (15.75 + 6.16 + 4.55 + 32.51
The efficiency score in the final column is 0.6273= 1.08 1.72 .
The results for other columns all follow this calculation procedure.
Table 7.19 shows the efficiency score comparison calculated by means of different
approaches. It reveals there to be eight scores for airport efficiency. The first column
presents the efficiency scores, which are calculated using the basic DEA-CCR BCC
model. The third column shows the efficiency scores calculated using the AHP/DEA-
AR model. The forth column shows the efficiency scores calculated using the
straightforward AHP approach. Columns five to eight reveal the efficiency scores,
which are computed by the weights of each variable considered in VS II.
205
Table: 7.17: Progress on AHP approach: VSI (1)
Input Output Input Output
1 2 3 4 5 6 1 2 3 4 1 2 3 4 5 6 1 2 3 4
Max 17996 264 6 1382000 4090 1851.29 67056 3400000 559812 2891.66
Amsterdam
2579 94 6 591885 3244 985.64 47430 1567712 446693 1602.42 14.33 35.61 100 42.83 79.32 53.24 70.73 46.11 79.79 55.42
(AMS)
Barcelona
569 101 3 155200 2850 290.36 30208 104239 321491 324.03 3.16 38.26 50.00 11.23 69.68 15.68 45.05 3.07 57.43 11.21
(BCN)
Frankfurt
17996 174 3 800000 4000 1851.29 53467 2111116 485783 2104.52 100 65.91 50.00 57.89 97.80 100.00 79.73 62.09 86.78 72.78
(FRA)
Istanbul (IST) 1750 32 3 318500 2767 172.04 28533 766221 254531 283.98 9.72 12.12 50.00 23.05 67.65 9.29 42.55 22.54 45.47 9.82
London (LGW) 2186 107 1 202519 3316 584.43 34214 112366 263653 854.14 12.15 40.53 16.67 14.65 81.08 31.57 51.02 3.30 47.10 29.54
London (LHR) 5516 264 2 364800 3780 1820.23 67056 1486262 478693 2891.66 30.65 100 33.33 26.40 92.42 98.32 100.00 43.71 85.51 100
Madrid (MAD) 797 76 4 300000 3863 580.71 50846 328985 469740 647.92 4.43 28.79 66.67 21.71 94.45 31.37 75.83 9.68 83.91 22.41
Munich (MUC) 7400 200 2 458000 4000 1063.23 34552 274464 432296 1368.4 41.12 75.76 33.33 33.14 97.80 57.43 51.53 8.07 77.22 47.32
Paris (CDG) 3858 124 4 542595 3454 1084.6 60875 2040000 559812 1768.92 21.44 46.97 66.67 39.26 84.45 58.59 90.78 60.00 100 61.17
Paris (ORY) 3304 102 3 371500 3123 497.65 26210 140000 230167 732.32 18.36 38.64 50.00 26.88 76.36 26.88 39.09 4.12 41.12 25.33
Rome (FCO) 3278 86 4 285000 3677 477.66 35227 137424 346654 948.93 18.22 32.58 66.67 20.62 89.90 25.80 52.53 4.04 61.92 32.82
Zurich (ZRH) 1254 67 3 138614 3167 401.63 22099 387671 274991 789.76 6.97 25.38 50.00 10.03 77.43 21.69 32.96 11.40 49.12 27.31
Bangkok
3245 120 2 563000 3850 432.04 46932 1291931 311435 733.69 18.03 45.45 33.33 40.74 94.13 23.34 69.99 38.00 55.63 25.37
(BKK)
Beijing (PEK) 1965 120 3 1382000 3600 517.26 55938 1367710 429646 561.6 10.92 45.45 50.00 100 88.02 27.94 83.42 40.23 76.75 19.42
Guangzhou
3482 74 2 320000 3700 235.65 33435 685868 280392 378.69 19.35 28.03 33.33 23.15 90.46 12.73 49.86 20.17 50.09 13.10
(CAN)
Hong Kong
1131 106 2 710000 3800 425.23 47700 3400000 296000 1120.93 6.28 40.15 33.33 51.37 92.91 22.97 71.13 100.00 52.87 38.76
(HKG)
Incheon (ICN) 933 90 3 600000 3833 396.64 29973 2423717 211102 973.92 5.18 34.09 50.00 43.42 93.72 21.43 44.70 71.29 37.71 33.68
Kuala Lumpur
1578 106 2 479404 4090 256.8 27529 667495 209681 292.24 8.77 40.15 33.33 34.69 100.00 13.87 41.05 19.63 37.46 10.11
(KUL)
Osaka (KIX) 388 52 2 330000 3750 458.45 16014 846522 133502 959.15 2.16 19.70 33.33 23.88 91.69 24.76 23.88 24.90 23.85 33.17
Tokyo (HRT) 720 87 2 789700 3250 1109.51 32654 2100448 193321 1830.69 4.00 32.95 33.33 57.14 79.46 59.93 48.70 61.78 34.53 63.31
Shanghai
6440 98 3 824000 3733 212.15 28236 2603027 265735 482.15 35.79 37.12 50.00 59.62 91.27 11.46 42.11 76.56 47.47 16.67
(PVG)
Singapore
1396 102 3 1043020 3583 402.63 22877 415726 185304 921.99 7.76 38.64 50.00 75.47 87.60 21.75 34.12 12.23 33.10 31.88
(SIN)
Shenzhen
3998 55 1 152000 3400 95.52 21401 598036 187942 217.22 22.22 20.83 16.67 11.00 83.13 5.16 31.92 17.59 33.57 7.51
(SZX)
Sydney (SYD) 306 65 3 387487 2978 136.47 32900 470000 298964 773.69 1.70 24.62 50.00 28.04 72.81 7.37 49.06 13.82 53.40 26.76
206
Ch7
Table 7.18: Progress on AHP approach: VSI (2)

Input Output Ratio Final
1 2 3 4 5 6 1 2 3 4
weight 0.0228 0.1173 0.0864 0.0946 0.0547 0.6241 0.2227 0.1335 0.0570 0.5867 1.72
Amsterdam
(AMS) 0.33 4.18 8.64 4.05 4.34 33.23 15.75 6.16 4.55 32.51 1.08 0.6273
Barcelona
(BCN) 0.07 4.49 4.32 1.06 3.81 9.79 10.03 0.41 3.27 6.57 0.86 0.5021
Frankfurt
(FRA) 2.28 7.73 4.32 5.48 5.35 62.41 17.76 8.29 4.95 42.70 0.84 0.4902
Istanbul
(IST) 0.22 1.42 4.32 2.18 3.70 5.80 9.48 3.01 2.59 5.76 1.18 0.6880
London
(LGW) 0.28 4.75 1.44 1.39 4.43 19.70 11.36 0.44 2.68 17.33 0.99 0.5794
London
(LHR) 0.70 11.73 2.88 2.50 5.06 61.36 22.27 5.84 4.87 58.67 1.09 0.6339
Madrid
(MAD) 0.10 3.38 5.76 2.05 5.17 19.58 16.89 1.29 4.78 13.15 1.00 0.5837
Munich
(MUC) 0.94 8.89 2.88 3.14 5.35 35.84 11.48 1.08 4.40 27.76 0.78 0.4568
Paris (CDG) 0.49 5.51 5.76 3.71 4.62 36.56 20.22 8.01 5.70 35.89 1.23 0.7179
Paris
(ORY) 0.42 4.53 4.32 2.54 4.18 16.78 8.70 0.55 2.34 14.86 0.81 0.4704
Rome
(FCO) 0.42 3.82 5.76 1.95 4.92 16.10 11.70 0.54 3.53 19.25 1.06 0.6188
Zurich
(ZRH) 0.16 2.98 4.32 0.95 4.24 13.54 7.34 1.52 2.80 16.02 1.06 0.6161
Bangkok
(BKK) 0.41 5.33 2.88 3.85 5.15 14.56 15.59 5.07 3.17 14.89 1.20 0.7007
Beijing
(PEK) 0.25 5.33 4.32 9.46 4.81 17.44 18.58 5.37 4.37 11.39 0.95 0.5560
Guangzhou
(CAN) 0.44 3.29 2.88 2.19 4.95 7.94 11.10 2.69 2.85 7.68 1.12 0.6535
Hong Kong
(HKG) 0.14 4.71 2.88 4.86 5.08 14.34 15.84 13.35 3.01 22.74 1.72 1
Incheon
(ICN) 0.12 4.00 4.32 4.11 5.13 13.37 9.95 9.52 2.15 19.76 1.33 0.7766
Kuala
Lumpur 0.20 4.71 2.88 3.28 5.47 8.66 9.14 2.62 2.13 5.93 0.79 0.4584
(KUL)
Osaka
(KIX) 0.05 2.31 2.88 2.26 5.02 15.46 5.32 3.32 1.36 19.46 1.05 0.6137
Tokyo
(HRT) 0.09 3.87 2.88 5.41 4.35 37.40 10.84 8.25 1.97 37.14 1.08 0.6280
Shanghai
(PVG) 0.82 4.35 4.32 5.64 4.99 7.15 9.38 10.22 2.71 9.78 1.18 0.6853
Singapore
(SIN) 0.18 4.53 4.32 7.14 4.79 13.57 7.60 1.63 1.89 18.71 0.86 0.5031
Shenzhen
(SZX) 0.51 2.44 1.44 1.04 4.55 3.22 7.11 2.35 1.91 4.41 1.20 0.6964
Sydney
(SYD) 0.04 2.89 4.32 2.65 3.98 4.60 10.93 1.85 3.04 15.70 1.71 0.9933
207
Ch7
Table 7.19: Efficiency scores using different approaches

VSI VSII
DEA/ DEA/
Airport CCR BCC AHP CCR BCC AHP
AHP-AR AHP-AR
Amsterdam (AMS) 1 1 0.7589 0.6273 0.712 0.901 0.7854 0.3720
Barcelona (BCN) 1 1 0.9088 0.5021 1 1 0.9325 0.4564
Frankfurt (FRA) 0.9899 1 0.8574 0.4902 0.7384 0.9996 0.9362 0.4799
Istanbul (IST) 1 1 0.7360 0.6880 0.693 0.7046 0.6183 0.4104
London (LGW) 1 1 0.6167 0.5794 1 1 0.6429 0.8563
London (LHR) 1 1 1 0.6339 1 1 1 1
Madrid (MAD) 1 1 0.8645 0.5837 1 1 0.8588 0.5267
Munich (MUC) 1 1 0.4965 0.4568 0.8263 0.903 0.5566 0.4541
Paris (CDG) 1 1 1 0.7179 0.9725 1 1 0.6261
Paris (ORY) 0.6500 0.6551 0.3541 0.4704 0.4305 0.481 0.3399 0.2910
Rome (FCO) 0.9711 0.9948 0.6094 0.6188 0.6803 0.7629 0.6115 0.3535
Zurich (ZRH) 1 1 1 0.6161 1 1 1 0.3924
Bangkok (BKK) 1 1 0.6436 0.7007 0.7817 0.8229 0.6090 0.6145
Beijing (PEK) 1 1 0.6579 0.5560 0.9276 1 0.7416 0.3731
Guangzhou (CAN) 0.9983 1 0.7168 0.6535 0.7036 0.7039 0.6089 0.5556
Hong Kong (HKG) 1 1 1 1 1 1 1 0.7569
Incheon (ICN) 0.9391 0.9418 0.8531 0.7766 0.8642 0.9061 0.8529 0.4670
Kuala Lumpur
0.789 0.8844 0.4379 0.4584 0.6447 0.6493 0.3902 0.4016
(KUL)
Osaka (KIX) 1 1 0.5923 0.6137 0.8151 1 0.5872 0.3545
Tokyo (HRT) 1 1 0.6343 0.6280 0.9956 1 0.6317 0.4654
Shanghai (PVG) 1 1 0.7907 0.6853 0.6671 0.8065 0.8026 0.3863
Singapore (SIN) 0.7878 0.8318 0.2471 0.5031 0.4423 0.4579 0.2852 0.1760
Shenzhen (SZX) 1 1 1 0.6964 0.9609 1 1 0.7059
Sydney (SYD) 1 1 0.5793 0.9933 1 1 0.5028 0.4117
S.D. 0.0896 0.0791 0.2141 0.1417 0.1782 0.1698 0.2213 0.1863
An examination of the straightforward AHP approach with the AHP/DEA-AR model

reveals that the number of efficient airports is similar, even in the AHP/DEA-AR model.
Some other findings are as follows:
(1) The results from the straightforward AHP approach have only one airport is
efficient for each variable set. In VSI is Hong Kong airport and in VSII is London
Heathrow. The results are similar to other models. However, the numbers of
efficient airport are much less than other approaches.
(2) On the opposite, the most inefficient airport calculated using the AHP weight
approach in VSI is Munich, and in VSII is Singapore. The result in VSI is very
different from the others, but the result in VSII is the same.
208
Ch7
(3) The result shows that the proposed straightforward AHP approach can produce
different numbers of efficient airports.
(4) The Standard Deviation (S.D.) of efficiency score for the whole sample, in the
AHP weight approach decreased from 0.2141 to 0.1417 in VSI and decreased from
0.2213 to 0.1863 in VSII whereas the S.D. is decreased as compared to the
AHP/DEA-AR model. However, the S.D. in the AHP weight approach was still
higher than that of the basic DEA BCC and CCR models. Therefore, these relative
efficiency scores and the S.D. indicate that the proposed AHP weight approach
possesses better discriminatory power than either the DEA-BCC model or the
DEA-CCR model.
(5) Table 7.19 indicates that the straightforward AHP approach can also provide
higher discriminatory power results. It will be a very interesting approach for
future research.
7.5.2 REFERRAL CLUSTER APPROACH

In this section, another airport efficiency calculation approach is discussed. In the basic
DEA model, the outcome from the clustering analysis referral (details presented in
Section 7.1.3) can help determine the relatively efficient airports, which are then used
as the references and frequencies that improve the efficiencies of relatively inefficient
enterprises (Zhu, 2009). The more an airport is referred, the more relatively efficient
the airport is in the DMU group. Therefore, this approach can also provide some
outputs which can generate efficiency scores. In addition, the clustering analysis
referral can also be conducted when applying the AHP/DEA-AR model. Therefore, in
this section, the results from these two models are discussed. Table 7.20 shows the
clustering analysis referrals and efficiency from the basic DEA-BCC model and the
AHP/DEA-AR model in two variable sets.
In Table 7.20, the first column presents the efficiency scores, which are calculated
using the basic DEA-BCC model. The second column shows the referral frequency.
The third column shows the rank reference, which is calculated from the referral
frequency. For Sydney Airport, the referral frequency is the highest (5 in this model);
subsequently we can give it the amount 24. London Heathrow Airport is the second
highest among these 24 airports, hence we can assign 23 to it. Other airports can be
followed in this order. Columns four to six show the results calculated using the
209
Ch7
AHP/DEA-AR model. The right side of the table shows the results that are considered
in VS II. In this table, relatively efficient airports can get higher scores. Therefore, in
VSI, when calculating the efficiency scores using the BCC model, Sydney Airport is
the most efficient airport, followed by London Heathrow. By means of the AHP/DEA-
AR model, Paris (CDG) and Hong Kong are the most efficient airports. If comparison
is made with Table 7.19, the results are similar. The comparison in VSII also shows a
similar situation.
Table 7.20: The clustering analysis referrals and efficiency scores

VSI VSII
BCC AHP/DEA-AR BCC AHP/DEA-AR
Efficiency Referral Rank Efficiency Referral Rank Efficiency Referral Rank Efficiency Referral Rank
DMU
score Frequency reference score Frequency reference score Frequency reference score Frequency reference
Amsterdam
(AMS)
1 1 15.5 0.7589 0 14 0.9010 0 9 0.7854 0 14
Barcelona
(BCN)
1 1 15.5 0.9088 0 19 1 3 19.5 0.9325 0 18
Frankfurt
(FRA)
1 0 9.5 0.8574 0 17 0.9996 0 12 0.9362 0 19
Istanbul
(IST)
1 2 20 0.7360 0 13 0.7046 0 5 0.6183 0 10
London
(LGW)
1 0 9.5 0.6167 0 8 1 3 19.5 0.6429 0 12
London
(LHR)
1 3 23 1 10 21.5 1 6 22 1 13 24
Madrid
(MAD)
1 1 15.5 0.8645 0 18 1 7 23 0.8588 0 17
Munich
(MUC)
1 0 9.5 0.4965 0 4 0.9030 0 10 0.5566 0 5
Paris
(CDG)
1 2 20 1 12 23.5 1 5 21 1 10 23
Paris
(ORY)
0.6551 0 1 0.3541 0 2 0.4810 0 2 0.3399 0 3
Rome
(FCO)
0.9948 0 5 0.6094 0 7 0.7629 0 6 0.6115 0 9
Zurich
(ZRH)
1 2 20 1 10 21.5 1 1 17.5 1 8 21.5
Bangkok
(BKK)
0.9937 0 9.5 0.6436 0 10 0.8229 0 8 0.6090 0 8
Beijing
(PEK)
1 0 9.5 0.6579 0 11 1 2 18 0.7416 0 13
Guangzhou
(CAN)
1 0 9.5 0.7168 0 12 0.7039 0 4 0.6089 0 7
Hong
Kong 1 2 20 1 12 23.5 1 9 24 1 8 21.5
(HKG)
Incheon
(ICN)
0.9418 0 4 0.8531 0 16 0.9061 0 11 0.8529 0 16
Kuala
Lumpur 0.8845 0 3 0.4379 0 3 0.6493 0 3 0.3902 0 4
(KUL)
Osaka
(KIX)
1 2 20 0.5923 0 6 1 1 17.5 0.5872 0 6
Tokyo
(HRT)
1 0 9.5 0.6343 0 9 1 0 14 0.6317 0 11
Shanghai
(PVG)
1 0 9.5 0.7907 0 15 0.8065 0 7 0.8026 0 15
Singapore
(SIN)
0.8318 0 2 0.2471 0 1 0.4579 0 1 0.2852 0 2
Shenzhen
(SZX)
1 1 15.5 1 3 20 1 0 14 1 4 20
Sydney
(SYD)
1 5 24 0.5793 0 5 1 0 14 0.5028 0 1
210
Ch7
7.5.3 THE COMPARISON BETWEEN MODELS

In this research, including the previous sections, six different approaches are applied to
evaluate airport efficiency with two variable sets (i.e. the basic DEA-CCR and BCC
models, the integrated AHP/DEA model, the AHP/DEA-AR model, the straightforward
AHP approach, and the referral cluster approach in the BCC and AHP/DEA-AR
models). In this section, a correlation coefficient analysis is adopted to show which
models or approaches can provide the most reliable results. Table 7.21 shows all of the
efficiency scores acquired in this research. As mentioned in Section 6.12, in this
research, only the BCC model is discussed in the following section because the
assumption of the BCC model is closer to an actual situation. In addition, when
conducting AHP/DEA-AR analysis, the clustering analysis referral can also be
undertaken. Therefore, the efficiency scores for the referral cluster approach calculated
using the AHP/DEA-AR model are also included. The outcome of correlation
coefficient analysis is displayed in Table 7.22. The reason for applying a correlation
coefficient analysis is to help the author choose proper analysis models, which means
that if two approaches have higher relationships, only one of them can be replaced by
another one. Table 7.22 reveals that there are few of the results higher than 0.8 that
implies when adopting these two approaches can provide similar results. In this case,
the author can only choose one of them to evaluate airport efficiency. In variable set I,
one of approaches between the AHP/DEA-AR and the AHP/DEA-AR rank reference
can be reserved. In variable set II, AHP/DEA can replace both DEA-BCC and BCC
referral frequency. One of approaches between the AHP/DEA-AR and AHP/DEA-AR
rank references can be reserved.
211
Table 7.21: The efficiency scores using different models and approaches (1)
VSI VSII
AHP/DEA-AR AHP/DEA-AR
BCC Rank BCC Rank AHP/DEA-AR AHP/DEA-AR
DMU BBCI AHP/DEAI Rank AHPI BBCII AHP/DEA II AHP II
reference I I reference/AR I
reference II II Rank reference II
Amsterdam 1 15.5 1 0.7589 14 0.6273 0.9010 9 0.8460 0.7854 14 0.3720

Barcelona 1 15.5 1 0.9088 19 0.5021 1 19.5 1 0.9325 18 0.4564
Frankfurt 1 9.5 1 0.8574 17 0.4902 0.9996 12 0.9115 0.9362 19 0.4799
Istanbul 1 20 1 0.7360 13 0.6880 0.7046 5 0.6776 0.6183 10 0.4104
London (LGW) 1 9.5 1 0.6167 8 0.5794 1 19.5 1 0.6429 12 0.8563
London (LHR) 1 23 0.8205 1 21.5 0.6339 1 22 1 1 24 1
Madrid 1 15.5 1 0.8645 18 0.5837 1 23 1 0.8588 17 0.5267
Munich 1 9.5 1 0.4965 4 0.4568 0.9030 10 0.5153 0.5566 5 0.4541
Paris (CDG) 1 20 1 1 23.5 0.7179 1 21 1 1 23 0.6261
Paris (ORY 0.6551 1 1 0.3541 2 0.4704 0.4810 2 0.4405 0.3399 3 0.2910
Rome 0.9948 5 1 0.6094 7 0.6188 0.7629 6 0.6711 0.6115 9 0.3535
Zurich 1 20 1 1 21.5 0.6161 1 17.5 1 1 21.5 0.3924
Bangkok 0.9937 9.5 1 0.6436 10 0.7007 0.8229 8 0.8212 0.6090 8 0.6145
Beijing 1 9.5 0.9996 0.6579 11 0.5560 1 18 1 0.7416 13 0.3731
Guangzhou 1 9.5 1 0.7168 12 0.6535 0.7039 4 0.7355 0.6089 7 0.5556
Hong Kong 1 20 1 1 23.5 1 1 24 1 1 21.5 0.7569
Incheon 0.9418 4 1 0.8531 16 0.7766 0.9061 11 0.9061 0.8529 16 0.4670
Kuala Lumpur 0.8845 3 1 0.4379 3 0.4584 0.6493 3 0.6137 0.3902 4 0.4016
Osaka 1 20 1 0.5923 6 0.6137 1 17.5 1 0.5872 6 0.3545
Tokyo 1 9.5 1 0.6343 9 0.6280 1 14 1 0.6317 11 0.4654
Shanghai 1 9.5 0.9944 0.7907 15 0.6853 0.8065 7 0.7656 0.8026 15 0.3863
Singapore 0.8318 2 0.8975 0.2471 1 0.5031 0.4579 1 0.4294 0.2852 2 0.1760
Shenzhen 1 15.5 1 1 20 0.6964 1 14 1 1 20 0.7059
Sydney 1 24 1 0.5793 5 0.9933 1 14 1 0.5028 1 0.4117
212
Table 7.22: The efficiency scores using different models and approaches (2)
BCC AHP/DEA
AHP/DEA
BCC Rank AHP/DEA AHP/DEA Referral AHP/DEA AHP/DEA -AR Rank
BCCI -AR Rank AHPI BCCII AHP II
reference I I -AR I Frequency II -AR II reference
reference I
II II
BCCI 1
BCC
0.5890 1
Rank reference I
AHP/DEAI 0.1179 -0.1206 1
AHP/DEA-AR 0.6005 0.6282 -0.0062 1

AHP/DEA-AR
0.5053 0.5747 -0.0670 0.9782 1
Rank reference I
AHPI 0.3536 0.5359 0.1009 0.4014 0.3580 1
BCCII 0.7581 0.6442 0.1352 0.6956 0.6021 0.3364 1

BCC Referral
0.5302 0.6629 -0.0700 0.6723 0.6494 0.3127 0.8677 1
Frequency II
AHP/DEA II 0.6643 0.6554 0.0854 0.7318 0.6606 0.4476 0.9093 0.8549 1
AHP/DEA-AR II 0.5871 0.5562 -0.0285 0.9752 0.9643 0.2986 0.7396 0.7159 0.7326 1
AHP/DEA-AR
0.4565 0.4626 -0.1491 0.9306 0.9554 0.1941 0.6287 0.6847 0.6545 0.9659 1
Rank reference II
AHP II 0.4034 0.4159 -0.3230 0.5767 0.5442 0.3083 0.5108 0.5954 0.5250 0.5489 0.5736 1
213
Ch7
7.5.4 ANOTHER HYPOTHESES TESTING METHOD

According to Hair and Black (2010), multiple regression is the use of two or more
independent variables in the prediction of the dependent variable. The task for a
researcher is to expand upon the simple regression model by adding independent
variable(s) that have the greatest additional predictive power. In this section, we might
expect the ownership of an airport and the location of an airport to result in higher
efficiency scores. In this research, we hope to answer the question: Does airport
ownership affect airport operational efficiency? In there, efficiency scores derived from
conducting the AHP/DEA-AR model are applied as the dependent variable, and airport
ownership is applied as the independent variable. Table 7.23 determines if there is
overwhelming evidence at the 𝛂 = 𝟎. 𝟎𝟓 level of a linear relationship between
operational efficiency and airport ownership. A t-test with n-2 degrees of freedom is
used. The critical values are 2.06 and -2.06. The t-test static is -0.05, and since -0.05 >-
2.06, we accept the null hypothesis that there is no significant relation between
efficiency and airport ownership.
Table 7.23: The result of regression analysis (1)

Regression Statistics
Multiple R 0.0111
R Square 0.0001
Adjusted R Square -0.0475
Standard Error 0.4984
Observations 23.0000
ANOVA
df SS MS F Significance F
Regression 1 0.0006 0.0006 0.0026 0.9600
Residual 21 5.2168 0.2484
Total 22 5.2174
Coefficients Standard Error t Stat P-value Lower 95% Upper 95%
Intercept 0.3650 0.3533 1.0329 0.3134 -0.3698 1.0997
Privatised -0.0239 0.4705 -0.0507 0.9600 -1.0023 0.9545
Furthermore, if adding another variable (location of airport), the question becomes: Do

airport ownership and airport location affect airport operational efficiency? From Table
7.24, it can be seen that another method is used to test the hypothesis: the p value
method. The p value is the probability of observing a test statistic more extreme than
214
Ch7
what we observed. In this research, the p values are 0.702 and 0.323, and both of them
are large than 0.05. Therefore the null hypothesis is accepted. This means that both
airport ownership and location have no relationship with airport operational efficiency.
Table 7.24: The result of regression analysis (2)

Regression Statistics
Multiple R 0.2242
R Square 0.0503
Adjusted R Square -0.0402
Standard Error 0.2184
Observations 24
ANOVA
df SS MS F Significance F
Regression 2 0.0530 0.0265 0.5556 0.5819
Residual 21 1.0017 0.0477
Total 23 1.0548
Coefficients Standard Error t Stat P-value Lower 95% Upper 95%
Intercept 0.7966 0.0995 8.0095 0.0000 0.5898 1.0034
Privatised -0.0446 0.1154 -0.3867 0.7029 -0.2846 0.1953
Asian -0.1135 0.1117 -1.0157 0.3213 -0.3458 0.1189
7.6 SUMMARY
The first section of this chapter described the concepts and analysis process of
efficiency scores using the AHP/DEA-AR model. The second and third sections
described the efficiency scores computed on VS I and VS II. The last section described
the results and compared them with those of the DEA-BCC model and the integrated
AHP/DEA model.
The Mann-Whitney test shows that the hypothesis that airports under private
management are more efficient than those under public management was not significant
in the case of both variable sets. The results reveal that the outcome of airport
efficiency is influenced by combining different evaluation techniques. In addition, the
AHP/DEA-AR model can also provide better discriminatory power for examiners to
use when evaluating airport performance.
215
Ch7
In addition, by using Saaty’s 1-9 scale, it can be seen that the AHP/DEA-AR model
provided a fair and useful technique by which to evaluate the performance of airports in
terms of their relative efficiencies, not only on the basis of the quality of variables but
also on the basis of the integration of diverse viewpoints. Moreover, when introducing
alternative judgement scales to the AHP model, the results were found to be more
robust as compared to the integrated AHP/DEA model, and these differences were very
significant. Therefore, this kind of calculation method can provide realistic results when
benchmarking airports. This means that relatively inefficient airports can be identified
to policy makers and civil aviation authorities.
In the last section, alternative thoughts about airport efficiency evaluation approaches
and hypothesis testing were presented. In this section, more possibilities for future
research are suggested.
216
Ch 8
CHAPTER 8
RESEARCH CONCLUSIONS
This final chapter starts with a brief summary of the entire research and then focuses
primarily on the researching findings that provide the answers to the research questions,
followed by the theoretical and methodological contributions and the managerial
implications of this study. Finally, the research concludes with some limitations and
highlights for future research.
217
Ch 8
8.1 RESEARCH CONCLUSIONS

The aims of this research were to examine the relationships between airport efficiency
and airport ownership or governance and to establish a reliable AEES. The research
was conducted using a number of different analytical methods. To achieve these aims,
it was necessary first to hypothesise this relationship and then to empirically examine
the relationship, and these two main tasks were accomplished through the use of a pilot
study intended to develop the AEES variables and an empirical study intended to
evaluate the relative efficiency of the airports under consideration.
A structured literature review defined the research gaps in research on this topic and
provided the inspiration for preliminary evaluation variables. In addition, a number of
methodological findings were also addressed (see Chapter 2). The recent revolution in
airport ownership and governance were described in order to help the author decide the
target regions and sample airports (see Chapter 3). A theoretical basis for the
development of the research design was then described and addressed (see Chapter 4).
The analytical methods for this research were expanded from the common approach
(the DEA method) to include an integrated AHP/DEA model, which has not yet been
employed previously to evaluate airport efficiency. In addition, the recommendations of
one of the experts who was interviewed were followed, and the AHP was subsequently
adopted in this research. Alternative scales were also used when conducting the AHP
method (see Chapter 4).
The AHP questionnaire survey was conducted in two fields (i.e. practice and academia).
A general picture of survey participants and their responses to the questions were
provided by the use of descriptive statistics (see Chapter 5). The airport efficiency
computed using the basic DEA models and an integrated AHP/DEA model were then
compared (see Chapter 6). A sensitivity analysis of the DEA model was also conducted.
Subsequently, another efficiency evaluation method (i.e. an AHP/DEA-AR model) was
used in this research, and the results were compared with the other evaluation methods.
In addition, some other thoughts about efficiency evaluation and hypothesis testing
approaches were also described (see Chapter 7).
218
Ch 8
8.2 RESEARCH FINDINGS

Revisiting the proposed Research Questions in Chapter 1, the following section
presents a brief summary of the key findings in accordance with of the each individual
research questions.
evaluation methods?
One the aims of this research is to determine airport efficiency using several different
evaluation techniques, (i.e. basic DEA models, which include the CCR and BCC
models, the integrated AHP/DEA model and the AHP/DEA-AR model), to evaluate 24
sample airports. The results show very clearly that adopting different evaluation
methods have an effect on the evaluation results in several ways, such as increases in
both the number of inefficient airports and the discriminatory power of the results
(addressed by mean and Standard Deviation of the efficiency scores). From Table 8.1,
the evidence shows that conducting different evaluation methods does change the
efficiency scores obviously.
Would an airport privatisation policy (airport ownership) influence the performance
of an airport’s operation?
The second Research Question examines if airports under private management are more
efficient than those under public management (as proposed and empirically tested in
Chapters 6 and 7). In this research, two variable sets (i.e. ten variables and six variables)
and three evaluation techniques are used. In addition, six alternative scales are
conducted by means of an AHP analysis. Table 8.1 shows a summary of efficiency
scores in the different models and the results of the Mann-Whitney test.
A hypothesis was set that the efficiency of privately operated airports should be shown
to be significantly different from those that are publicly operated. However, the results
using DEA-BCC models, the integrated AHP/DEA model, and the AHP/DEA-AR
219
Ch 8
model to evaluate airport efficiency, indicated no statistically significant difference in

airport efficiency between privately operated and publicly operated airports, even with
the application of two variable sets.
Although all of the results show that there to be no statistically significant difference,
efficiency scores which were calculated using the AHP/DEAAR model still provided
some evidence indicating that privately operated airports are more efficient than those
that are operated publicly.
Table 8.1: Summary of efficiency scores and the Mann-Whitney test

Method DEA AHP/DEA AHP/DEA-AR
VSI 0.9780 0.9706 0.7134
S.D. 0.0500 0.0880 0.2139
Public
VSII 0.8833 0.8466 0.7196
Average S.D. 0.1663 0.1946 0.2184
efficiency
scores VSI 0.9615 0.9801 0.7394
S.D. 0.1149 0.0598 0.1951
Private
VSII 0.8612 0.8484 0.7223
S.D. 0.1799 0.2067 0.2396
Z −0.1260 −0.7980 −0.4490
VSI
Mann- p 0.9000 0.5990 0.6530
Whitney
test Z −0.1270 −0.1910 −0.2700
VSII
p 0.8990 0.8610 0.8150
Does the influence of alternative scales on the results of the AHP analysis cause a
different weight for each variable?
The use of alternative comparison scales within AHP is discussed in Chapter 5. As

described, a series of pair-wise comparisons and a unit scale are used in the AHP to
play a fundamental role in quantifying a DM’s preference judgements. To date, the
Saaty 1-9 scale is the most used 9-unit scale in the AHP. Furthermore, although some
authors (see Chapter 4) have tried to debate the appropriateness of a 1-9 scale, the AHP
literature has not yet addressed which of the available alternative scales are most
appropriate for the process of making pair-wise comparisons. The influence of
alternative scales on the results of the AHP analysis is assessed in this research.
220
Ch 8
To overcome the deficiencies of the 1-9 scale, various judgement scales for a pair-wise
comparison were proposed and evaluated. In this research, Saaty’s 1-9 scale and other
five alternative scales were used to calculate the weights of each variable (i.e. e1-9, 21-9
9/9-9/1, 10/10-18/2, ø mapping). The results of the AHP analysis in Chapter 6 show
that alternative scales cause a different weight for each variable; however, the
difference was not shown to be significant (see Table 5.15). In addition, the outcome
also revealed that scales 1-9, e1-9 and 21-9 can produce similar weights and that scales
10/10-18/2 and ø mapping can result in similar weights. Moreover, applying these
weights with DEA, the efficiency scores did not demonstrate obvious differences, but
with the DEA-AR models, the differences among the efficiency scores can be easily
recognised.
Research Question 4
Does the number of input and output variables affect the results of airport efficiency
evaluation?
In order to answer this research question, two variables (ten variables and six variables)
were undertaken in the DEA analysis in this research. An insufficient number of
variables for a DEA model will tend to rate all DMUs as being 100% efficient because
of an inadequate number of degrees of freedom. Hence, a proper variablesnumber is
required for identifying a true performance frontier (Zhang and Bartels, 1998). A rule
of thumb for maintaining an adequate number of degrees of freedom when using DEA
is to obtain at least two variables for each input or output measure (Bowlin, 1987). In
addition, Cooper et al. (2006) found that this also to be a part of DEA sensitivity
analysis. The results from Chapter 6 and Chapter 7 show that the numbers of variables
do affect the results related to airport efficiency. Furthermore, using fewer numbers of
variables (in this research in VSII) or the use of the AHP/DEA-AR model can provide a
higher discriminatory power for the AEES.
8.3 CONTRIBUTIONS AND IMPLICATIONS

The key highlights of the proposed research contributions and managerial implications
are discussed in this section.
221
Ch 8
8.3.1 THEORETICAL CONTRIBUTIONS

Firstly, to the best of author’s knowledge, this research is the first effort to make a
comparative assessment of the impact of airport ownership or governance on efficiency
involving European and Asia-Pacific airports. There have been some previous studies
that have not focused on a specific region, such as Oum et al. (2006), who focused on
109 airports around the world, or Lin and Hong (2006), who evaluated 20 major
airports around the world. Although Oum et al. (2006) showed that airport ownership
affects airport efficiency, Lin and Hong (2006) found that airport efficiency is not
affected by airport ownership. This difference may be due to the time differences
between these two studies: Oum et al.’s (2006) study covered a four year period
between 2003 and 2005 while Lin and Hong (2006) only used data drawn from a one
year period.
Secondly, this research has objectively established an Airport Efficiency Evaluation

System (AEES). A two steps AHP analysis was conducted in this research (i.e. a pilot
semi-structured interview and structured interview or questionnaire survey). Following
this survey, 38 experts in total were selected from academia and practice for interviews
(i.e. three experts in the pilot interview and 35 experts in the questionnaire survey).
Thirdly, this research is the first research to conduct a structured literature review on
the development of benchmarking techniques in airport performance evaluation
research. From this literature review, the measurement methodologies, the variables
used, and the results associated with various airport activities have been realised. In
addition, some attention has been given to the increased employment of mathematical
modelling and advanced statistical analysis methods.
Finally, this research is also the first research to apply two different evaluation models
(i.e. the AHP/DEA model and the AHP/DEA-AR model) to evaluate airport efficiency
with an AEES.
8.3.2 METHODOLOGICAL CONTRIBUTIONS

This research has made an important contribution in terms of methodology. It is the
first effort to adopt AHP alternative scales, an integrated AHP/DEA model, and an
integrated AHP/DEA-AR model on airport performance. Although the AHP method
222
Ch 8
has been widely employed in many other studies (which were mentioned in Chapter 5),
it has not previously been used to assess airport efficiency. Therefore, the analysis
process described and practiced in this study can provide guidance to air transport
researchers who wish to use these techniques.
In particular, the analytical technique used in this research addressed various issues and
included advanced techniques in AHP and DEA. Firstly, DEA was compared with
other similar techniques, such as SFA and TFP (see Chapter 2). This comparison
provided some useful information on the advantages and disadvantages of DEA over
the other techniques and the situations where DEA needs to be used.
Secondly, in this study, alternative scales of AHP were conducted, which have been
argued in AHP literature for a long time. This research selected five other scales to
calculate the weights of variables, which were compared with Saaty’s 1-9 scale.
Thirdly, multiple groups of experts were interviewed in the AHP analysis. The multiple
group analysis in this study covered most of the possible interviewees in airport
efficiency. This means that the AEES can be established more reliably.
Fourthly, some other thoughts about efficiency evaluation implied from AHP weights
and cluster analysis as well as hypothesis testing are introduced in this research.
Although the whole process is still not very mature, it also provides some guidance for
future research.
8.3.3 MANAGERIAL IMPLICATIONS AND PRACTICAL CONTRIBUTIONS

For practitioners, the identification of the variables affecting airport operations explored
in this study may stimulate more carefully considered transport decision-making by
providing a more accurate and precise framework for airport planning. The slack
analysis can also provide guidance by which to set up targets to improve airport
efficiency.
Secondly, the review of the devolution of airport governance and ownership in different
countries can also help policy makers and practitioners gain some experience and
feedback when deciding to change the ownership or governance of airports. In addition,
223
Ch 8
the results of this research also can provide some evidence to persuade people to
support airport privatisation policy.
Finally, an AEES was built in this research, and this system provided several variables
based on finance, service, and airport capacity. By means of this system, airport
managers or airport authorities can assess their airports easily or provide some
guidelines that will enable them to establish their own variable set.
8.4 LIMITATIONS AND SUGGESTIONS FOR FURTHER RESEARCH

Any study encounters limitations, and this study is no exception. The limitations in this
study can provide potential directions for further research.
Firstly, this study only successfully interviewed 22 experts in the area of air transport.
Future research can try to expand the number of experts; in that case, the AEES or
research results may have different results or concepts. In addition, the experts in this
research only included two points of view (i.e. scholars and airport managers), whereas
additional research could investigate the proposed variables from the civil aviation
authority perspective concurrently. Therefore, a new viewpoint may aid in the
accumulation of information and provide new insights into this topic. Secondly, other
qualitative approaches (such as a focus group or Delphi studies) could be adapted to
construct the weights of each variable. Thirdly, the application of the AEES could be
taken into other geographical areas to cross-validate the findings of this research. As
pointed out in Chapter 1, the Asia-Pacific region and Europe were selected as the
location of interest for this study. Therefore, confidence in the applicability of the
research model can be increased if the cross-validation and invariance of the model are
verified in other geographical locations.
As mentioned in Chapter 5, only one year’s worth of panel data of the sample airports
was applied in this research. Therefore, it is recommended that a study that gathers
more panel data would be able to increase the confidence of the applicability of the
research model. Finally, only major airports were selected in this research because of
the problem of data availability. It is therefore recommended that future research apply
this model to medium or small airports to increase the confidence in regard to the
applicability of the research model.
224
Bibliography
b BIBLIOGRAPHY a
Aastrup, J. and Halldorsson, A. 2008. Epistemological role of case studies in logistics:

A critical realist perspective, International Journal of Physical Distribution &
Logistics Management, 38(10), pp. 746-763.
Abbott, M. and Wu S. 2002. Total Factor Productivity and efficiency of Australian
airports. The Australian Economic Review 35(3), pp. 244-260.
Ablanedo-Rosas, J.H. and Gemoets, L.A. 2010. Measuring the efficiency of Mexican
airports. Journal of Air Transport Management 16(6), pp. 343-345.
Adler, N. and Berechman, J. 2001. Measuring airport quality from the airlines’
viewpoint: An application of data envelopment analysis. Transport Policy 8(3), pp.
171-181.
Aguarón, J., and Moreno-Jiménez, J.M. 2004. The geometric consistency index.
Approximated thresholds. European journal of operational research 147(1), pp.137-
145.
Airport Council International. 2011. ACI Airport Economics Survey 2010. Geneva: ACI.
Aigner, D.J., Lovell, C.A.K., and Schmidt, P. 1977. Formulation and estimation of
stochastic frontier production function models. Journal of Econometrics 6(1), pp. 21-
37.
Airbus. 2012. Global market forecast. Lavaur: Airbus.
Ali, I., Charnes, A., Cooper, W., Divine, D., Klopp, G.A., and Stutz, J. 1987. An
application of data envelopment analysis to US Army recruitment districts. In:
Schultz, R.L. ed. Applications of management science. New York: JAI press.
Ali, A., Seiford, L.M. 1993. The mathematical programming approach to efficiency
analysis. In: Fried, H.O., Lovell, C.A.K., and Schmidt, S.S. ed. The Measurement of
Productive Efficiency: Techniques and Applications. New York: Oxford University
Press.
Allen, R., Athanassopoulos, A., Dyson, R.G., and Thanassoulis, E. 1997. Weights
restrictions and value judgements in Data Envelopment Analysis: Evolution,
development and future directions. Annals of operations research 73, pp. 13-34.
Anderson L., Walberg H.J., and Weinstein, T. 1998. Efficiency and effectiveness
analysis of Chicago public elementary schools: 1989, 1991, 1993. Educational
Administration Quarterly 34(4), pp. 484-504.
Arbel, A. 1989. Approximate articulation of preference and priority derivation.
European journal of operational research 43(3), pp. 317-326.
Armitage, A. and Keeble-Allen, D. 2008. Undertaking a Structured Literature Review
or Structuring a Literature Review: Tales from the Field. The Electronic Journal of
Business Research Methods 6(2), pp. 103-114.
Ashford, N. and Moore, C. A. 1992. Airport finance. New York: Van Nostrand
Reinhold.
Assaf, A. 2009. Accounting for size in efficiency comparisons of airports. Journal of
Air Transport Management 15(5), pp. 256-258.
225
Bibliography
Assaf, A. 2010. The cost efficiency of Australian airports post privatization. Tourism
Management 31(2), pp. 267-273.
Assaf, A. 2011. Bootstrapped Malmquist indices of Australian airports. The Service
Industries Journal 31(5), pp. 829-846.
Assaf, A and Gillen, D. 2012. Measuring the joint impact of governance form and
economic regulation on airport efficiency. European journal of operational research
220(1), pp. 187-198.
Assaf, A, Gillen, D., and Barros, C. 2012. Performance assessment of UK airports:
Evidence from a Bayesian dynamic frontier model. Transportation research part E
48(3), pp. 603-615.
Avison, D.E. and Fitzgerald. G. 1995. Information systems development: methodologies,
techniques and tools. London: McGraw-Hill.
Azadeh, A., Ghaderi, S.F. and Izadbakhsh, H. 2008. Integration of DEA and AHP with
computer simulation for railway system improvement and optimization. Applied
Mathematic and computation 195, pp. 775-785.
BAA. (2011). BAA history. [online]. London: BAA. Available at: http://www.baa.com/.
[Accessed 25th January 2011].
BAA. (2012). BAA history. [online]. London: BAA. Available at: http://www.baa.com/.
[Accessed 31st August 2012].
Banker, R.D., Charnes, A., and Cooper, W. 1984. Some models for estimating technical
and scale inefficiencies in Data Envelopment Analysis. Management science 30(9),
pp. 1078-1092.
Barros, C.P. 2008a. Airports in Argentina: Technical efficiency in the context of an
economic crisis. Journal of Air Transport Management 14(6), pp.315-319.
Barros, C.P. 2008b. Technical efficiency of UK airports. Journal of Air Transport
Barros, C.P. 2009. The measurement of efficiency of UK airports: Using a stochastic
latent class frontier model. Transport Reviews 29(4), pp.479-498.
Barros. C. P. 2011. Cost efficiency of African airports using a finite mixture model.
Transport Policy 18(6), pp. 807-813.
Barros, C.P. and Dieke, P.U.C. 2007. Performance evaluation of Italian airports: A data
envelopment analysis. Journal of Air Transport Management 13(4), pp. 184-191.
Barros, C.P. and Dieke, P.U.C. 2008. Measuring the economic efficiency of airports: A
Simar-Wilson methodology analysis. Transportation Research Part E 44(6), pp.
1039-1051.
Barros, C. P., Managi, S., and Yoshida, Y. 2010. Technical efficiency, regulation and
heterogeneity in Japanese airports. Pacific Economic Review 15(5), pp. 685-696.
Barros, C.P. and Sampaio, A. 2004. Technical and allocative efficiency in airports.
International Journal of Transport Economics 31(3), pp. 355-378.
Bazargan, M. and Vasigh, B. 2003. Size verus efficiency: a case study of US
commercial airports. Journal of Air Transport Management 9(3), pp. 187-193.
226
Bibliography
Bech, M. and Kristensen, M.B. 2009. Differential response rates in postal and web-
based survey among older respondents. Survey research methods 3(1), pp. 1-6.
Beesley, M. E. 1997. Privatization, Regulation and Deregulations, 2nd ed. London:
Routledge.
Behzadian, M., Kazemzadeh, R.B., Albadvi, A., and Aghdasi, M. 2009. PROMETHEE:
A comparative analysis literature review on methodologies and applications.
European Journal of Operational Research 200(1), pp. 198-215.
Belton, V. 1986. A comparison of the analytical hierarchy process and a simple multi-
attribute value function. European journal of operational research 26(1), pp. 7-21.
Benayoun, R., Roy, B., and Sussman, B. 1966. ELECTRE: une methode pour guider le
choix en presence de points de vue multiples SEMA-METRA. International
Direction Scientifique 49.
Benton, T. and Craib, I. 2001. Philosophy of social science. New York: Palgrave.
Berger, A.N. and Humphrey, D.B. 1997. Efficiency of Financial Institutions:
International Survey and Directions for Future Research. European Journal of
Operational Research 98(2), pp. 175-212.
Berrittella, M., Franca, L.L., and Zito, P. 2009. An analytic hierarchy process for
ranking operating costs of low cost and full service airlines. Journal of Air Transport
Beynon, M.J. 2002. An analysis of distributions of priority values from alternative
comparison scales within AHP. European Journal of Operational Research. 140(1),
pp. 104-117.
Benayoun, R., Roy, B., and Sussman, B. 1966. ELECTRE: Une methode pour guider le
choix en presence de points de vue multiples. Sema (Metra International), Direction
Scientifi que, Note de Travail , No. 49.
Berger, P. L., and Luckmann, T. 1966. The social construction of reality: A treatise in
the sociology of knowledge. New York: Doubleday & Company.
Berrittella, M., Franca, L.L., and Zito, P. 2009. An analytical hierarchy process for
ranking operating cost of low cost and full service airlines. Journal of Air Transport
Beyth-Marom, R. how probable is probable? A numerical translation of verbal
probability expressions. Journal of forecasting 1, pp. 257-269.
Bishop, M. and Thompson, D. 1992. Privatisation in the UK: internal organization and
productive efficiency. Annals of Public & Cooperative Economics 63(2), pp. 171-
189.
Brans J.-P. and Mareschal, B. 2005. PROMETHEE METHODS. In: Figueria J, Greco
S, and Ehrgott, M. ed. Multiple criteria decision analysis: State of the art surveys.
New York: Springer.
Boeing, 2012. Current market outlook 2012 [online]. Settle: Boeing. Available at:
http://www.boeing.com/commercial/cmo/ [Accessed: 19th July 2012].
Bojkovic, N., Anic, I. and Pejcic-Tarle, S. 2010. One solution for cross-country
transport-sustainability evaluation using a modified ELECTRE method. Ecological
Economics 69(5), pp. 1176-1186.
227
Bibliography
Bolloju, N. 2001. Aggregation of analytic hierarchy process models based on

similarities in decision makers’ preference. European Journal of Operational
Research 128(3), pp. 499-508.
Bowlin, W. 1987. Evaluating the Efficiency of US Air Force Real-Property
Maintenance Activities. The Journal of the Operational Research Society 38(2), pp.
127-136.
Brans, J.-P., and Mareschal, B. 2005. PROMETHEE METHODS. In: Figueria, J. Greco,
S.,and Ehrgott, M. ed. Multiple Criteria Decision Analysis: State of the Art Surveys.
New York: Springer.
Bureau, J. C, Fare, R, and Grosskopf, S. 1995. A comparison of 3 nonparametric
measures of productivity growth in European and United-States agriculture. Journal
of Agricultural Economics 46(3), pp. 309-26.
Burton, D. 2000. The use of case studies in social science research. In: Burton, D. ed.
Research Training for Social Scientists. London: Sage.
Brugha, C.M. 2000. Relative measurement and the power function. European journal
of operational research 121(3), pp. 627-640.
Bryman, A. 2008. Why do researchers integrate/combine/mesh/blend/mix/merge/ fuse
quantitative and qualitative research. In: Bergman, M.M. ed. Advances in mixed
methods research. London: Sage.
Bryman, A. and Bell, E. 2003. Business Research Method. Oxford: Oxford University
Press.
Bryman, A. and Bell, E. 2007. Business Research Method, 2nd ed. Oxford: Oxford
University Press.
Bryman, A. and Bell, E. 2011. Business Research Method, 3rd ed. Oxford: Oxford
University Press.
Carney, M. and Mew, K. 2003. Airport governance reform: A strategic management
perspective. Journal of Air Transport Management 9 (4), pp. 221-232.
Carmmone, Jr. F.J., Kara, A., and Zanakis, S.H. 1997. A Monte Carlo investigation of
incomplete pairwise comparision matrices in AHP. European Journal of Operational
Research 102(3), pp. 538-553.
Castelli, L. and Pellegrini, P. 2011. An AHP analysis of air traffic management with
target windows. Journal of Air Transport Management 17(2), pp. 68-73.
Chang, Y. H., Yeh, C. H., and Liu, Y. L. 2006. Prioritizing management issues of
moving dangerous goods by air transport. Journal of Air Transport Management
12(4), pp. 191-196.
Charnes, A., Cooper, W., and Rhodes, E. 1978. Measuring the efficiency of decision
making units. European Journal of Operational Research 2(6), pp. 429-444.
Charnes, A., Cooper, W., Golany, B. Seiford, I. and Stutz, J. 1985. Foundations of data
envelopment analysis for Pareto-Koopmans efficient empirical production functions.
Journal of Econometrics, 30, pp. 91-107.
Charnes, A., Cooper, W., Lewin, A.Y., Seiford, L.M. 1994. Data Envelopment Analysis:
Theory, Methodology and Applications. Boston: Kluwer Academic.
228
Bibliography
Charnes, A., and Neralic, L. 1990. Sensitivity Analysis of the Additive Model in Data
Envelopment Analysis. European Journal of Operations Research 48(4), pp. 332-
341.
Chen, T.Y., and Chen, L.H. 2007. DEA performance evaluation based on BSC
indicators incorporated: the case of semiconductors industry. International Journal
of Productivity and Performance Management 56(4), pp. 335-57.
Chen, C.-C., Chen, J., and Lin, P.-C. 2009. Identification of significant threats and
errors affecting aviation safety in Taiwan using the analytic hierarchy process.
Journal of Air Transport Management 15(5), pp. 261-263.
Chen, C.J. and Yen, Q.J. 2005. A comparative performance evaluation of Taiwan’s
high-tech industries. International journal of business performance management 7(1),
pp. 16-33
Chi-Lok A.Y. and Zhang, A. 2009. Effects of competition and policy changes on
Chinese airport productivity: An empirical investigation. Journal of Air Transport
Chow, C.K.W. and Fung, M.K.Y. 2012. Estimating indices of airport productivity in
grater china. Journal of Air Transport Management 24, pp. 12-17.
Churchill, G.A. 1976. Marketing research: Methodological foundation. Illinois: The
Dyrden Press.
Churchill, G.A. and Iacobucci, D. 2002. Marketing research: Methodological
foundation 8th ed. Orlando: Harcourt College Publishers.
Civil Aviation Administration of China, 2010. The Civil Aviation Airport Production
Statistics Communique´ of 2010 [Online] Beijing: CAAC. Available at:
http://www.caac.gov.cn/E_Pub- WebApp/index.aspxS [Accessed: 10 April 2012].
Clark, D., Lotto, L. and Astuto, T. 1984. Effective schooling and school improvement:
A comparative analysis of two lines to inquiry, Educational Administration
Quarterly 20, pp. 41-68.
Coelli, T. J., Prasada Rao, D.S., O’Donnell, C. J., and Battese, G. E. 2005. An
Introduction to Efficiency and Productivity Analysis. 2nd ed. New York: Springer.
Comin, D. 2006. Total Factor Productivity. In: Durlauf, S. N. and Blume, L.E. ed. The
New Palgrave Dictionary of Economics, 2nd ed. London: Palgrave Macmillan.
Competition Commission. 2008. Working paper on benchmarking the operating
performance of BAA airports. [Online] London: Competition Commission. Available
at: http://www.competitioncommission.org.uk/. [Accessed 30th May 2010].
Cooper, W. W., Seiford, L. M., and Tone, K. 2006. Introduction to Data Envelopment
Analysis and it uses. New York: Springer.
Cullinane, K., Firth, M., Fung, M.K.Y, Gong, S.X.H., Law, J.S., and Zang, A. 2011.
Development of air transport industry in the Asia-Pacific region. Pacific economic
review 16(1), pp. 42-46.
Curi, C., Gitto, S. and Mancuso, P. 2010. The Italian airport industry in transition: a
performance analysis. Journal of Air Transport Management 16(4), pp. 218-221.
229
Bibliography
Curi, C., Gitto, S. and Mancuso, P. 2011. New evidence on the efficiency of Italian
airports: A bootstrapped DEA analysis. Socio-Economic Planning Sciences 45(2), pp.
84-93.
DeVaus, D. 2002. Survey in Social Science, 5th ed. London: Routeledge.
Denscombe, M. 2003. The Good Research Guide for Small-Scale Social Research
Projects. 2nd ed. Maidenhead: Open University Press.
Dezin, N.K. and Lincoln, Y.S. 1994. Introduction: entire the field of qualitative
research. In: Dezin, N.K. and Lincoln, Y.S. ed. Handbook of qualitative research.
Thousand Oaks: Sage.
Doganis, R. 1992. The Airport Business. London: Routledge.
Doganis, R., Graham, A. and Lobbenberg, A. 1995. The economic performance of
European airports. Research Report 3. London: University of Westminster.
Doganis, R. 2002. Flying off course: The economics of international airlines. 3rd
London: Routledge.
Donegan, H.A., Dodd, F.J., McMaster, T.B., 1992. A new approach to AHP decision-
making. The Statistician 41(3), pp. 295-302.
Doyle, J., Green, R. 1994. Efficiency and cross-efficiency in DEA: Derivations,
meanings and uses. Journal of the Operational Research Society 45(5), pp. 567-578.
Dyer, J.S. 1990. Remarks on the Analytic hierarchy process. Management science
36(3), pp. 249-258.
Efron, B. and Tibshirani, R. 1994. An Introduction to the Bootstrap. New York:
Chapman & Hall.
Ertay, T. Ruan, D., and Tuzkaya U.R. 2006. Integrating data envelopment analysis and
analytic hierarchy for the facility layout design in manufacturing systems.
Information Sciences, 176 (3), pp. 237-262.
Eisenhardt, K. M. 1989. Building theories from case study research. Academy of
Management Review 14(4), pp. 532-550.
Escobar, M.T., Aguarón, J., and Moreno-Jiménez, J.M. 2004. A note on AHP group
consistency for the row geometric mean priorization procedure. European journal of
operational research 153(2), pp. 318-322.
Federal Aviation Administration. 2006. Airport categories. [Online]. Washington:
Federal Aviation Administration. Available at:
http://www.faa.gov/airports_airtraffic/airports/planning_capacity/passenger_allcargo
_stats/categories/ [Accessed on October 15, 2011].
Federal Aviation Administration. 2011. National Plan of Integrated Airport Systems
(NPIAS), 2001–2005. Washington, D.C.: FAA.
Feng, Y. J., Lu, H., and Bi, K. 2004. An AHP/DEA method for measurement of the
efficiency of R&D management activities in university. International Transactions
in Operational Research 11, pp. 181-191.
Fernandes, E. and Pacheco, R.R. 2002. Efficient use of airport capacity. Transportation
Research Part A 36(3), pp. 225-238.
230
Bibliography
Fodness, D. and Murry, B. 2007. Passengers’ expectation of airport service quality.

Journal of Services Marketing. 21, pp. 492-506.
Forman, E. and Peniwati, K. 1998. Aggregating individual judgments and priorities
with the analytic hierarchy process. European Journal of Operational Research.
108(1), pp. 165-169.
Forsyth, P.J. 2000. Models of airport performance. In: Hensher, D.A. and Button, K.J.
ed. Handbook of Transport Modelling. Oxford: ELSEVIER.
Forsyth, P.J. 2006. Airport competition: Regulatory issues and policy implications. In:
Lee, D. ed. Advances in Airline Economics: Competition policy and antitrust. Oxford:
Elsevier.
Francis, G., Humphreys, I. and Fry, J. 2002. The benchmarking of airport performance.
Freeling, A.N.S. 1983. Belief and decision aiding. PhD thesis, Cambridge University.
French, S. 1980. On the methodology of decision analysis III: On various non Bayesian
aids. Notes in decision theory No. 89. Department of decision theory. University of
Manchester.
Friedman, L., and Sinuany-Stern, Z. 1998. Combining ranking scales and selecting
variables in the DEA context: the case of industrial branches. Computers and
Operations research 25 (9), pp. 781-791
Fung, M.K.Y., Wan, K.K.H. Hui, Y.V. and Law, J.S. 2008. Productivity changes in
Chinese airports 1995–2004. Transportation Research Part E 44(3), pp. 521-542.
Fung, M.K.Y. and Chow, C.K.W. 2011. Note on the productivity convergence of
airports in China. Pacific economic review 16(1), pp. 120-133.
Gattoufi, S., Oral, M., and Reisman, A. 2004. Data envelopment analysis literature: a
bibliography update (1951-2001). Socio-Economic planning science 38, pp. 159-229.
Getwick airport. (2011). Getwick history. [Online]. London: Gatwick Airport Limited.
Available at: http://www.gatwickairport.com/. [Accessed: 25th January 2011].
Ghauri, P. N. and Grønhaug, K. 2002. Research methods in business studies: A
practical guide. London: Pearson Education.
Giddens, A. 1974. Positivism and Sociology. London: Heinemann.
Gillen, D. and Lall, A. 1997. Developing measures of airport productivity and
performance: an application of Data Envelopment Analysis. Transportation
Research Part E 33(4), pp. 261-273.
Gillen, D. and Lall, A. 2001. Non-parametric measures of efficiency of U.S. airports.
International Journal of Transport Economics 28(3), pp. 283-305.
Gillen, D. 2011. The evolution of airport ownership and governance. Journal of Air
Transport Management 17(1), pp. 3-13
Giokas, D. 1991. Bank branch operating efficiency: A comparative application of DEA
and the loglinear model. Omega 19(6), pp. 549-557.
Gitto, S. and Mancuso, P. 2012a. Two facts of airports business: A non-parametric
analysis of the Italian airport industry. Journal of Air Transport Management 20, pp.
39-42.
231
Bibliography
Gitto, S. and Mancuso, P. 2012b. Bootstrapping the Malmquist indexes for Italian
airports. International Journal of production economics 135, pp. 403-411.
Golany, B., Roll, Y., 1989. An application procedure for DEA. Omega 17(3), pp. 237-
250.
Gong, S.X.H., Cullinane, K., and Firth, M. 2012. The impact of airport and seaport
privatization on efficiency and performance: A review of the international evidence
and implications for developing countries. Transport policy 24, pp. 37-47.
Gonzalez, M.M. and Trujillo, L. 2009. Efficiency measurement in the port industry: A
survey of Empirical evidence. Journal of Transport Economics and Policy 43(2), pp.
157–192.
Graham, A. 2005. Airport benchmarking: a review of the current situation.
Benchmarking: An International Journal 12(2), pp. 99–111.
Graham, A. 2006. Competition in airports. In: Papatheodorou, A. ed. Rivalry and
market power. New York: I.B. Tauris & Co. Ltd.
Graham, A. 2008. Managing airports: An international perspective. 3rd. Oxford:
Elsevier.
Graham, A. 2009. How important are commercial revenues to today’s airports? Journal
of Air Transport Management 15(3), pp. 106-111.
Graham, A. 2011. The objective and outcomes of airport privatisation. Research in
Transportation business & management 1(1), pp. 3-14.
Greene, W. 2005. Fixed and random effects in stochastic frontier models. Journal of
Productivity Analysis 23(1), pp. 7-32.
Grupo Ferrovial (2007). Annual report 2006. Madrid: Grupo Ferrovial.
Guo, J.Y., Liu, J., and Qiu, L. 2006. Research on supply chain performance evaluation:
based on DEA/AHP model. Proceedings of the IEEE Asia-Pacific Conference on
Services Computing (APSCC'02).
Guba, E.G. and Lincoln, Y.S. 1994. Competing paradigms in qualitative research. In:
Denzin, N.K. and Lincoln, Y.S. ed. Handbook of qualitative research. Thousand
Oaks: Sage.
Guba, E.G. and Lincoln, Y.S. 2005. Paradigmatic Controversies, Contradictions, and
Emerging Confluences. In: Denzin, N.K. and Lincoln, Y.S. ed. Handbook of
qualitative research. 3rd ed. Thousand Oaks: Sage.
Hair, J. and Black, W. 2010. Multivariate data Analysis. 7th . London: Prentice Hall.
Hämäläinen, R.P. and Salo, A.A. 1997. The issue is understanding the weights, Journal
of multi-criteria decision analysis 6, pp. 340-343.
Hamm, R.M. 1991. Selection of verbal probabilities: a solution for some problems of
verbal probability expression. Organization behaviour human decision process 48,
pp. 193-223.
Hamzaee, R.G. and Vasigh, B. 2000. A simple revenue-cost perspective on US airport
operations. Journal of Air Transport Management 6(1), pp. 61-64.
Harker, P.T. and Vargas, L.G. 1987. The theory of ratio scale estimation: Saaty’s
analytic hierarchy process. Management science 33(11), pp. 1383-1403.
232
Bibliography
Harker, P.T. 1987. Alternative models of questioning in the analytic hierarchy process.
Mathematical modelling 9(11), pp. 353-360.
Hershey, J.C., Kunreuther, H.C., and Schoemaker, P.J.H. 1982. Sources of bias in
assessment procedures for utility functions. Management science 28(8), pp.936-939.
Ho, W. 2008. Integrated analytical hierarchy process and its applications-A literature
review. European journal of operational research 186(1), pp. 211-228.
Hooper, P.G. and Hensher, D.A. 1997. Measuring total factor productivity of airports:
An index number approach. Transportation Research Part E 33(4), pp. 249-259.
Hsu, K.H. 2005. Using balanced scorecard and fuzzy data envelopment analysis for
multinational R&D project performance assessment. Journal of America Academy of
Business 7(1), pp. 189-96.
Hu, F. H., Jiang, J., Liu, L., Sun, L. Y. and Ji, Y. C. 2010. A New Multi-perspective
Framework for Multi-attribute Decision Making, Expert Systems with Applications
37(12), pp. 8575-82.
Huang, H. 2004. Estimation of technical inefficiencies with heterogeneous technologies.
Journal of Productivity Analysis 21(3), pp. 277-296.
Humphreys, I. 1999. Privatisation and commercialisation change in UK airport
ownership patterns. Journal of Transport Geography 7(2), pp. 121-134.
Humphreys, I. and Francis, G. 2002a. Airport privatisation. British economy review
31(2), pp. 51-56
Humphreys, I. and Francis, G. 2002b. Performance measurement: a review of airports.
International Journal of Transport Management 1(2), pp. 79-85.
Hwang, C.L. and Yoon, K. 1981. Multiple attribute decision making. New York:
Spring-Verlag.
International Civil Aviation Organisation (ICAO). 2006. Airport Economics manual.
Montréal: ICAO.
Ison, S., Francis, G., Humphreys, I., and Page, R. 2011. UK regional airport
commercialisation and privatisation: 25 years on. Journal of transport geography
19(6), pp. 1341-1349.
Jahanshahloo, G. R., Hosseinzadeh, Lotfi, F., and Moradi, M. 2004. Sensitivity and
stability analysis in DEA with interval data. Applied Mathematics and Computation
156(6), pp. 463-477.
Jarach, D. 2005. Airport Marketing. Surrey: Ashgate.
Jensen, R.E. 1984. An alternative scaling method for priorities in hierarchical structures.
Journal of mathematical psychology 28(3), pp. 317-332.
Jing-yuan, G., Jia, L., and Li, Q. 2006. Research on supply chain performance
evaluation based on DEA/AHP model. Proceedings of the 2006 IEEE Asia-Pacific
Conference on Services Computing (APSCC'06).
Jyoti, T. and Banwet, D. K. and Deshmukh, S. G. 2008. Evaluating performance of
national R&D organizations using integrated DEA-AHP technique. International
Journal of Productivity and Performance Management 57(5), pp. 370-388.
Joshi, G. 2000. Overview of Privatization in South Asia. New Delhi: ILO.
233
Bibliography
Kao, C. 2009. Efficiency decomposition in network data envelopment analysis: A

relational model. European journal of operational research 192(3), pp. 949-962.
Kalirajan, K.P. and Obwona, M.B. 1994. Frontier production function: the stochastic
coefficients approach. Oxford Bulletin of Economics and Statistics 56(1), pp. 87-96.
Kinnear, T.C. and Taylor, J.R. 1991. Marketing Research: an applied approach.
London: McGraw-Hill.
Kok, M. and Lootsma, F.A. 1984. Pairwise comparsion methods in multiobjective
programming, with applications in a long-term energy-planning model. Technical
report 84(19). Delft university of technology.
Korpela, J., Lehmusvaara, A., and Nisonen, J. 2007. Warehouse operator selection by
combining AHP and DEA methodologies. International Journal of Production
Economics 108, pp. 135-42.
Kumbhakar, S., Park, B., Simar, L., and Tsionas, E. 2007. Nonparametric stochastic
frontiers: a local maximum likelihood approach. Journal of Econometrics 137(1), pp.
1-27.
Lai, P., Potter, A. and Beynon, M. 2012. The development of benchmarking techniques
in airport performance evaluation research. Transportation Journal 51 (3), pp. 305-
337.
Lam, S., Low, J. and Tang, L. 2009. Operational efficiencies across Asia Pacific
airports. Transportation Research Part E 45(4), pp. 654-665.
Lee, T. Z., and Tseng, Y. F. 2006. Explore the optimal Allocation Problem with
AHP/DEA. Asia Pacific Management Review 11(6), pp. 389-392.
Legrady, K. Lootsma, F.A., Meisner, J. Schellemans, F. 1984. Multicriteria decision
analysis to aid budget allocation. In: Grower, M. Woerzbick, A.P. ed. Interactive
Decision Analysis. NEW York: Spring-Berlag
Leung, L.C., and Cao, D. 2000. On consistency and ranking of alternatives in Fuzzy
AHP, European journal of operational research 124(1), pp. 102-113.
Levin, D.M. 1988. The opening of vision: Nihilism and the postmodern situation.
London: Routledge.
Lewin, A. Y. and Minton, J. W. 1986. Determining Organizational Effectiveness:
Another Look and an Agenda for Research. Management Science 32(5), pp. 514-538.
Lin, L. C. and Hong, C.H. 2006. Operational performance evaluation of international
major airports: An application of data development analysis. Journal of Air
Transport Management 12(6), pp. 342-351.
Lootsma, F.A. 1989. Fuzzy performance evaluation of nonlinear optimization methods,
with sensitivity analysis of the final scores. Journal of information and optimisation
science 10, pp.15-44.
Lozano, S. and Villa, G. 2007. Integer DEA models. How DEA models can handle
integer inputs and outputs. In: Zhu, J., Cook, W. ed. Modelling data irregularities
and structural complexities in data envelopment analysis. New York: Springer,
Lozano, S. and Gutierrez, E. 2011. Efficiency analysis and target setting of Spanish
airports. Network and Spatial Economics 11(2), pp. 139-157.
234
Bibliography
Ma, J. and Zheng, X. 1991. 9/9-9/1 scale method of AHP. Proceedings of the second
international symposium on the AHP, pp. 197-202.
May, T. 2001. Social research: issues, methods and process. 3rd ed. Buckingham: Open
University Press.
Manataki, I.E. and Zografos, K.G. 2009. Assessing airport terminal performance using
a system dynamics model. Journal of Air Transport Management 16(2), pp. 86-93.
Martin-Cejas, R.R. 2002. An approximation to the productive efficiency of the Spanish
airports network through a deterministic cost frontier. Journal of Air Transport
Management 8(4), pp. 233–238.
Martin, J.C. and Roman, C. 2001. An application of DEA to measure the efficiency of
Spanish airports prior to privatization. Journal of Air Transport Management 7(3),
pp. 149-157.
Martin, J.C. and Roman, C. 2006. A benchmarking analysis of Spanish commercial
airports: A comparison between SMOP and DEA ranking methods. Networks and
Spatial Economics 6(2), pp. 111-134.
Martin, J.C., Roman, C., and Voltes-Dorta, A. 2009. A Stochastic Frontier Analysis to
estimate the relative efficiency of Spanish airports. Journal of Productivity Analysis
31(3), pp. 163-176.
Maylor, H. and Blackmon, K. 2005. Researching Business and Management, New
York: Palgrave Macmillan.
McKenzie, G. 1997. Understanding Social Research: Perspectives on Methodology
and Practice. London: Sage.
Mentzer, J.T. and Kahn, K.B. 1995. A framework of logistics research, Journal of
Business Logistics 16(1), pp. 231-250.
Miller, D. 1983. Handbook of Research Design and Social Measurement, 4th ed.
London: Longman.
Mon, D.L., Cheng, C.H. and Lin, J.C. 1994. Evaluating weapon system using fuzzy
hierarchy process based on entropy weight. Fuzzy Sets and Systems. 62(2), pp. 127-
134.
Morgan, P. 1995. The privatisation and the welfare state: a case for back to the future.
In: Morgan, P. ed. Privatisation and the welfare state: Implication for consumers
and the workforce. New York: Dartmouth.
Murillo-Melchor, C. 1999. An analysis of technical efficiency and productivity changes
in Spanish airports using the Malmquist index. International Journal of Transport
Economics 26(2), pp. 271-292.
Näslund, D. 2002. Logistics need qualitative research-especially action research,
International Journal of Physical Distribution & Logistics Management 32(5), pp.
321-338.
Neufville, R. 1999. Airport privatization: issues for the United States. Transportation
Research Record: Journal of the Transportation Research Board 1662, pp. 24-31.
Önü, S. and Soner, S. 2008. Transshipment site selection using AHP and TOPSIS
approaches under fuzzy environment. Waste Management 28 (9), pp. 1552-1559.
235
Bibliography
Oum TH, Yu C. 1994. Economic efficiency of railways and implications for public
policy. Journal of Transport Economics & Policy, 28(2), pp. 121–38.
Oum, T.H., Yu, C., and Fu, X. 2003. A comparative analysis of productivity
performance of the world’s major airports: summary report of the ATRS global
airport benchmarking research report 2002. Journal of Air Transport Management
9(5), pp. 285-297.
Oum, T.H. and Yu, C. 2004. Measuring airports’ operating efficiency: a summary of
the 2003 ATRS global airport benchmarking report. Transportation Research Part E
40(6), pp. 515-532.
Oum, T.H., Adler, N. and Yu, C. 2006. Privatization, corporatization, ownership forms
and their effect on the performance of the world’s major airports. Journal of Air
Transport Management 12(3), pp. 109-121.
Oum, T.H., Yan, J. and Yu, C. 2008. Ownership forms matter for airport efficiency: A
stochastic frontier investigation of worldwide airports. Journal of Urban Economics
64(2), pp. 422-435.
Ozcan, Y. A. and McCue, M. J. 1996. Development of a Financial Performance Index
for Hospitals: DEA Approach. The journal of the operational research society 47(1),
pp. 18-26.
Pacheco, R.R. and Fernandes, E. 2003. Managerial efficiency of Brazilian airports.
Transportation Research Part A 37(8), pp. 667-680.
Pallis, A.A., Vistounis, T.K. and De Langen, P.W. 2010. Port Economics, Policy and
Management: Review of an Emerging Research Field. Transport Review 30(1), pp.
115-161.
Parker, D. 1999. The performance of BAA before and after privatisation: A DEA study.
Journal of Transport Economics and Policy 33(2), pp. 133-146.
Pan, J. and Rahman, S. 1998. Multiattribute utility analysis with imprecise information:
an enhanced decision support technique for the evaluation of electric generation
expansion strategies. Electric power systems research 46(1), pp. 101-109.
Pathomsiri, S., Haghani, A. Dresner, M., and Windle, R.J. 2008. Impact of undesirable
outputs on the productivity of US airports. Transportation Research Part E 44(2), pp.
235-259
Pels E, Nijkamp, P. and Rietveld, P. 2001. Relative efficiency of European airports.
Transport Policy 8(3), pp. 183-192.
Pels E, Nijkamp, P. and Rietveld, P. 2003. Inefficiencies and scale economies of
European airports operations. Transportation Research Part E 39(5), pp. 341-361.
Peng, J.L. and Zhan, C.X. 2011. A case study on evaluation of airport logistics
competitiveness based on AHP. Advanced material research 159(2), pp. 307-312.
Perelman, S. and Serebrisky, T. 2012. Measuring the technical efficiency of airports in
Latin America. Utilities policy 22(1), pp. 1-7.
Pirie, M. 1985. Privatisation in Theory and Practice. London: Adam Smith Institute.
Polit, D. F.,Beck, C. T., Hungler, B. P. 2001. Essentials of nursing research: methods,
appraisl and utilization, 5th ed. Philadelphia: Lippincott Williams & Wilkins.
236
Bibliography
Pöyhönen, M.A., Hämäläinen, R.P., and Salo, A.A. 1997. An experiment on the
numerical modelling of verbal ratio statement. Journal of Multi-Criteria decision
analysis 6(1), pp. 1-10.
Ramanathan, R. 2001. A note on the use of the analytic hierarchy process for
environmental impact assessment. Journal of environmental management 63(1), pp.
27-35.
Ramanathan, R. 2006. Data envelopment analysis for weight derivation and
aggregation in the analytic hierarchy process. Computers and Operations Research
33 (5), pp. 1289-1307.
Reed, M. 1997. In praise of duality and dualism: rethinking agency and structure in
organisational analysis. Organisational Studies 18, pp. 21-24.
Robson, C. 1993. Real world research: A resource for social scientist and practitioner-
researchers. Oxford: Blackwell Publishers.
Robson,C. 2002. Real World Research, 2nd ed. Oxford: Blackwell.
Roy, B. 1971. Problems and methods with multiple objective functions. Mathematical
programming 1(1), pp. 239-266.
Roy, B. 1991. The outranking approach and the foundations of ELECTRE methods.
Theory and decision 31(1), pp. 49-73.
Roy, B. 2005. Paradigms and challenges. In Figueira, J., Greco, S., and Ehrgott, M. ed.
2005. Multiple criteria decision analysis. London: Springer.
Russell, R. 1985. Measures of technical efficiency. Journal of economic theory 35, pp.
109-126.
Saen, R. F., Memariani, A., and Lotfi, F. H. 2005. Determining relative efficiency of
slightly non-homogeneous decision making units by data envelopment analysis: A
case study in IROST. Applied Mathematics and Computation 165(2), pp. 313-328.
Saaty, R.W. 2003. Decision making in complex environment. Pittsburgh: Creative
Decision Foundation.
Saaty, T.L. 1978. Exploring the interface between hierarchy, multiple objective and
fuzzy sets. Fuzzy sets and theory 1, pp. 57-68.
Saaty, T.L. 1980. The Analytic Hierarchy Process. New York: McGraw-Hill.
Saaty, T.L. 1987. Rank generation, preservation and reversal in the analytic hierarchy
decision process. Decision science 18, pp. 157-177.
Saaty, T.L. 1990. How to make a decision: the analytic hierarchy process. European
Journal of Operational Research 48(1), pp. 9-26.
Saaty, T.L. 1996. Decision Making with Dependence and Feedback: The Analytic
Network Process. Pittsburgh: RWS publication.
Saaty, T.L. 2008. Decision making with the analytic hierarchy process. Journal of
service science 1(1), pp. 83-98.
Saaty, T.L. and Vargas, L.G. 1987. Uncertainty and rank order in the analytic hierarchy
process. European Journal of Operational Research 32(1), pp. 107-117.
237
Bibliography
Salo, A.A. and Hämäläinen, R.P. 1993. On the measurement of preferences in the
analytic hierarchy process. Research report A47 system analysis laboratory. Helsinki:
Helsinki University.
Sarkis, J. 2000. An analysis of the operational efficiency of major airports in the United
States. Journal of Operations Management 18(3), pp. 335-351.
Sarkis, J. and Talluri, S. 2004. Performance based clustering for benchmarking of US
airports, Transportation Research Part A 38(5), pp. 329-346.
Saunders, M., Lewis, P. and Thornhill, A. 2000. Research Methods for Business
Students. 3th ed. London: Prentice Hall.
Sayer, A. 2000. Realism and Social Science. London: Sage.
Schoner, B. and Wedley, W.C. 1989. Alternative scales in AHP. In: Lockett, A.G. and
Islei, G. ed. Improving decisions in organizations LNEM 335. Berlin: Springer.
Scotti, D., Malighetti, P. Martini, G. and Volta, N. 2012.The impact of airport
competition efficiency: A stochastic frontier analysis applied to Italian airports.
Journal of Air Transport Management 22, pp. 9-15.
Shahroudi, K., Taleghani, M., and Mohammadi, G. 2012. Application of two-stage
DEA technique for efficiencies measuring of private insurance companies in Iran.
International journal of applied operational research 1(3), pp. 91-104
Shearman, P. 1992. Air transport: Strategic issues in planning and development.
London: Pitman Publishing.
Sinuany-Stern, Z., Mehrez, A., and Hadad, Y. 2000. An AHP/DEA methodology for
ranking decision making units. International Transactions in Operational Research
7(2), pp. 109-124.
Smith, P. K. 2000. Philosophy of science and its relevance for the social science. In:
Burton, D. ed. Research Training for Social Scientist. London: SAGE Publication.
Spens, K.M. and Kovács, G. 2005. A content analysis of research approaches in
logistics research, International Journal of Physical Distribution and Logistics
Management. 36(5), pp. 374-390.
Starkie, D. 2008. Aviation market. Surrey: Ashgate.
Takamura, Y. & Tone, K. 2003. A Comparative Site Evaluation Study for Relocating
Japanese Government Agencies Out of Tokyo. Socio-Economic Planning Sciences
37(2), pp. 85-102.
Timmermanns, D. 1994. The roles of experience and domain of expertise in using
numerical and verbal probability terms in medical decisions. Medical decision
making 14, pp. 146-156.
Thompson, G. 1990. The political economy of the new right. London: Printer.
238
Bibliography
Tovar, B. and Martin-Cejas, R.R. 2010. Technical efficiency and productivity changes
in Spanish airports: A parametric distance function approach. Transportation
Triantaphyllou, E.V. 2000. Multi-criteria decision-making methods: A comparative
study. London: Springer.
Triantaphyllou, E.V., Loostma, F.A., Pardalos, P.M., and Mann, S.H. 1994. On the
evaluation and application of different scales for quantifying pairwise comparison in
Fuzzy sets. Journal of Multi-Criteria Decision Analysis. 3(1), pp. 133-155.
Tsai, M. C. and Su, Y. S. 2002. Political risk assessment on air logistics hub
developments in Taiwan. Journal of Air Transport Management 8(6), pp. 373-380.
Tsekeris, T. 2011 Greek airports: Efficiency measurement and analysis of determinants.
Tsionas, E. 2002. Stochastic frontier models with random coefficients. Journal of
Applied Econometrics 17(2), pp. 127-147.
Turner, J. R. 2005. The role of pilot studies in reducing risk on projects and
programmes. International Journal of Project Management 23(1), pp. 1-6
Ülkü, T. 2009. Efficiency of German airports and influencing factors, MSc Thesis.
Humboldt University.
Varian, H. 2005. Bootstrap Tutorial. Mathematica Journal 9 (4), pp. 768-75.
Vasigh, B. and Gorjidooz, J. 2006. Productivity analysis of public and private airports:
a casual investigation. Journal of Air Transportation 11(3), pp. 144-163.
Vickers, J. and Yarrow, G. 1988. Privatization: An economic analysis. Boston: MIT
press.
Vickers, J. and Yarrow, G. 1991. Economic perspectives on privatization. Journal
economic perspectives 5, pp. 111-132.
Wagner, S.M. and Kemmerling, R. 2010. Handling nonresponse in logistics research,
Journal of Business Logistics. 31(2), pp. 357-381.
Wanke, P.F 2012a. Efficiency of Brazil’s airports: Evidence from bootstrapped DEA
and FDH estimates. Journal of Air Transport Management 23, pp. 47-53.
Wanke, P.F 2012b. Capacity shortfall and efficiency in Brazilian airports: Evidence
from bootstrapped DEA estimates. Socio-Economic planning sciences 46(3), pp.
216-229.
Wang R.-T., Ho, C.-T. Feng, C.-M., and Yang, Y.-K. 2004. A comparative analysis of
the operational performance of Taiwan’s major airports. Journal of Air Transport
Wang, Y.M., Liu, J., and Elhag, M.S. 2008. An integrated AHP-DEA methodology for
bridge risk assessment. Computer & Industrial Engineering 54(3), pp. 513-525.
Wells, A. T. and Young, S. B. 2004. Airport planning and management. 5th ed. London:
McGraw-Hill.
Wenseveen, J. G. 2007. Air transportation: a management perspective sixth ed. Surrey:
Ashgate.
239
Bibliography
Woo, S.-H., and Pettit, S.J. and Kwak, D.-K. 2010. Trends and themes in port research
since 1980– a decadal approach. International Association of Maritime Economists
Conference 2010, Lisbon.
Wragg, D. 2009. The world’s top 500 airports. Somerset: Haynes publishing.
Xing, T.P. and Tseng, G.X. 2002. Accessing efficiency of criminal investigation and
crime prevention in Taiwan police department: an application of DEA and AHP.
Journal of Information, Technology and Society 1, pp. 33-56.
Yang, H.-H. 2010. Measuring the efficiencies of Asia-Pacific international airports-
Parametric and non-parametric evidence. Computer & Industrial Engineering 59(4),
pp. 697–702.
Yang, X and Houng, Y. U. 2010. Deregulatory reform of China's airports. East Asia
policy 2(2), pp. 62-72.
Yang,T. and Kuo, C. 2003. A hierarchical AHP/DEA methodology for the facilities
layout design problem. European Journal of Operational Research, 147 (1), pp.
128–136
Yang, X., Tok, S. K., and Su, F. 2008.The privatization and commercialization of
China’s airports. Journal of Air Transport Management 14(5), pp. 243-251.
Yeh, C. H., Deng, H. and Pan, H. 1999. Multi-Criteria Analysis for Dredger
Dispatching under Uncertainty. Journal of the Operational Research Society 50, pp.
35-43.
Yeh C.H., Deng H., and ChangY. 2000. Fuzzy multicriteria analysis for performance
evaluation of bus companies. European Journal of Operational Research. 126(3), pp.
459-473.
Yoo, K. E. and Choi, Y. C. 2006. Analytic hierarchy process approach for identifying
relative importance of factors to improve passenger security checks at airports.
Journal of Air Transport Management 12(3), pp.135-142.
Yoshida, Y. 2004. Endogenous-weight TFP measurement: Methodology and its
Application to Japanese-airport benchmarking. Transportation Research Part E
40(2), pp. 151-182.
Yoshida, Y and Fujimoto, H. 2004. Japanese-airport benchmarking with the DEA and
endogenous-weight TFP methods Testing the criticism of overinvestment in Japanese
regional airports, Transportation Research Part E 40(6), pp. 533-546.
Yu, M.-M. 2004. Measuring physical efficiency of domestic airports in Taiwan with
undesirable outputs and environmental factors. Journal of Air Transport
Yu, M.-M., Hsu, S.-H., Chang, C.-C., and Lee, D.-H. 2008. Productivity growth of
Taiwan’s major domestic airports in the presence of aircraft noise. Transportation
Yu, M.-M. 2010. Assessment of airport performance using the SBM-NDEA model.
Omega 38(6), pp. 440-452.
Zeleny, M. 1983. Multiple criteria decision making. New York: McGraw Hill.
240
Bibliography
Zhang, A., Yuen, A., 2008. Airport policy and performance in Mainland China and
Hong Kong. In: Winston, C., de Rus, G. ed. Aviation Infrastructure Performance: A
Study in Comparative Political Economy. Washington, D.C.: Brookings Institution
Press.
Zhang, B., Wang, J., Liu, C. and Zhao, Y. 2012. Evaluating the technical efficiency of
Chinese airport airside activities. Journal of Air Transport Management 20, pp. 23-
27.
Zhang, Y. and Bartels, R. 1998. The effect of sample size on the mean efficiency in
DEA with an application to electricity distribution in Australia, Sweden and New
Zealand. Journal of productivity 9(3), pp.187-204.
Zhu, J. 2009. Quantitative models for performance evaluation and benchmarking, 2nd
ed. New York : Springer.
241
Appendix III: Relative weighted input and output variables scores
Table 1: Relative weighted input and output scores (scale 𝑒 !!! )
!!!
𝑒 weights 0.0124 0.1153 0.0825 0.0971 0.0490 0.6436 0.2670 0.1644 0.0628 0.5057
Employees Gates Runways Terminal Area Runway Expenditure Passengers Freight and Mail Movement Revenues
Amsterdam (AMS) 0.1777 4.1054 8.25 4.1586 3.8864 34.2657 18.8854 7.5803 5.0110 18.6599
Barcelona (BCN) 0.0392 4.4111 4.125 1.0904 3.4144 10.0944 12.0281 0.5040 3.6065 10.1082
Frankfurt (FRA) 1.2400 7.5993 4.125 5.6208 4.7922 64.3600 21.2892 10.2079 5.4495 18.6801
Istanbul (IST) 0.1206 1.3976 4.125 2.2378 3.3150 5.9810 11.3611 3.7049 2.8553 4.4961
London (LGW) 0.1506 4.6731 1.375 1.4229 3.9727 20.3177 13.6231 0.5433 2.9577 13.8381
London (LHR) 0.3801 11.5300 2.75 2.5631 4.5286 63.2802 26.7000 7.1865 5.3700 50.5700
Madrid (MAD) 0.0549 3.3193 5.50 2.1078 4.6281 20.1884 20.2456 1.5907 5.2696 20.2167
Munich (MUC) 0.5099 8.7349 2.75 3.2179 4.7922 36.9631 13.7577 1.3271 4.8495 21.0410
Paris (CDG) 0.2658 5.4156 5.50 3.8123 4.1381 37.7061 24.2389 9.8640 6.2800 33.3936
Paris (ORY) 0.2277 4.4547 4.125 2.6102 3.7415 17.3008 10.4362 0.6769 2.5820 14.3054
Rome (FCO) 0.2259 3.7560 5.50 2.0024 4.4052 16.6058 14.0265 0.6645 3.8888 19.0985
Zurich (ZRH) 0.0864 2.9262 4.125 0.9739 3.7942 13.9626 10.9494 1.8745 3.0849 11.6485
Bangkok (BKK) 0.2236 5.2410 2.75 3.9557 4.6125 15.0198 18.6871 6.2469 3.4937 20.0504
Beijing (PEK) 0.1354 5.2410 2.75 9.7100 4.3130 17.9825 22.2731 6.6133 4.8198 10.0066
Guangzhou (CAN) 0.2399 3.2319 2.75 2.2483 4.4328 8.1924 13.3130 4.6522 2.3682 4.2477
Hong Kong (HKG) 0.0779 4.6295 2.75 4.9885 4.5526 14.7831 18.9929 16.4400 3.3205 12.9456
Incheon (ICN) 0.0643 3.9307 4.125 4.2156 4.5921 13.7892 11.9345 11.7194 2.3682 13.6289
Kuala Lumpur (KUL) 0.1087 4.6295 2.75 3.3683 4.9000 8.9276 10.9614 3.2275 2.3522 6.4607
Osaka (KIX) 0.0267 2.2711 2.75 2.3186 4.4927 15.9380 6.3764 4.0932 1.4976 18.0399
Tokyo (HRT) 0.0496 3.7997 2.75 5.5485 3.8936 38.5721 13.0020 10.1563 2.1687 33.2237
Shanghai (PVG) 0.4437 4.2801 4.125 5.7895 4.4723 7.3754 11.2429 12.5864 2.9810 15.4322
Singapore (SIN) 0.0962 4.4547 4.125 7.3283 4.2926 13.9974 9.1090 2.0102 2.0787 11.3934
Shenzhen (SZX) 0.2755 2.4020 1.375 1.0680 4.0734 3.3207 8.5213 2.8917 2.1083 4.5156
Sydney (SYD) 0.0211 2.8388 4.125 2.7225 3.5678 4.7444 13.0999 2.2726 3.3538 8.6798
250
Table 2: Relative weighted input and output scores (scale 2!!! )
2!!! weights 0.0051 0.1108 0.0695 0.0869 0.0336 0.6941 0.2991 0.1527 0.0402 0.5079
Amsterdam (AMS) 0.0731 3.9451 6.9500 3.7218 2.6650 36.9544 21.1559 7.0409 3.2077 18.7411
Barcelona (BCN) 0.0161 4.2390 3.4750 0.9759 2.3413 10.8864 13.4741 0.4682 2.3086 10.1522
Frankfurt (FRA) 0.5100 7.3027 3.4750 5.0304 3.2861 69.4100 23.8487 9.4814 3.4884 18.7613
Istanbul (IST) 0.0496 1.3430 3.4750 2.0027 2.2731 6.4503 12.7270 3.4412 1.8278 4.5157
London (LGW) 0.0620 4.4907 1.1583 1.2734 2.7242 21.9119 15.2610 0.5047 1.8933 13.8983
London (LHR) 0.1563 11.0800 2.3167 2.2939 3.1053 68.2455 29.9100 6.6751 3.4375 50.7900
Madrid (MAD) 0.0226 3.1897 4.6333 1.8864 3.1735 21.7724 22.6796 1.4775 3.3732 20.3047
Munich (MUC) 0.2097 8.3940 2.3167 2.8799 3.2861 39.8634 15.4117 1.2327 3.1043 21.1325
Paris (CDG) 0.1093 5.2043 4.6333 3.4118 2.8375 40.6647 27.1530 9.1620 4.0200 33.5389
Paris (ORY) 0.0936 4.2809 3.4750 2.3360 2.5656 18.6583 11.6908 0.6288 1.6528 14.3676
Rome (FCO) 0.0929 3.6094 4.6333 1.7921 3.0207 17.9088 15.7128 0.6172 2.4893 19.1815
Zurich (ZRH) 0.0355 2.8120 3.4750 0.8716 2.6017 15.0582 12.2658 1.7411 1.9747 11.6992
Bangkok (BKK) 0.0920 5.0364 2.3167 3.5401 3.1628 16.1984 20.9338 5.8023 2.2364 20.1376
Beijing (PEK) 0.0557 5.0364 2.3167 8.6900 2.9575 19.3935 24.9509 6.1426 3.0853 10.0501
Guangzhou (CAN) 0.0987 3.1058 2.3167 2.0122 3.0396 8.8352 14.9135 4.3211 1.5159 4.2661
Hong Kong (HKG) 0.0321 4.4488 2.3167 4.4645 3.1218 15.9431 21.2763 15.2700 2.1256 13.0019
Incheon (ICN) 0.0264 3.7773 3.4750 3.7728 3.1489 14.8711 13.3693 10.8853 1.5159 13.6882
Kuala Lumpur (KUL) 0.0447 4.4488 2.3167 3.0145 3.3600 9.6281 12.2792 2.9978 1.5057 6.4889
Osaka (KIX) 0.0110 2.1824 2.3167 2.0750 3.0807 17.1886 7.1430 3.8019 0.9587 18.1184
Tokyo (HRT) 0.0204 3.6514 2.3167 4.9656 2.6699 41.5986 14.5652 9.4335 1.3882 33.3682
Shanghai (PVG) 0.1825 4.1130 3.4750 5.1813 3.0667 7.9541 12.5945 11.6907 1.9082 15.4993
Singapore (SIN) 0.0396 4.2809 3.4750 6.5585 2.9435 15.0957 10.2042 1.8671 1.3307 11.4430
Shenzhen (SZX) 0.1133 2.3083 1.1583 0.9558 2.7932 3.5813 9.5458 2.6859 1.3496 4.5352
Sydney (SYD) 0.0087 2.7280 3.4750 2.4365 2.4465 5.1166 14.6749 2.1109 2.1469 8.7175
251
Table 3: Relative weighted input and output scores (scale 9/9-9/1)
9/9-9/1 weights 0.0522 0.1103 0.1013 0.1066 0.0825 0.5436 0.1971 0.1719 0.1286 0.4976
Amsterdam (AMS) 0.7480 3.9278 10.1300 4.5657 6.5439 28.9413 13.9409 7.9263 10.2610 27.5770
Barcelona (BCN) 0.1650 4.2201 5.0650 1.1971 5.7486 8.5236 8.8794 0.5277 7.3855 5.5781
Frankfurt (FRA) 5.2200 7.2699 5.0650 6.1711 8.0685 54.3600 15.7148 10.6733 11.1599 36.2153
Istanbul (IST) 0.5074 1.3368 5.0650 2.4571 5.5811 5.0500 8.3866 3.8746 5.8474 4.8864
London (LGW) 0.6342 4.4705 1.6887 1.5617 6.6891 17.1615 10.0560 0.5673 6.0571 14.6991
London (LHR) 1.5999 11.0300 3.3763 2.8142 7.6247 53.4468 19.7100 7.5137 10.9966 49.7600
Madrid (MAD) 0.2312 3.1755 6.7537 2.3143 7.7921 17.0527 14.9461 1.6640 10.7908 11.1512
Munich (MUC) 2.1465 8.3563 3.3763 3.5327 8.0685 31.2189 10.1566 1.3872 9.9305 23.5464
Paris (CDG) 1.1192 5.1808 6.7537 4.1851 6.9671 31.8495 17.8927 10.3140 12.8600 30.4382
Paris (ORY) 0.9584 4.2620 5.0650 2.8654 6.2997 14.6120 7.7046 0.7082 5.2880 12.6042
Rome (FCO) 0.9511 3.5936 6.7537 2.1981 7.4168 14.0249 10.3537 0.6945 7.9629 16.3312
Zurich (ZRH) 0.3638 2.7994 5.0650 1.0692 6.3880 11.7907 8.0831 1.9597 6.3168 13.5895
Bangkok (BKK) 0.9412 5.0142 3.3763 4.3429 7.7657 12.6876 13.7950 6.5322 7.1540 12.6241
Beijing (PEK) 0.5700 5.0142 3.3763 10.6600 7.2617 15.1882 16.4421 6.9155 9.8701 9.6634
Guangzhou (CAN) 1.0101 3.0917 3.3763 2.4678 7.4630 6.9200 9.8274 4.8648 4.8495 6.5186
Hong Kong (HKG) 0.3278 4.4285 3.3763 5.4760 7.6651 12.4865 14.0197 17.1900 6.7991 19.2870
Incheon (ICN) 0.2704 3.7601 5.0650 4.6286 7.7319 11.6493 8.8104 12.2548 4.8495 16.7592
Kuala Lumpur (KUL) 0.4578 4.4285 3.3763 3.6980 8.2500 7.5397 8.0910 3.3744 4.8174 5.0307
Osaka (KIX) 0.1128 2.1729 3.3763 2.5456 7.5644 13.4595 4.7067 4.2803 3.0671 16.5054
Tokyo (HRT) 0.2088 3.6355 3.3763 6.0911 6.5555 32.5779 9.5988 10.6200 4.4406 31.5031
Shanghai (PVG) 1.8682 4.0943 5.0650 6.3555 7.5298 6.2297 8.2999 13.1607 6.1046 8.2950
Singapore (SIN) 0.4051 4.2620 5.0650 8.0451 7.2270 11.8233 6.7251 2.1023 4.2567 15.8635
Shenzhen (SZX) 1.1599 2.2975 1.6887 1.1726 6.8582 2.8050 6.2914 3.0237 4.3171 3.7370
Sydney (SYD) 0.0887 2.7156 5.0650 2.9891 6.0068 40.0633 9.6697 2.3757 6.8672 13.3158
252
Table 4: Relative weighted input and output scores (scale 10/10-18/2)
10/10-18/2 weights 0.0417 0.1128 0.0986 0.1058 0.0786 0.5626 0.2111 0.1722 0.1140 0.5027
Amsterdam (AMS) 0.5976 4.0164 9.8600 4.5312 6.2342 29.9532 14.9315 7.9400 9.0964 18.5492
Barcelona (BCN) 0.1318 4.3155 4.9300 1.1881 5.4770 8.8239 9.5098 0.5279 6.5468 10.0482
Frankfurt (FRA) 4.1700 7.4345 4.9300 6.1245 7.6871 56.2600 16.8320 10.6922 9.8925 18.5692
Istanbul (IST) 0.4055 1.3672 4.9300 2.4383 5.3175 5.2282 8.9825 3.8807 5.1833 4.4695
London (LGW) 0.5065 4.5718 1.6433 1.5504 6.3726 17.7606 10.7710 0.5691 5.3690 13.7560
London (LHR) 1.2782 11.2800 3.2867 2.7928 7.2643 55.3161 21.1100 7.5275 9.7481 50.2700
Madrid (MAD) 0.1847 3.2473 6.5733 2.2967 7.4238 17.6476 16.0069 1.6662 9.5658 20.0968
Munich (MUC) 1.7147 8.5455 3.2867 3.5063 7.6871 32.3112 10.8774 1.3901 8.8033 20.9162
Paris (CDG) 0.8940 5.2982 6.5733 4.1539 6.6378 32.9606 19.1641 10.3320 11.4000 33.1955
Paris (ORY) 0.7656 4.3581 4.9300 2.8440 6.0017 15.1234 8.2512 0.7091 4.6871 14.2205
Rome (FCO) 0.7596 3.6746 6.5733 2.1818 7.0663 14.5159 11.0899 0.6960 7.0593 18.9852
Zurich (ZRH) 0.2906 2.8628 4.9300 1.0612 6.0862 12.2054 8.6570 1.9634 5.5999 11.5794
Bangkok (BKK) 0.7519 5.1273 3.2867 4.3101 7.3988 13.1295 14.7747 6.5433 6.3421 19.9315
Beijing (PEK) 0.4553 5.1273 3.2867 10.5800 6.9184 15.7193 17.6099 6.9270 8.7493 9.9472
Guangzhou (CAN) 0.8068 3.1618 3.2867 2.4498 7.1105 7.1613 10.5257 4.8729 4.2989 4.2225
Hong Kong (HKG) 0.2621 4.5291 3.2867 5.4355 7.3027 12.9226 15.0165 17.2200 6.0277 12.8688
Incheon (ICN) 0.2162 3.8455 4.9300 4.5933 7.3661 12.0537 9.4358 12.2754 4.2989 13.5481
Kuala Lumpur (KUL) 0.3657 4.5291 3.2867 3.6701 7.8600 7.8041 8.6664 3.3807 4.2699 6.4224
Osaka (KIX) 0.0899 2.2218 3.2867 2.5263 7.2066 13.9321 5.0414 4.2874 2.7186 17.9329
Tokyo (HRT) 0.1668 3.7173 3.2867 6.0456 6.2457 33.7176 10.2799 10.6382 3.9368 33.0266
Shanghai (PVG) 1.4923 4.1872 4.9300 6.3082 7.1739 6.4472 8.8890 13.1836 5.4114 15.3407
Singapore (SIN) 0.3235 4.3581 4.9300 7.9849 6.8857 12.2358 7.2019 2.1055 3.7735 11.3258
Shenzhen (SZX) 0.9264 2.3500 1.6433 1.1636 6.5340 2.9028 6.7373 3.0289 3.8272 4.4888
Sydney (SYD) 0.0709 2.7772 4.9300 2.9664 5.7230 4.1473 10.3573 2.3804 6.0881 8.6283
253
Table 5: Relative weighted input and output scores (scale ø mapping)
Ø mapping weights 0.0613 0.1087 0.1009 0.1050 0.0886 0.5355 0.1919 0.1711 0.1354 0.5016
Amsterdam (AMS) 0.8785 3.8704 10.090 4.4970 7.0273 28.5104 13.5735 7.8893 10.8040 18.5087
Barcelona (BCN) 0.1938 4.1586 5.0450 1.1792 6.1738 8.3989 8.6449 0.5246 7.7758 10.0262
Frankfurt (FRA) 6.1300 7.1643 5.0450 6.0781 8.6651 53.5500 15.3011 10.6239 11.7495 18.5286
Istanbul (IST) 0.5961 1.3176 5.0450 2.4199 5.9941 4.9764 8.1655 3.8559 6.1563 4.4597
London (LGW) 0.7446 4.4056 1.6817 1.5387 7.1833 16.9051 9.7913 0.5655 6.3769 13.7259
London (LHR) 1.8789 10.8700 3.3633 2.7716 8.1885 52.6516 19.1900 7.4794 11.5780 50.1600
Madrid (MAD) 0.2715 3.1293 6.7267 2.2793 8.3683 16.7975 14.5510 1.6556 11.3615 20.0528
Munich (MUC) 2.5207 8.2349 3.3633 3.4797 8.6651 30.7548 9.8880 1.3812 10.4558 20.8704
Paris (CDG) 1.3142 5.1056 6.7267 4.1225 7.4823 31.3729 17.4211 10.2660 13.5400 33.1229
Paris (ORY) 1.1254 4.1997 5.0450 2.8225 6.7652 14.3949 7.5007 0.7045 5.5670 14.1894
Rome (FCO) 1.1166 3.5410 6.7267 2.1653 7.9653 13.8167 10.0812 0.6916 8.3844 18.9436
Zurich (ZRH) 0.4272 2.7587 5.0450 1.0531 6.8606 11.6175 7.8696 1.9509 6.6511 11.5540
Bangkok (BKK) 1.1053 4.9410 3.3633 4.2775 8.3401 12.4971 13.4309 6.5015 7.5326 19.8879
Beijing (PEK) 0.6693 4.9410 3.3633 10.5000 7.7986 14.9621 16.0083 6.8828 10.3917 9.9254
Guangzhou (CAN) 1.1861 3.0469 3.3633 2.4313 8.0152 6.8164 9.5684 4.8418 5.1059 4.2132
Hong Kong (HKG) 0.3853 4.3645 3.3633 5.3944 8.2318 12.3001 13.6507 17.1100 7.1593 12.8406
Incheon (ICN) 0.3178 3.7057 5.0450 4.5586 8.3032 11.4731 8.5776 12.1970 5.1059 13.5184
Kuala Lumpur (KUL) 0.5375 4.3645 3.3633 3.6424 8.8600 7.4281 7.8782 3.3591 5.0715 6.4084
Osaka (KIX) 0.1322 2.1411 3.3633 2.5072 8.1235 13.2610 4.5829 4.2600 3.2290 17.8937
Tokyo (HRT) 0.2453 3.5822 3.3633 5.9999 7.0403 32.0934 9.3449 10.5702 4.6758 32.9543
Shanghai (PVG) 2.1937 4.0351 5.0450 6.2605 8.0866 6.1366 8.0805 13.0994 6.4272 15.3071
Singapore (SIN) 0.4755 4.1997 5.0450 7.9245 7.7617 11.6464 6.5469 2.0921 4.4819 11.3010
Shenzhen (SZX) 1.3618 2.2645 1.6817 1.1548 7.3653 2.7630 6.1245 3.0095 4.5457 4.4790
Sydney (SYD) 0.1042 2.6763 5.0450 2.9440 6.4511 3.9475 9.4153 2.3652 7.2310 8.6094
254
Appendix IV: The weight distribution of variables in AHP/DEA model
Table 1: The weight distribution of variables in AHP/DEA model (1-9 and 𝑒 !!! )
1-9 𝑒 !!!
Number of Gates
Number of Gates
Size of Terminal
Freight and Mail
Size of Terminal
Freight and Mail

Expenditure
Expenditure
Operational
Operational
Employees
Employees
Passengers
Passengers
Number of
Number of
Number of
Amount of
Number of
Number of
Number of
Amount of
Movement
Movement
Length of
Length of
Revenues
Revenues
Runways
Runways
Runway
Runway
Aircraft
Aircraft
Total
Total
Area
Area
DMUs
1 0.1511 0.5355 0 0 0.3134 0 0.0499 0.2373 0.7128 0 0.1546 0.5321 0 0.0112 0.3021 0 0 0.2473 0.7527 0
2 0.4963 0 0 0.2889 0 0.2148 1 0 0 0 0.4693 0 0 0.5307 0 0 1 0 0 0
3 0 0 0.2924 0 0 0 0 0.2932 0.7068 0 0 0 0.2924 0 0 0 0 0.2932 0.7068 0
4 0.3089 0.6911 0 0 0 0 1 0 0 0 0 0.4191 0.5809 0 0 0 0.9911 0.0089 0 0
5 0.6148 0 0.3514 0.0338 0 0 0.1713 0 0.6692 0.1595 0 0 0.0656 0.8236 0 0.1108 1 0 0 0
6 0 0 0.3657 0 0 0.6343 0 0.0642 0.3520 0.5839 0.1909 0 0.8091 0 0 0 0.3647 0.1174 0.5179 0
7 0.4417 0.5583 0 0 0 0 1 0 0 0 0.3004 0 0.3646 0 0.3351 0 1 0 0 0
8 0 0 0.3326 0.0722 0.3109 0.2843 0 0 0.9512 0.0488 0 0 0.4298 0.0617 0.3157 0.1928 0 0 1 0
9 0 0.3743 0.4029 0.0817 0.1411 0 1 0 0 0 0 0.3743 0.4029 0.0817 0.1411 0 1 0 0 0
10 0 0.1324 0.0934 0.2000 1.0823 0.7198 0 0 0 1 0 0.1324 0.0934 0.2000 1.0824 0.7198 0 0 0 1
11 0 0.0202 0 0.5056 0 0.4742 0.1768 0 0 0.8232 0 0.1086 0 0.1179 0 0.7736 0 0 0 1
12 0 0.1475 0.1966 0.6559 0 0 1 0 0 0 0 0 0 0.7319 0 0.2681 0.6306 0 0 0.3694
13 0 0.4659 0.4851 0 0 0.0489 0.7370 0 0 0.2630 0.1063 0 0.1187 0 0 0.7749 0 0 0 1
14 0.3470 0 0.6530 0 0 0 1 0 0 0 0.3470 0 0.6530 0 0 0 1 0 0 0
15 0 0.1742 0.1971 0.1402 0.1576 0.2190 1 0 0 0 0 0.1742 0.1971 0.1402 0.1576 0.2190 1 0 0 0
16 0 0 0 0.1233 0.6665 0.2102 0.9408 0.0592 0 0 0 0 0 0.1233 0.6665 0.2102 0.9408 0.0592 0 0
17 0.0714 0 0 0 0 0.6663 0 0.2626 0 0.7374 0.0714 0 0 0 0 0.6664 0 0.2626 0 0.7374
18 3.0677 0 7.2334 0 0 2.2682 1 0 0 0 3.0677 0 7.2334 0 0 2.2682 1 0 0 0
19 1 0 0 0 0 0 0 0.1262 0.1696 0.7042 1 0 0 0 0 0 0 0.1262 0.1696 0.7042
20 1 0 0 0 0 0 0.2565 0.7375 0 0.0060 1 0 0 0 0 0 0.2565 0.7375 0 0.0060
21 0 0 0 0 0 1 0.4175 0.5825 0 0 0 0.7500 0 0 0.1832 0.0668 0 0.5011 0.2130 0.2858
22 0.1306 0 0 0 0.3365 0.7957 0 0 0.0613 0.9387 0.1306 0 0 0 0.3365 0.7957 0 0 0.0613 0.9387
23 0 0 0 0.8066 0 0.1934 1 0 0 0 0 0 0 0.8066 0 0.1934 1 0 0 0
24 1 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 1 0 0 0
255
Table 2: The weight distribution of variables in AHP/DEA model (2!!! and 9/9-9/1)
2!!! 9/9-9/1
Number of Gates
Number of Gates
Size of Terminal
Freight and Mail
Size of Terminal
Freight and Mail

Expenditure
Expenditure
Operational
Operational
Employees
Employees
Passengers
Passengers
Number of
Number of
Number of
Amount of
Number of
Number of
Number of
Amount of
Movement
Movement
Length of
Length of
Revenues
Revenues
Runways
Runways
Runway
Runway
Aircraft
Aircraft
Total
Total
Area
Area
DMUs
1 0.1560 0.5449 0 0 0.2991 0 0 0.2459 0.7437 0.0104 0.1560 0.5449 0 0 0.2991 0 0 0.2459 0.7437 0.0104
2 0.4693 0 0 0.5307 0 0 1 0 0 0 0.4693 0 0 0.5307 0 0 1 0 0 0
3 0 0 0.2924 0 0 0 0 0.2932 0.7068 0 0 0 0.2924 0 0 0 0 0.2932 0.7068 0
4 0 1 0 0 0 0 0.7815 0 0 0.2185 0 0.2235 0 0 0.7141 0.0623 0.9677 0.0323 0 0
5 0.2729 0 0.2699 0.0090 0 0.4482 0 0 0.5514 0.4486 0 0.4156 0.5844 0 0 0 1 0 0 0
6 0.1909 0 0.8091 0 0 0 0.3647 0.1174 0.5179 0 0 0 0.8145 0.1855 0 0 1 0 0 0
7 0.3636 0 0.5807 0.0557 0 0 1.0000 0 0 0 0.4417 0.5583 0 0 0 0 1 0 0 0
8 0 0.1065 0.3621 0.0711 0.3129 0.1474 0 0 1 0 0 0.1065 0.3621 0.0711 0.3129 0.1474 0 0 1 0
9 0 0.3743 0.4029 0.0817 0.1411 0 1 0 0 0 0 0.3743 0.4029 0.0817 0.1411 0 1 0 0 0
10 0 0.1324 0.0934 0.2001 1.0823 0.7198 0 0 0 1 0 0.1324 0.0934 0.2001 1.0823 0.7198 0 0 0.0000 1
11 0 0.2747 0 0.3846 0 0.3407 0 0 0.0110 0.9890 0 0.3092 0 0.0758 0 0.6150 0 0 0.1422 0.8578
12 0.3615 0 0 0.6385 0 0 0.7353 0.2647 0 0 0.3615 0 0 0.6385 0 0 0.7353 0.2647 0 0
13 0 0.2887 0.0491 0.3218 0 0.3404 0.2468 0.1550 0 0.5982 0.3477 0 0.2253 0 0.0326 0.3944 0.3269 0 0 0.6731
14 0.1830 0 0.1536 0 0.6634 0 0 0 1 0 0.2725 0 0.7275 0 0 0 0.9283 0 0 0.0717
15 0 0.1742 0.1971 0.1402 0.1576 0.2190 1 0 0 0 0 0.1742 0.1971 0.1402 0.1576 0.2190 1 0 0 0
16 0 0 0 0.1233 0.6665 0.2102 0.9408 0.0592 0 0.0000 0.1267 0.6001 0.2732 0 0 0 1 0 0 0
17 0.0714 0 0 0 0 0.6664 0 0.2626 0 0.7374 0.0714 0 0 0 0 0.6664 0 0.2626 0 0.7374
18 3.0677 0 7.2334 0 0 2.2682 1 0 0 0 3.0677 0 7.2334 0 0 2.2682 1 0 0 0.0000
19 1 0 0 0 0 0 0 0.1262 0.1696 0.7042 1 0 0 0 0 0 0 0.1262 0.1696 0.7042
20 1 0 0 0 0 0 0.2565 0.7375 0 0.0060 1 0 0 0 0 0 0.2565 0.7375 0 0.0060
21 0 0 0 0 0 1 0.4175 0.5825 0 0 0 0 0 0 0 1 0.4175 0.5825 0 0
22 0.1306 0 0 0 0.3365 0.7957 0 0 0.0613 0.9387 0.1306 0 0 0 0.3365 0.7957 0 0 0.0613 0.9387
23 0 0 0 0.8066 0 0.1934 1 0 0 0 0 0.5550 0.3435 0.0931 0 0.0084 1 0 0 0
24 1 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 1 0 0 0
256
Table 3: The weight distribution of variables in AHP/DEA model (10/10-18/2 and ∅ mapping)
10/10-18/2 ∅ mapping
Number of Gates
Number of Gates
Size of Terminal
Freight and Mail
Size of Terminal
Freight and Mail

Expenditure
Expenditure
Operational
Operational
Employees
Employees
Passengers
Passengers
Number of
Number of
Number of
Amount of
Number of
Number of
Number of
Amount of
Movement
Movement
Length of
Length of
Revenues
Revenues
Runways
Runways
Runway
Runway
Aircraft
Aircraft
Total
Total
Area
Area
DMUs
1 0.1560 0.5449 0 0 0.2991 0 0 0.2459 0.7437 0.0104 0.1560 0.5449 0 0 0.2991 0 0 0.2459 0.7437 0.0104
2 0.4693 0 0 0.5307 0 0 1 0 0 0 0.4693 0 0 0.5307 0 0 1 0 0 0
3 0 0 0.2924 0 0 0 0 0.2932 0.7068 0 0 0 0.2924 0 0 0 0 0.2932 0.7068 0
4 0 0.2237 0 0.7454 0 0.0309 0.3918 0.3410 0.2671 0 0.3089 0.6911 0 0 0 0 1 0 0 0
5 0 0.4156 0.5844 0 0 0 1 0 0 0 0 0.4156 0.5844 0 0 0 1 0 0 0
6 0 0 0.4574 0 0 0.5426 0.5739 0.0329 0 0.3932 0 0 0.8145 0.1855 0 0 1 0 0 0
7 0.3636 0 0.5807 0.0557 0 0 1 0 0 0 0.3636 0 0.5807 0.0557 0 0 1 0 0 0
8 0 0 0.4298 0.0617 0.3157 0.1928 0 0 1 0 0 0 0.3147 0.0868 0.3049 0.2936 0 0 1 0
9 0 0.5601 0 0 0.4399 0 1 0 0 0 0.2664 0 0.1031 0.0495 0.5809 0 1 0 0 0
10 0 0.1324 0.0934 0.2001 1.0823 0.7198 0 0 0 1 0 0.1324 0.0934 0.2001 1.0823 0.7198 0 0 0 1
11 0 0.0202 0 0.5056 0 0.4742 0.1768 0 0 0.8232 0 0.3092 0 0.0758 0 0.6150 0 0 0.1422 0.8578
12 0 0.1475 0.1966 0.6559 0 0 1 0 0 0 0.0175 0.0094 0 0.2921 0 0.6810 0 0.1055 0.1225 0.7719
13 0.0545 0 0.0347 0.5333 0 0.3776 0.1748 0.1950 0 0.6302 0 0.5471 0.3507 0.0579 0.0443 0 0.8229 0 0 0.1771
14 0.3470 0 0.6530 0 0 0 1 0 0 0 0.3470 0 0.6530 0 0 0 1 0 0 0
15 0 0.1742 0.1971 0.1402 0.1576 0.2190 1 0 0 0 0 0.1742 0.1971 0.1402 0.1576 0.2190 1 0 0 0
16 0 0 0 0.1233 0.6665 0.2102 0.9408 0.0592 0 0 0.1267 0.6001 0.2732 0 0 0 1 0 0 0
17 0.0714 0 0 0 0 0.6664 0 0.2626 0 0.7374 0.0714 0 0 0 0 0.6664 0 0.2626 0 0.7374
18 3.0677 0 7.2334 0 0 2.2682 1 0 0 0 3.0677 0 7.2334 0 0 2.2682 1 0 0 0
19 1 0 0 0 0 0 0 0.1262 0.1696 0.7042 1 0 0 0 0 0 0 0.1262 0.1696 0.7042
20 1 0 0 0 0 0 0.2565 0.7375 0 0.0060 1 0 0 0 0 0 0.2565 0.7375 0 0.0060
21 0 0.7500 0 0 0.1832 0.0668 0 0.5011 0.2130 0.2858 0 0.9284 0 0 0 0.0716 0 0.6008 0.1021 0.2971
22 0.1306 0 0 0 0.3365 0.7957 0 0 0.0613 0.9387 0.1306 0 0 0 0.3365 0.7957 0 0.0000 0.0613 0.9387
23 0 0 0.8203 0 0 0.1797 1 0 0 0 0 0 0 0.8066 0 0.1934 1 0 0 0
24 1 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 1 0 0 0
257
Appendices
Appendix V: Upper and lower bounds of variables weight ratios: VS I

Table 1: 𝑒 !!! scale
Output Weight
Input Weight Ratio Upper Lower Upper Lower
Ratio
WI1/WI2 3.4666 0.0246 WO1/WO2 8.7298 0.1889
WI1/WI3 4.4920 0.0187 WO1/WO3 20.0869 0.2636
WI1/WI4 3.5136 0.0235 WO1/WO4 15.7794 0.0233
WI1/WI5 3.4009 0.0419 WO2/WO3 20.869 0.1889
WI1/WI6 1.6280 0.0009 WO2/WO4 3.5209 0.0233
WI2/WI3 4.4721 0.1129 WO3/WO4 2.742 0.002
WI2/WI4 6.8199 0.2730
WI2/WI5 12.3694 0.4699
WI2/WI6 2.2217 0.0113
WI3/WI4 15.0178 0.1547
WI3/WI5 5.5249 0.3473
WI3/WI6 4.6168 0.0087
WI4/WI5 4.7428 0.2091
WI4/WI6 3.4928 0.0093
WI5/WI6 1.4852 0.0067
Table 2: 2!!! scale

Output Weight
Ratio
WI1/WI2 0.0425 0.0112 WO1/WO2 20.1588 0.0992
WI1/WI3 0.0929 0.0245 WO1/WO3 65.3478 0.1574
WI1/WI4 0.1233 0.0123 WO1/WO4 56.1261 0.0076
WI1/WI5 0.2353 0.477 WO2/WO3 65.3478 0.0992
WI1/WI6 0.019 0.0004 WO2/WO4 8.2006 0.0076
WI2/WI3 4.167 0.459 WO3/WO4 6.0223 0.0002
WI2/WI4 14.5695 0.1532
WI2/WI5 31.9222 0.3364
WI2/WI6 4.7176 0.0033
WI3/WI4 44.0653 0.0647
WI3/WI5 11.0174 0.2346
WI3/WI6 12.09 0.0022
WI4/WI5 16.6049 0.1156
WI4/WI6 9.735 0.0024
WI5/WI6 2.6948 0.0017
259
Appendices
Table 3: 9/9-9/1scale
Output Weight
Ratio
WI1/WI2 1.4133 0.2666 WO1/WO2 1.8418 0.6215
WI1/WI3 1.4722 0.2058 WO1/WO3 3 0.6933
WI1/WI4 1.3904 0.2779 WO1/WO4 1.5208 0.1304
WI1/WI5 1.383 0.3261 WO2/WO3 3 0.641
WI1/WI6 0.3686 0.0359 WO2/WO4 0.9212 0.1304
WI2/WI3 1.6836 0.5146 WO3/WO4 0.6216 0.0635
WI2/WI4 1.7987 0.6664
WI2/WI5 2.2654 0.8051
WI2/WI6 0.4627 0.0821
WI3/WI4 2.7444 0.5265
WI3/WI5 2.0256 0.7052
WI3/WI6 0.6791 0.0788
WI4/WI5 1.9665 0.6166
WI4/WI6 0.5409 0.0835
WI5/WI6 0.3797 0.0741
Table 4: 10/10-18/2 scale

Output Weight
Ratio
WI1/WI2 1.6649 0.184 WO1/WO2 2.4753 0.4966
WI1/WI3 1.8174 0.1564 WO1/WO3 4 0.5757
WI1/WI4 1.6609 0.1869 WO1/WO4 2.2533 0.1029
WI1/WI5 1.652 0.2346 WO2/WO3 4 0.5036
WI1/WI6 0.4857 0.0203 WO2/WO4 1.1548 0.1029
WI2/WI3 1.956 0.3948 WO3/WO4 0.8231 0.037
WI2/WI4 2.2793 0.5787
WI2/WI5 3.0221 0.7336
WI2/WI6 0.6322 0.0596
WI3/WI4 3.404 0.4165
WI3/WI5 2.2196 0.6218
WI3/WI6 0.9389 0.0546
WI4/WI5 2.5286 0.5049
WI4/WI6 0.7828 0.0598
WI5/WI6 0.4957 0.0491
260
Appendices
Table 5: ø mapping scale

Output Weight
Ratio
WI1/WI2 1.3297 0.3737 WO1/WO2 1.6642 0.6744
WI1/WI3 1.391 0.3279 WO1/WO3 2.2361 0.7339
WI1/WI4 1.3251 0.3819 WO1/WO4 1.0423 0.1707
WI1/WI5 1.3224 0.4314 WO2/WO3 2.236 0.6813
WI1/WI6 0.3298 0.0554 WO2/WO4 0.7101 0.1707
WI2/WI3 0.14805 0.5928 WO3/WO4 0.556 0.0977
WI2/WI4 1.5958 0.7339
WI2/WI5 1.8892 0.8414
WI2/WI6 0.3981 0.101
WI3/WI4 2.0644 0.6064
WI3/WI5 1.6213 0.7609
WI3/WI6 0.5127 0.0965
WI4/WI5 1.711 0.6777
WI4/WI6 0.4444 0.1013
WI5/WI6 0.3448 0.0922
261
Appendices
Appendix VI: Upper and lower bounds of variables weight ratios: VS II

Table 1: 𝑒 !!! scale
WI1/WI2 3.7938 0.0216 WO1/WO2 8.7303 0.5134
WI1/WI3 3.2116 0.0256 WO1/WO3 20.0869 0.2636
WI2/WI3 14.3937 0.1353 WO3/WO4 20.0863 0.1889
Table 2: 2!!! scale

WI1/WI2 6.3495 0.0049 WO1/WO2 20.159 0.3969
WI1/WI3 5.0396 0.0062 WO1/WO3 64.0072 0.1575
WI2/WI3 40.33 0.0625 WO3/WO4 64.0072 0.0992
Table 3: 9/9-9/1scale
WI1/WI2 1.4249 0.0098 WO1/WO2 1.8419 0.8502
WI1/WI3 1.3535 0.0062 WO1/WO3 3.0001 0.6934
WI2/WI3 2.6888 0.5095 WO3/WO4 3.0001 0.641
Table 4: 10/10-18/2 scale

WI1/WI2 0.1418 1.7296 WO1/WO2 2.4753 0.7642
WI1/WI3 0.1651 1.6107 WO1/WO3 4 0.5757
WI2/WI3 0.4099 3.2076 WO3/WO4 4 0.5036
Table 5: ø mapping scale

WI1/WI2 1.3585 0.2955 WO1/WO2 1.6643 0.861
WI1/WI3 1.305 0.3406 WO1/WO3 2.2361 0.7339
WI2/WI3 1.9645 0.5996 WO3/WO4 2.2361 0.6813
262
Appendices
Appendix I: Research of Airport Performance Evaluation by DEA model

Indictors
Authors Method1 Sample
Input Output
Number of runways, gates,
employees, collection belts,
Gillen and Lall 21 US airports (1989– Number of passengers
DEA parking spots
1997 93) Amount of cargos
Length of runway
Airport and terminal areas
Hooper and
TFP model and 6 Australian airports Labour, capital, and other Non-aeronautical revenue
Hensher
Tornquist index (1988/89–91/92) cost Aeronautical revenue
1997
Number of workers
Murillo-
DEA and 33 Spanish airports Accumulated capital stock
Melchor Number of passengers
Malmquist (1992–94) proxied by amortization
1999
Intermediate expenses
32 UK regulated Number of labour
Parker Number of passengers
DEA airports (1979/80– Amounts of capital stock
1999 Amount of cargo
1995/96) Non-labour and capital cost
Number of passengers
Operational cost Aircraft movements
Sarkis 44 US airports (1990–
DEA Number of employees, gates, Amounts of operational
2000a 94)
runways revenue
Amount of cargo
Number of gates,
Gillen and Lall DEA and 22 US airports (1989– runways, employees, Number of passengers
2001 Malmquist 93) collection belts, Amount of cargo
parking spots
Number of aircraft
Martin and
37 Spanish airports Labour, capital, and movements
Roman DEA
(1997) materials cost Number of passengers
2001
Amount of cargo
Terminal size in square
meters
(i) Terminal model:
Number of aircraft parking
Pels, ijkamp, Number of passengers
34 European airports positions at the terminal.
and Rietveld DEA and SFA (ii) Movement model:
(1995–97) check-in desks,
2001 Aircraft transport
collection belts,
movement
remote aircraft parking
positions
Abbott and Number of employees
Malmquist TFP 12 Australian airports Number of passengers
Wu Amount of capital stock
index and DEA (1989/90–1999/2000) Amount of cargo
2002 Length of runway
Areas of apron
Area of departure lounges,
Fernandes and baggage claim Number of domestic
DEA 35 Brazilian airports
Pacheco 2002 Number of check-in desks, passengers
vehicle parking spots
Length of curb frontage
Operational cost
Aeronautical revenue
Bazargan and Non-operating expense
45 US airports (1996– Non-aeronautical revenue
Vasigh DEA Number of runways, gates,
2000) Percentage of on time
2003 passengers, air carrier
operations
operations, other operations
Areas of apron
Area of departure lounges,
Pacheco and
35 Brazilian airports baggage claim
Fernandes DEA Number of passengers
(1998) Number of check-in desks,
2003
vehicle parking spots
Length of curb frontage
Number of passengers,
aircraft movements
Oum, Yu, and Number of employees,
TFP model, gross 50 major airports Amount of cargo,
Fu runways, gates
TFP, residual TFP around the world non-aeronautical revenue
2003 Terminal size
1
TFP=Total Factor Productivity; VFP= Variable Factor Productivity; DEA= Data Envelopment Analysis;
SPF= Stochastic Production Frontier; FA=Stochastic Frontier Analysis
241
Appendices
Indictors
Authors Method Sample
Input Output
(i) Terminal model:

Terminal size in square
meters
Number of aircraft parking
(i) Terminal model:
positions at the terminal,
Pels, Nijkamp, Number of passengers
34 Europeans airports remote aircraft parking
and Rietveld SPA and DEA (ii) Movement model:
(1995–97) positions, runways
2003 Aircraft transport
Dummy z variables for time
movement
restrictions and for slot-
coordinated airport
(ii) Movement model:
Number of check-in desks
Number of aircraft
movements, passengers
Barros and Portuguese (1999– Number of labour Amount of general cargo,
DEA
Sampaio 2004 2000) Capital cost mail cargo
Non-aeronautical revenue
Oum and Yu 76 major airports Number of labour aircraft movements
VEP
2004 around the world Soft cost input2 Amount of non-
aeronautical revenue
Sarkis and Operational cost
44 major US airports aircraft movements
Talluri DEA Number of employees, gates,
(1990–94) Amount of operational
2004 runways
revenue, cargo
Aircraft movement
Yoshida Endogenous-weight 30 Japanese airports Size of terminal
2004 TFP (2000) Total length of runways
cargo
Length of runway
DEA and Number of passengers,
Yoshida and 67 Japanese airports Terminal size
endogenous-weight aircraft movements
Fujimoto 2004 (2000) Access cost
TFP Amount of cargo
Number of employees
Area of runway, apron,
Yu 14 Taiwan airports terminal
DEA aircraft movements
2004 (1994–2000) Active route
Aircraft noise
Population
Number of employees,
check-in desks, runways,
20 major airports Number of aircraft
Lin and Hong parking spots, baggage
DEA around the world movements, passengers
2006 claims, aprons, boarding
(2003) Amount of cargo
gates
Number of aircraft
movements,
Martin and Labor cost of passengers
34 Spanish airports
Roman DEA and SMOP Capital cost Amount of general cargo,
(1997)
2006 Materials cost mail cargo
Non-aeronautical revenue
Oum, Adler, 116 major airports
Number of employees Soft aircraft movements
and Yu VEP around the world
cost input Amount of non-
2006 (2003–5)
aeronautical revenue
Operational revenue
Vasigh and 22 US and European Operation cost Non-operational revenue
TFP model and
Gorjidooz major airports (2000– Net total assets Total terminal passengers
regression model
2006 2004) Runway area Total aircraft movements
Landing fee
Labour costs planes
Barros and
31 Italian airports Capital costs General cargo
Dieke DEA
(2001–3) Operational costs excluding Handing receipt
2007
labour costs Non-aeronautical revenue
2
Soft-cost input is a catch-all input other than labour and capital costs, including costs of outsourced
services, consultant services, utility costs, travel expenses, non-labour building and equipment
maintenance expenses, and repair costs.
242
Appendices
Indictors
Input Output
Number of labour, runways Number of passengers,
Barros 32 Argentina airports
DEA Area of aprons aircraft movements
2008a (2003–6)
Terminal area Amount of general cargo
Operational cost Price of capital-investment
Barros Random SFA 27 UK airports
Price of workers Number of passengers,
2008b model (2000/1–2004/5)
Price of capital –premises aircraft movements
Labor costs aircraft movements
Barros and
31 Italian airports Capital costs General cargo
Dieke Two-stages DEA
(2001–3) Operational costs excluding Handing receipt
2008
labor costs Non-aeronautical revenue
Fung, Wan, DEA and
25 Chinese airports Length of runway aircraft movements
Hui, and Law Malmquist TFP
(1995–2004) Terminal area Amount of cargo
2008 index
Number of labor, runways
Non-labor variable cost
Oum, Yan, and Terminal size Number of passengers,
109 world’s airports
Yu SFA Wage rate aircraft movements
(2007)
2008 Non-labor variable input Non-aeronautical revenue
price
Labor cost share
(i) Desirable
Pathomsiri, Number of passengers,
Haghani, TFP and Land area Non-delayed flights
56 US airports (2000–
Dresner, and Malmquist- Number of runways Amount of cargo
2003)
Windle Lumberger index Size of runway area (ii) Undesirable
2008 Number of delayed flights
Time delays
Yu, Hsu, Number of employees
Chang, and 4 Taiwan airports The accumulated capital
DEA Number of passengers
Lee (1995–99) stock
2008 Intermediate expense
Operational cost Price of capital-investment
Barros Random SFA 27 UK airports
Price of workers, capital – Number of passengers,
2009 model (2000–2006)
premises aircraft movements
Chi-Lok and
25 China airports Terminal area aircraft movements
Zhang DEA
(1995–2006) Length of runway Amount of cargo
2009
Number of labor Air traffic movement
Lam, Low, and 11 major airports in The value of capital (ATM)
DEA
Tang 2009 Asia Pacific (2001–5) Soft input Number of passengers
Trade value Amount of cargo.
Markov Chain
Martin,
Monte Carlo Number of labor Air traffic movement
Roman, and 37 Spanish airports
(MCMC) Capital costs (ATM)
Voltes-Dorta (1991–97)
simulation and SFA Material Work-load units (WLU)
2009
model
Number of operations per Air traffic movement
Ablanedo-
37 Mexican airports hour (ATM)
Rosas and DEA
(2009) Number of passengers per Number of passengers
Gemoets 2010
hour Amount of cargo
Air traffic movement
Number of FTE
Assaf DEA and (ATM)
27 UK airports (2007) Size of airport area
2010 bootstrapped Number of passengers
Number of runway
Amount of cargo
(ATM)
Tovar and SFA and
26 Spanish airports Number of labor, gates Average size of aircraft
Martin-Cejas Malmquist TFP
(1993–99) Airport area Share of non-aeronautical
2010 index
revenues in total airport
revenue
Number of employees,
Yang 12 Asia-Pacific
DEA and SFA runways Operational revenue
2010 airports (1998–2006)
Operational cost
243
Appendices
Indictors
Input Output
Assaf No. of employees Number of passengers,
2011 Malmquist Size of terminal area aircraft movements
13 Australian
bootstrapped Operational costs Amount of cargo
(2002-2007)
methodology
Barros SFA Operational cost Price of capital-investment
2011 17 Angola and Trend variables Number of passengers,
Mozambique airports Price of workers, capital – aircraft movements
(2000-2010) premises
Curi, Gitto, Number of employees, Number of passengers,
and Mancuso 18 Italian airports runways aircraft movements
2011 Bootstrapped DEA Size of apron Amount of cargo
(2000-2004)
Fung, Chow Length of runway, size of Number of passengers,

2011 Malmquist TFP 41 Chinese airports terminal. aircraft movements
index (1995-2004) Amount of cargo
Lozano and Size of runway area, apron, Number of passengers,

Gutierrez 2011 41 Spanish airports terminal area aircraft movements
DEA Number of check-in desks, Amount of cargo
(2006)
baggage claims, gates
Tsekeris DEA 39 Greek airports Operating hours Number of passengers
2011 (2007) Number of runways Amount of cargo,
Size of terminal area Air traffic movement
Size of airplane parking area
Assaf and SFA 73 world’s airports Number of employees, Number of passengers,
Gillen And Bayesian (2002-2008) runways, size of terminal, aircraft movements
2012 model Other operational costs Non-aeronautical revenue
Assaf et al. Bayesian model 27 UK airports The price of labour, the total of aeronautical,
2012 (1998-2008) The price of capital, non-aeronautical revenues
The price of materials
Chow and TFP 30 airports in greater Size of terminal, Number of passengers

Fung China Length of runway Amount of cargo,
2012 (2000-2006) Air traffic movement
Gitto and DEA and 28 Italian airports Labour cost, capital invested, Aeronautical revenue
Mancuso bootstrapped (2000-2006) soft costs Non-aeronautical revenue
2012a technique
Gitto and TFP and 28 Italian airports Labour cost, capital invested, Number of passengers
Mancuso bootstrapped (2000-2006) soft costs Amount of cargo,
2012b technique Air traffic movement
the total of aeronautical,
non-aeronautical revenues
Perelman and DEA 21 Latin American Number of employees, Number of passengers
Serebrisky airports runways, Size of terminal Amount of cargo,
2012 (2000-2007) Air traffic movement
Scotti et al. SFA 38 Italian airports Number of authorised flights Number of passengers
2012 (2005-2008) per hours, number of aircraft Amount of cargo,
parking positions, size of Air traffic movement
terminal, number of check-in
desk, number of baggage
claims, number of employees.
Wanke Bootstrapped DEA 65 Brazilian airports Air traffic movement Number of passengers
2012a and FDH (2009) Amount of cargo,
Wanke Bootstrapped DEA 63 Brazilian airports Size of terminal, size of Number of passengers
2012b (2009) apron, number of runways, Amount of cargo,
length of runway, number of Air traffic movement
aircraft parking positions,
size of airport, number of
parking space
Zhang et al. DEA 37 Chinese airports Take-off distance available Number of passengers
2012 (2009) Landing distance available Amount of cargo,
244
Appendices
Appendix II: Analytical Hierarchy Process (AHP) questionnaire
Airport Efficiency Evaluation Questionnaire

Dear Sirs/Madam,
I am currently undertaking a PhD research programme in Cardiff Business School,

Cardiff University, examining airport efficiency in Europe and the Asia Pacific region
using quantitative, analytical techniques. As part of this process, I am conducting a
questionnaire survey which aims to rank airport efficiency indictors. It is expected that
data collected through this questionnaire will help to develop an appropriate airport
efficiency evaluation framework. As a senior manager in the airport industry, you are
invited to provide your perceptions of airport efficiency. Your opinions are extremely
crucial to this research; the attached questionnaire is part of the research. There are no
right and wrong answers.
Your participation in this questionnaire survey is entirely voluntary. The information

gathered in this survey will be treated in the strictest confidence and be used only for
academic research purposes and you are entitled to withdraw your answer at anytime.
This survey will take you about 10 minutes to complete. If you consent to participate in
this survey, please fill out the questionnaire and send it back to us in the attached return
envelope. If you have any queries or concerns regarding the survey, please contact
either myself or my supervisor, Dr. Andrew Potter (PotterAT@cf.ac.uk). If you wish to
receive a summary of the survey findings, please indicate this at the end of the
questionnaire or e-mail us and I will be happy to send the summary to you when the
research is over.
Please accept my thanks for your anticipated co-operation.
Yours faithfully,
PO-LIN LAI
Postgraduate Research Student
Logistics and Operations Management Section,
Room D46 Aberconway Building,
Cardiff Business School,
Cardiff University, Cardiff CF10 3EU, UK
Tel: +44-(0)2920 875480
Email: laip@cardiff.ac.uk
245
Appendices
Questionnaire Explanation
In this questionnaire you will be asked to compare the following criteria which are used
to evaluate airport operational efficiency. In this research, both variables can be
separated into two hierarchies.
Input
Main criteria Sub-criteria Definition of the criteria
The number of full-time equivalent employees directly
Number of employees
employed by the airport.
The number of gates with jet ways and other non jet-way
Number of gates
Airport gates.
capacity Number of runways The available number of runways at each airport.
factors
Size of terminal area The total area of passenger terminals.
The average runway length of every runway in each
Length of runway
airport.
The financial resources needed to run an airport including

Financial salaries and benefits, communications and utilities,
Operational cost
factors supplies, materials, repairs and maintenance, services and
other expenses.
Output
Main criteria Sub-criteria Definition of the criteria

The number of passengers arriving or departing at an
Number of passengers airport (including terminal passengers and transit
passengers).
Service Amount of freight and The weight of property carried on an aircraft and the
performance mails weight of Post Office mail carried.
The number of landings or take-offs of aircraft engaged in
Aircraft movement the transport of passengers, cargo or mail on commercial
terms
The revenues that are generated by aviation activities such

Financial
Aeronautical revenues as landing fees, terminal fees, apron charges, fuel flowage,
performance
fixed base operators (FBOs), rentals and utilities.
246
Appendices
How to complete the questionnaire

You are invited to tick the most appropriate box according to your opinion on how
important one criterion over another when you are evaluating an airport operational
efficiency. If your preference is between two levels of importance, e.g. between Strong
Importance and Very Strong Importance, please tick the intermediate box between
them.
Intensity of influence Definition

EI Equal Importance
MI Moderate Importance for one over another
SI Strong Importance
VSI Very Strong Importance
ExI Extreme Importance
Examples
Each row has a single comparison for you to make. As stated above, between two
criteria “EI” means that both criteria are of Equal Importance. If you think, for example,
the importance of Numbers of Employees over Number of Gates is Strong Importance,
your answer should be placed on the left side subject to the degree of relative
importance, and then you would tick as follows:
Criterion Intensity of Importance Criterion
ExI VSI SI MI EI MI SI VSI ExI

Number of
□ □ □ □ √ □ □ □ □ □ □ □ □ □ □ □ □ Number of gates
employees
Tick √ means: the importance of Number of Employees over the criterion Number of
Gates is a Strong Importance.
If, however, you think the importance of Number of Gates over the criterion Number of
Employees is an Extreme Influence, then you should thick as follow:

Number of Number
□ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ √
employees of gates
If the importance is the same, tick Equal Importance will be the answer.

Number
of □ □ □ □ □ □ □ □ √ □ □ □ □ □ □ □ □ Number of gates
employees
The Survey of Input variables

1. Comparison of main criteria

Airport Financial
□ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □
capacity factor factor
247
Appendices
2. Comparison of the relative importance of each sub-criterion in second hierarchy:

The Input variables: Airport capacity factors
□ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ Number of gates
Number □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ Number of runways

of
employees
□ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ Size of terminal area
□ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ Length of runway
□ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ Number of runways
Number
of gates □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ Size of terminal area
□ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ Length of runway
□ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ Size of terminal area

Number
of
runways
□ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ Length of runway
Size of
terminal □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ Length of runway
area
The Survey of Output variables

1. Comparison of main criteria

Service Financial
□ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □
performance performance
2. Comparison of the relative importance of each sub-criterion in second hierarchy: The

output variables: Service performance
Amount of cargos and

□ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □
Number mails
of
passengers
□ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ Aircraft movement
Amount
of cargos □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ Aircraft movement
and mails
248
Appendices
3. Personal information:
This information will be used to enable clusters to be formed from the responses.
However, individual responses will not be identifiable.
A. What is your position in your company (choose one):

Vice president or above Manager / Assistant manager Director /Vice
director
Sales representative Clerk Other (please specify):
B. How long have you worked in airport related sector? years
If you would like to receive a summary of the results of this research, please contact us
by e-mail so that we can send the report to you.
Thanks for your patience and help
249

A Study On The Relationship Between Airport Privatisation and Airport Efficiency

Uploaded by

Copyright:

Available Formats

A Study On The Relationship Between Airport Privatisation and Airport Efficiency

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Study On The Relationship Between Airport Privatisation and Airport Efficiency

Uploaded by

Copyright:

Available Formats

A STUDY ON THE RELATIONSHIP BETWEEN

AIRPORT PRIVATISATION AND AIRPORT EFFICIENCY

AN APPLICATION OF USING AHP/DEA METHODS

A thesis submitted in fulfillment of the requirements for the degree of

Logistics and Operations Management Section

Signed Po-Lin Lai

Date 10 April 2013

This thesis is being submitted in partial fulfilment of the requirements for

Signed Po-Lin Lai

Date 10 April 2013

This thesis is the result of my own independent work/investigation, except where

Signed Po-Lin Lai

Date 10 April 2013

Signed Po-Lin Lai

Date 10 April 2013

First and foremost, I would like to express my greatest appreciation to my supervisor,

It is a pleasure to thank the participants of the questionnaire survey of this research.

Preface Declaration and statements i

Chapter 2 ANALYSIS OF AIRPORT PERFORMANCE RESEARCH 12

Chapter 3 DEVOLUTION OF AIRPORT OWNERSHIP AND GOVERNANCE 34

Chapter 5 EMPIRICAL ANALYSIS: AHP ANALYSIS 102

Chapter 6 Airport Efficiency Analysis I: Basic DEA models and 131

Chapter 7 Airport Efficiency Analysis II: AHP/DEA-AR model 178

Chapter 8 RESEARCH CONCLUSIONS 217

1.1 World annual air traffic 2

2.1 Distribution of sample region classified by country and continent 24

3.1 Traditional airport organisation 38

4.1 The research onion 64

5.1 AHP hierarchies of Variable Set I (ten variables) 108

2.1 Analysis methods used in airport performance evaluation 20

3.1 Activities of non-aeronautical revenue and aeronautical revenue 39

4.1 Comparisons of philosophical research paradigms 66

5.1 The preliminary indictors of airport efficiency evaluation system 104

6.1 Correlation coefficients among inputs and outputs variables: VS I 135

8.1 Summary of efficiency scores and Mann-Whitney test 222

An overview of the research is presented in this chapter. It outlines the challenges of

1.1 INTRODUCTION AND MOTIVATION

Figure 1.1: World annual air traffic

Source: Airbus (2012).

Figure 1.2: Passenger development Figure1.3: Annual traffic growth

Source: Boeing (2012). Source: Boeing (2012).

1.2 RESEARCH QUESTIONS

Another aim of this research is to determine if airport privatisation influences airport

The research questions in the following section can be classified as methodological

1.3 THE SCOPE AND STRUCTURE OF THE RESEARCH

Chapter 2 presents a structured literature review that encompasses airport performance

Chapter 4 examines the technical considerations of the current research development.

Chapter 7 conducts another sensitivity analysis by applying the AHP/DEA-AR model

Figure 1.4: Structure of the research

Chapter Scope Research

·∙ Exploration of airport efficiency studies

Airport Privatisation and

Research Methodology and

Descriptive Analysis and

Research Conclusions and

1.4 RESEARCH CONTRIBUTIONS

2.2 METHODOLOGICAL APPROACHES

2.2.1 PARTIAL MEASURES