Airport Benchmarking: An Efficiency Analysis of European ...
Transcript of Airport Benchmarking: An Efficiency Analysis of European ...
i
Airport Benchmarking: An Efficiency Analysis of
European Airports from an Economic and
Managerial Perspective
by
Dipl. Volksw. (FH) Vanessa Philippa Liebert
A thesis submitted in partial fulfilment of the
requirements for the degree of
Doctor of Philosophy
in Economics
School of Humanities and Social Sciences
Approved, Dissertation Committee
Prof. Dr. Gert Brunekreeft, Jacobs University Bremen (Chair)
Prof. Dr. Adalbert FX Wilhelm, Jacobs University Bremen
Prof. Dr. Hans-Martin Niemeier, University of Applied Sciences Bremen
Date of Defense: 08.03.2011
ii
iii
So eine Arbeit wird eigentlich nie fertig, man muss sie für fertig
erklären, wenn man nach der Zeit und den Umständen das
Möglichste getan hat.
Johann Wolfgang von Goethe (1749-1832)
Acknowledgements
iv
ACKNOWLEDGEMENTS
The expenditure of time to build a new airport on a green field may take three to four years.
Extensive academic research on the efficiency of existing airports with the objective to improve the
airports performance may require twice as long but is certainly more cost efficient… By now, nearly
six years have passed since I began with my research and I would like to use the opportunity and
express my gratefulness to people who have (technically and mentally) supported me during the whole
process of writing this doctoral dissertation.
First and foremost, I owe my gratitude to my supervisors Prof. Gert Brunekreeft and Prof. Hans-
Martin Niemeier for the opportunity to write a dissertation. Both made available their encouragement
and support in a number of ways. I would also like to thank my co-supervisor Prof. Adalbert Wilhelm
for valuable feedback. Furthermore, I would like to acknowledge the Federal Ministry of Education
and Research for financial support of the research project German Airport Performance (GAP) on
which this dissertation is based.
I am grateful to Dr. Nicole Adler for helpful discussion and methodological support. She taught
me to Keep It Simple and Stupid (KISS). I wish to express my gratitude to Adél Németh for her
encouragement and unlimited supply of coffee and chocolate and Nathalie-Chantal McCaughey for
proofreading. Moreover, I would like to thank my PhD fellows Nele Friedrichsen, Karsten Fröhlich,
Roland Meyer, Tolga Ülkü, Volker Wannack and Katya Yazhemsky. The student workers from the
GAP project as well as Prof. Jürgen Müller and Prof. Hansjochen Ehmer are gratefully acknowledged
for assisting me with collecting the data. Furthermore, I thank Prof. David Gillen, Prof. Peter Forsyth
and Dr. Mike Tretheway for helpful remarks on earlier drafts of this thesis.
Without the unlimited patience of my friends ‘outside’ and their repeated gamesmanship I might
have gone mad (academics are strange people…) and I would like to thank them all.
Finally, I owe my deepest gratitude to my family for their unrestricted support. My parents were
always there when I needed them most; they deserve far more credit than I can ever give them. This
thesis would not have been possible without my husband and his continuous encouragement. I am
indebted for his support during his countless days as grass widower and full-time daddy. Last but not
least, I thank my children who shared their mother with her ‘other baby’. This thesis, though not
appropriate for bedtime stories, is dedicated to them.
Hamburg in May 2011,
Vanessa Liebert
Abstract
v
ABSTRACT Subsequent to airline deregulation, an increasing commercialization, privatization and
restructuring gradually changed a sovereign operated airport industry to modern business enterprises.
Where market power was likely to be exploited, airports may now face competition with nearby
airports or other transport modes. Consequently airport benchmarking became popular for
comparisons with competitors and to assess efficiency changes resulting from the structural change.
Within academic benchmarking a number of studies emerged utilizing parametric and non-parametric
approaches to estimate the productivity and efficiency of airports. Building on the limitations and
discussions from previous research the general objective of this thesis is to further the understanding
of the airport industry and to improve airport benchmarking in order to enhance its usefulness for
managerial, political and regulatory purposes. Particular emphasis is given on the consideration of the
heterogeneous character of airports and how to explain efficiency difference across airports. The
cumulative thesis presents the results of three research articles. The first article provides a survey on
the methods, data and findings of empirical research from the current literature in airport
benchmarking. The survey indicates substantial progress in the methodological application however
many issues still remain unresolved such as the appropriate measurement of capital. The second article
assesses the combined impact of ownership form, economic regulation and competition on airport
performance and pricing in order to search for the most desirable combination. Australian and
European are analyzed using non-parametric data envelopment analysis (DEA) in a first stage
efficiency measurement and regression analysis in a second stage environmental study. The results
reveal that airports not facing competition should be regulated to increase cost efficiency and prevent
exploitation of market power. However, in a competitive setting, regulation inhibits airports of any
ownership from operating efficiently. Nevertheless, unregulated private airports appear to remain
profit-maximizer within competition. The third article aims to improve the airport benchmarking
process. Most previous studies either treat the airport production technology as a black box or separate
terminal and airside activities, assessing them individually. This research analyzes European airports
as a single unit due to the direct complementarities but opening the black box through network DEA.
Combined with dynamic clustering appropriate benchmarks are identified based on pre-defined
characteristics. Compared to basic DEA models, the results of the network DEA structure provide
more meaningful benchmarks with comparable peer units and target values that are achievable in the
medium term.
Keywords: benchmarking, airport efficiency, data envelopment analysis (DEA), second-stage
regression, external heterogeneity
List of Abbreviations
vi
LIST OF ABBREVIATIONS
ADP Aéroports de Paris (Paris airports) AENA Aeropuertos Españoles y Navegación Aérea (Spanish Airports and Air Navigation)ATM Air Transport Movements a/c Aircraft ACI Airports Council International ANOVA Analysis of Variance ADV Arbeitsgemeinschaft Deutscher Verkehrsflughäfen (German Airports Association)ACCC Australia Competition & Consumer Commission BAA British Airports Authority capex Capital Expenditures CCD Caves, Christensen and Diewert CAA Civil Aviation Authority CRS Constant Returns-To-Scale CPI Consumer Price Index DHL Dalsey, Hillblom and Lynn (Parcel service) DEA Data Envelopment Analysis DMU Decision Making Unit dom. Domestic DAA Dublin Airport Authority EW-TFP Endogenous-Weight Total Factor Productivity EU European Union FDH Free Disposal Hull FTE Full-Time Equivalents GA General Aviation GAP German Airport Performance GDP Gross Domestic Product int. International IATA International Air Transport Association ICAO International Civil Aviation Organization km kilometres max. Maximum MLE Maximum Likelihood Estimation min. Minimum ND Non-Discretionary NIRS Non-Increasing Returns-To-Scale no. Number obs. Observations op. Operating OLS Ordinary Least Squares PAX Passengers PIM Perpetual Inventory Method
List of Abbreviations
vii
pos. Positions PIN Price-Index Number PC Principal Component PCA Principal Component Analysis reg. Regional RPI Revenue Price Index rev. Revenues RWY Runway SH&E Simat, Helliesen & Eichner (Aviation consultancy) SBM Slack-Based Measure SFA Stochastic Frontier Analysis SMOP Surface Measure of Overall Performance totex Total Expenditures TFP Total Factor Productivity UPS United Parcel Service VFP Variable Factor Productivity VRS Variable Returns-To-Scale WLU Work Load Units
List of Airports and Country Codes
viii
LIST OF AIRPORTS AND COUNTRY
CODES ABZ Aberdeen LEJ Leipzig AMS Amsterdam LGW London Gatwick ATH Athens LHR London Heathrow BFS Belfast LJU Ljubljana BHX Birmingham LTN London Luton BLQ Bologna LYS Lyon BRE Bremen MAN Manchester BRU Brussels MEL Melbourne BTS Bratislava MLA Malta BUD Budapest MLH Basel Mulhouse CGN Cologne Bonn MME Durham Tees Valley CPH Copenhagen MRS Marseille DRS Dresden MUC Munich DTM Dortmund NCE Nice DUB Dublin NUE Nuremberg DUS Dusseldorf OSL Oslo EDI Edinburgh PER Perth EMA East Midlands (Nottingham) RIX Riga FLR Florence SOU Southampton FRA Frankfurt STN London Stansted GLA Glasgow STR Stuttgart GOA Genoa SYD Sydney GVA Geneva SZG Salzburg HAJ Hanover TLL Tallinn HAM Hamburg VCE Venice LBA Leeds Bradford VIE Vienna LCY London City ZRH Zurich
AT Austria IE Ireland AU Australia IT Italy BE Belgium LV Latvia CH Switzerland MT Malta DE Germany NL The Netherlands DK Denmark NO Norway EE Estonia SI Slovenia FR France SK Slovakia GR Greece UK United Kingdom HU Hungary US United States
List of Figures
ix
LIST OF FIGURES
Fig. 1: Quantitative benchmarking approaches ..................................................................20
Fig. 2: Stochastic frontier analysis .....................................................................................22
Fig. 3: Data envelopment analysis......................................................................................25
Fig. 4: Structure of this dissertation....................................................................................29
Fig. 5: Airport production function in DEA.......................................................................34
Fig. 6: Technical, allocative and economic efficiency .......................................................41
Fig. 7: Quantitative methods in productivity and efficiency analysis ................................42
Fig. 8: Models in data envelopment analysis .....................................................................45
Fig. 9: Models in stochastic frontier analysis .....................................................................48
Fig. 10: Inputs and outputs in previous airport benchmarking studies...............................53
Fig. 11: Airport network technology ................................................................................107
Fig. 12: Benchmark clustering..........................................................................................110
Fig. 13: Two-stage airport network technology ...............................................................114
Fig. 14: Kruskal-Wallis ANOVA for outsourcing ...........................................................122
Fig. 15: Co-Plot graphic display of Vienna’s input-oriented strategy..............................127
Fig. 16: Co-plot of input minimization results with emphasis on Hanover .....................129
Fig. 17: Catchment area of Hanover airport (2 hour drive)..............................................130
Fig. 18: Current and target output values for Lyon ..........................................................132
Fig. 19: Co-plot of output maximization results with emphasis on Lyon ........................133
List of Figures
x
LIST OF TABLES
Tab. 1: Comparison of DEA, SFA and PIN properties...................................................... 50
Tab. 2: Studies using non-parametric approaches ............................................................. 62
Tab. 3: Studies using parametric approaches..................................................................... 68
Tab. 4: Studies using price-based index approaches.......................................................... 71
Tab. 5: Regression analysis................................................................................................ 83
Tab. 6: Variables in analysis (DEA) .................................................................................. 84
Tab. 7: Combination of environmental variables analyzed ............................................... 89
Tab. 8: Second-stage regression results from the individual cost efficiency model.......... 94
Tab. 9: Second-stage regression results from the combined cost efficiency model .......... 96
Tab. 10: Second-stage regression results from the combined revenue model ................... 98
Tab. 11: List of airports.................................................................................................... 102
Tab. 12: DEA efficiency scores ....................................................................................... 103
Tab. 13: Variables in airport efficiency analysis ............................................................. 120
Tab. 14: Banker F-test for outsourcing ............................................................................ 123
Tab. 15: Benchmarking Hanover airport ......................................................................... 128
Tab. 16: Output benchmarks for Lyon airport ................................................................. 131
Tab. 17: Airport dataset ................................................................................................... 135
List of Figures
xi
TABLE OF CONTENTS
ACKNOWLEDGEMENTS .........................................................................................................iii
ACKNOWLEDGEMENTS .........................................................................................................iv
ABSTRACT ...............................................................................................................................v
LIST OF ABBREVIATIONS.......................................................................................................vi
LIST OF AIRPORTS AND COUNTRY CODES .........................................................................viii
LIST OF FIGURES ...................................................................................................................ix
1 INTRODUCTION .............................................................................................................14
1.1 Theoretical background ..........................................................................................15
1.1.1 The beginning of a liberalization process in the aviation industry................15
1.1.2 The changing nature of European airports - a motivation for
benchmarking .............................................................................................16
1.1.3 Users of airport benchmarking ...................................................................18
1.1.4 Quantitative approaches of benchmarking ..................................................20
1.2 Contribution of this PhD research ..........................................................................27
1.2.1 Discussions arising from previous research.................................................27
1.2.2 Motivation for research...............................................................................28
1.2.3 Article 1: A survey of empirical research on the productivity and
efficiency measurement of airports .............................................................31
1.2.4 Article 2: Joint impact of competition, ownership form and economic
Regulation on airport performance .............................................................32
1.2.5 Article 3: Benchmarking airports from a managerial perspective .................33
1.3 Concluding remarks ................................................................................................35
List of Figures
xii
2 A SURVEY OF EMPIRICAL RESEARCH ON THE PRODUCTIVITY AND EFFICIENCY
MEASUREMENT OF AIRPORTS....................................................................................... 37
2.1 Introduction ........................................................................................................... 38
2.2 Productivity and efficiency measurement concepts................................................. 40
2.2.1 Productivity and efficiency ......................................................................... 41
2.2.2 Price-based index number approaches........................................................ 42
2.2.3 Data envelopment analysis ......................................................................... 44
2.2.4 Stochastic frontier analysis ......................................................................... 47
2.2.5 Comparison of techniques.......................................................................... 50
2.3 Variable selection in airport studies ........................................................................ 51
2.3.1 Outputs...................................................................................................... 51
2.3.2 Inputs......................................................................................................... 52
2.4 Empirical results of productivity and efficiency studies .......................................... 55
2.4.1 Productivity and efficiency changes over time ............................................ 56
2.4.2 Empirical effects of ownership................................................................... 57
2.4.3 Scale effects................................................................................................ 59
2.5 Conclusion ............................................................................................................. 60
2.A Appendix ............................................................................................................... 62
3 JOINT IMPACT OF COMPETITION, OWNERSHIP FORM AND ECONOMIC
REGULATION ON AIRPORT PERFORMANCE ................................................................. 72
3.1 Introduction ........................................................................................................... 73
3.2 Literature on competition, regulation and ownership ............................................. 76
3.3 Methodology and model specification .................................................................... 80
3.3.1 Data envelopment analysis ......................................................................... 81
3.3.2 Second-stage regression.............................................................................. 82
3.4 Dataset ................................................................................................................... 83
3.4.1 Variables in the first-stage efficiency analysis .............................................. 84
3.4.2 Variables in the second-stage regression ..................................................... 86
3.5 Empirical results..................................................................................................... 89
3.5.1 Efficiency scores from data envelopment analysis ...................................... 89
3.5.2 Regression results explaining cost efficiency ............................................... 92
3.5.3 Regression results explaining airport charges .............................................. 97
3.6 Conclusions............................................................................................................ 99
List of Figures
xiii
3.A Appendix ..............................................................................................................102
4 AIRPORT BENCHMARKING FROM A MANAGERIAL PERSPECTIVE ..............................104
4.1 Introduction .........................................................................................................105
4.2 Methodology ........................................................................................................109
4.2.1 Dynamic clustering ...................................................................................110
4.2.2 Network DEA ..........................................................................................111
4.2.3 Principal component analysis integrated with DEA ..................................111
4.2.4 Visualizing multiple dimensions ................................................................112
4.2.5 Measuring efficiency variation across groups ............................................113
4.3 Model formulations ..............................................................................................113
4.4 Dataset .................................................................................................................118
4.5 Empirical results ...................................................................................................120
4.5.1 Efficiency variation across groups.............................................................121
4.5.2 Comparison of basic and network DEA ...................................................124
4.5.3 Benchmarking airports..............................................................................126
4.6 Conclusion............................................................................................................133
4.A Appendix ..............................................................................................................135
5 REFERENCES................................................................................................................136
14
1 INTRODUCTION
Since I was little, airports have been a fascinating place to me. My dad often took me to
the airport and let me watch the airplanes departing and arriving. However, at that time I have
certainly neither expected to examine the airports’ efficiency nor to submit a doctoral
dissertation 20 years later. The application of benchmarking (or efficiency analysis) aroused
my interest while I was studying one year abroad in England where Data Envelopment
Analysis (DEA) and other efficiency measurement techniques were already widely applied.
Hence what may be more interesting than airport benchmarking? I combined both issues and
together with my professor (Dr. Hans-Martin Niemeier, University of Applied Sciences
Bremen) we applied for a research grant aiming to assess the efficiency of German airports
which were hardly considered at that time. At the beginning of my position as researcher in
the German Airport Performance (GAP) project1 and as PhD candidate at Jacobs University I
have not expected how challenging but also exciting it may be to assess an industry that has
been under continuous structural changes over the last two decades…
1 The GAP project on “Efficiency Measurements of German Airports in Comparison to Europe, Australia and North America” is a joint research project of the University of Applied Sciences Bremen, the Berlin School of Economics and Law (HWR) and the International University of Applied Sciences Bad Honnef, which has been sponsored by the Federal Ministry of Education and Research (BMBF) from 2005 until 2009. Its aim is to investigate the changing nature and performance of European airports, their commercialization and competitive environment, as well as the need for further financial and environmental regulation. For details see http://www.gap-projekt.de.
Introduction
15
1.1 Theoretical background
During the last four decades, an upward trend in international tourism and globalization
substantially increased traffic rates in the aviation sector. Although several external shocks
(e.g. Gulf war, economic downturn, terror attacks in 2001) temporarily interrupted this trend
overall growth was not substantially impacted. One major influence on this growth has been
the deregulation of the airline industry, which began in the late nineteen-seventies and
resulted in lower airfares. This was the starting point of a gradual liberalization process in the
aviation industry. The opening of the aviation market increased competition of a previously
heavily restricted industry. As a result from the airline deregulation, many airports have felt
and still feel increasingly exposed to the cost pressure and are obliged to operate efficiently.
Increasing commercialization, privatization activities, changes in economic regulation or
restructuring lead to substantial changes in an initially sovereign operated industry.
Consequently, performance measurement of airports became increasingly important for
comparisons with competitors and to assess efficiency changes resulted from the structural
change.
The following section provides an introduction to the changing nature of the aviation
industry and highlights the usefulness of airport benchmarking. Efficiency measurement
techniques that are widely adopted in the academic benchmarking literature will be presented
thereafter.
1.1.1 The beginning of a liberalization process in the aviation industry
Historically, the worldwide civil aviation industry has been a fully regulated environment
under the ownership of public authorities. Political restrictions on market entries, ticket fares,
capacities and frequencies protected national carriers in a contestable market characterized by
low economies of scale (Doganis 2002). In the nineteen-seventies, economists began to
debate if there was indeed a need for a regulated airline industry in which passenger growth
and technological progress are likely to be constrained by political restrictions. Considering
the downsides of heavy handed regulation the US removed restrictions on domestic routes,
fares and schedules in 1978. Although many European governments were initially reluctant to
resign control of their flag carriers, they nevertheless followed the liberalization process in
1987. European liberalisation took place in stages. Initially tariff flexibility was increased and
market entry was made easier. Then in 1993 restrictions on routes, capacities and market
Introduction
16
entries were removed within the European Union (EU). Since 1997, cabotage is allowed,
which means that EU airlines can operate city-pairs in foreign countries within the EU. This
move turned the EU into the largest open aviation market in the world (Maurer 2003). A
major step in the liberalization process was reached with the EU-US open sky agreement.
This agreement became effective in March 2008 and displaced numerous bilateral
agreements. Nevertheless the liberalization process is an ongoing one. The next target of the
EU is an open sky with neighbour states and other countries outside Europe in order to
enlarge the single aviation market worldwide (European Commission 2007).
Exposed to increased competitive pressures several airlines in Europe began to complain
about markets of air transport service suppliers such as ground handling2 which remained
regulated. For example in Spain the state-owned carrier Iberia possessed market power in
ground handling operations at hub airports and other airlines were forced to purchase their
overcharged service. In Germany, Italy or Austria the airport provided ground handling
services in a fully regulated market where independent providers had little chance to enter. As
a further stage in the liberalization process of the aviation industry, ground handling services
within the EU were deregulated in 1996. In a gradual implementation schedule, the directive
allowed self handler (airlines) and independent third party providers to enter the market
(Templin 2007).
Following the liberalization process of the airline and ground handling market, a gradual
change in the nature of the airport industry may be expected.
1.1.2 The changing nature of European airports - a motivation for
benchmarking
Similar to the airline market prior liberalization, European airports were mostly deemed
state-owned entities with the objective to provide and operate the infrastructure for airlines.
As with other infrastructure based services and utilities airports were viewed as natural
monopolies enjoying both economies of scale and market power (Czerny 2006).
Consequently, in order to encourage efficiency and avoid market power exploitation, the
majority of commercial airports were subject to economic regulation. In Europe, passenger
and landing fees charged to airlines have traditionally been regulated according to a rate-of-
return or cost-plus principle. Such regulation permits airports to generate sufficient revenue to
2 Ground handling activities include the handling of passengers, baggage, freight and mail, ramp handling, fuel and oil handling, aircraft services and maintenance (Graham 2004).
Introduction
17
cover total expenditures, including the depreciation of capital and an expected rate of return
on capital (Reinhold et al. 2010).
With airline liberalization many airports moved away from being public utilities towards
operating as modern enterprises pursuing commercial objectives. After the successful
privatization of the British Airports Authority (BAA) in the late Eighties, a number of
privatization processes have been actively promoted by governments with the proclaimed
intention of reducing government involvement and increasing airport productivity and
innovation.
Given the assumed behaviour to maximize profits in a monopolistic environment, the
majority of privatized airports in Europe remained subject to economic regulation (Gillen
2010). However, as initial cost-based regulation procedures were assumed to engender
overcapitalization rather than productive efficiency (Averch and Johnson 1962) a number of
privatized airports adopted incentive regulation. Price-cap regulation as proposed by
Littlechild (1983) was introduced at the regulated BAA airports along with privatization.
Price-caps are generally set over a regulatory period of five years according to the RPI-X
formula where RPI represents the retail price index and X is the efficiency improvement that
the regulators consider reasonable within the timeframe. If the airport management achieves
greater cost reductions over the five year period, the gains are enjoyed by the company.
Having started to operate as modern businesses many airports changed their management
style towards increasing commercialization. Substantial investments in non-aeronautical
activities such as shopping malls were undertaken to augment their revenues from non-
aeronautical3 sources in order to cross-subsidize aviation charges and attract additional
airlines and passengers to the airport (Zhang and Zhang 2010). Furthermore vertical
boundaries have changed over time. A number of airports in Germany restructured their
labour-intensive ground handling segment in order to set flexible tariffs and compete with
independent providers.
Airline deregulation induced competition for airport services covering a multiplicity of
markets. The liberalization of bilateral air service agreements offered the opportunity to
attract international traffic to gateway airports (e.g. Amsterdam and Frankfurt) and secondary
hubs. Underutilized secondary airports (e.g. Lübeck) and former military airports started to
serve low cost carriers sharing local catchment areas with primary airports. Alternative modes
3 Throughout the thesis the terms non-aeronautical, non-aviation and commercial activities are used interchangeably.
Introduction
18
of transport such as high speed rail compete with aviation in the medium distance markets
(Tretheway and Kincaid 2010).
Furthermore, subsequent to airline deregulation lower airfares substantially increased the
traffic volume with growing tourism and globalization trends. In order to meet future
demands, congested airports needed to expand their capacity and to introduce new
technologies to increase runway and terminal system capacities. However at many major
airports the excess demand was rationed rather inefficiently through queuing and slot
allocation mechanisms having increased the number of non-weather related delays.
In summary the airport industry evolved into a dynamic market environment. Increasing
commercialization, privatization and restructuring processes, a shift towards incentive
regulation and advanced technologies changed the nature of the airport industry and may have
contributed to productivity and efficiency changes. For these reasons airports offer a rich field
for performance comparisons commonly defined as benchmarking4.
1.1.3 Users of airport benchmarking
Although benchmarking was already applied in other transport sectors and regulated
utilities in the nineteen seventies, it only became important in the airport industry twenty
years later. Graham (2005) argues that the increasing interest in airport benchmarking is a
result of the changes in ownership and the liberalization, commercialization and globalization
trends which have influenced airport business growth, complexity and competitiveness.
However, the comparison of airports generally appears to be challenging given their
unique character. Airports offer a heterogeneous mix of services. Some airports mainly serve
passengers while others fill empty capacities with cargo operations for parcel services (e.g.
Leipzig as the European hub of DHL). Furthermore, airports may be highly vertical integrated
and offer handling services and commercial products whereas others outsource this activity to
independent providers and concessionaires. Generally, the importance of commercial
activities which generate extra revenues may also vary among the airports. Lumpy
investments which are typical for airports further complicate financial comparisons when
airports are in different stages of their life cycles. Moreover, airports are often heavily
affected by multiple factors that are beyond the control of an airport manager. For example
4 Initially, the term of benchmarking was introduced by Xerox in 1979 and was referred to reverse engineering. Due to a decrease in market shares Xerox systematically compared their copy machine with a product by Canon who offer a similar product for a lower price (Dence 1995).
Introduction
19
the location may influence the airport’s operation. Night curfews at city airports reduce the
operating hours and airports near the coast such as Amsterdam require additional runways and
a special configuration to handle operations consistently irrespective of weather conditions. In
short, no airport is a smaller version of a large counterpart (Forsyth 2000). Nevertheless,
Adler et al. (2009) cited Peter Drucker who argues ’what you cannot measure, you cannot
manage’. For a multiple reasons as outlined below airport benchmarking has received
increasing interest by various airport stakeholders.
Airport benchmarking may for instance be utilized for managerial purposes. Airport
managers compare overall or partial processes (e.g. ground handling activities) with potential
competitors or best-practice airports to develop new strategies. In order to avoid improper
comparisons, Frankfurt may include other European hubs such as Amsterdam, London-
Heathrow, Paris or best-practice examples such as Hong Kong, Singapore and Dubai rather
than nearby airports (Kincaid and Tretheway 2006).
Customers, shareholders and investors are interested in benchmarking as decision-making
instrument. Airlines, as the intermediate between airports and passengers, prefer efficient
airports with low costs, high service standards and no delays. Moreover, passengers prefer
airports with low queue lengths that are located close to the city centre and are equipped with
shopping and entertainment facilities. Private shareholders and investors expect high and fast
returns on investments.
National and regional governments mostly assess airport performance from an economic
perspective. They may examine effects of policy changes in before and after comparisons or
with other countries. As an example privatization activities or changes in economic regulation
are assessed that may have lead to improving efficiency, pricing and investments.
Furthermore, benchmarking is applied to inform policy. In Australia, major infrastructure
service industries such as electricity, gas supply or airports were benchmarked in an
international comparison in 1995 in order to identify performance gaps between Australian
providers and the best-in-class worldwide (Kincaid and Tretheway 2006).
Proposed by Shleifer (1985) benchmarking may serve for regulatory purposes, widely
known as yardstick competition. This form of regulation implies virtual competition amongst
regulated firms by comparing their cost levels and determining the permitted price based on
an average level. The intention is to stimulate an airport to operate efficiently. Whereas
yardstick competition evolved into a standard approach in the British water and railway
industries, it has to-date rarely been applied to airports. To the best of our knowledge, the
Introduction
20
Dublin Airport Authority (DAA) is the only European example that attempted to implement
yardstick competition in 2001. However, it was highly criticized by airport management for
identifying inappropriate peer airports and was discontinued (Reinhold et al. 2010). The
British Civil Aviation Authority (CAA) argues that the heterogeneous character of airports
and the challenge to obtain appropriate data contribute to their reluctance to apply this type of
economic regulation (CAA 2000).
In order to improve the use of benchmarking and provide a valuable instrument for
managers, governments, regulators and other stakeholders, academic research continuously
aims to refine quantitative methods to assess the productivity and efficiency of airports. The
methods that are mostly applied are outlined below.
1.1.4 Quantitative approaches of benchmarking
Generally speaking, airports may be defined as a network consisting of multi-production
processes. Aeronautical activities include the handling of passengers, aircrafts and cargo. The
non-aeronautical side may operate car parking facilities, restaurants and retail. A number of
quantitative techniques have emerged that assess the productivity and efficiency of decision
making units (DMU) as can be taken from Figure 1.
Fig. 1: Quantitative benchmarking approaches
Productivity and Efficiency Analysis
One-dimensional
Multi-dimensional
Partial Performance
Average Approaches
Frontier Approaches
Non-Parametric(index numbers)
Parametric (Deterministic)
Parametric (Stochastic)
Non-Parametric (Deterministic)
Total Factor Productivity
(TFP)
Ordinary Least Squares (OLS)
Stochastic Frontier Analysis
(SFA)
Data Envelopment
Analysis (DEA)
Source: adapted from von Hirschhausen and Cullmann (2005)
Introduction
21
One-dimensional approaches are the simplest form to assess the productivity by dividing
one output (y) by one input (x). Being skeptical towards sophisticated overall quantitative
techniques, airport managers mostly prefer partial productivity measures. However this
measure should be treated with caution. As discussed by Forsyth et al. (1986) partial measures
should only be applied if data for overall measures is not available. Results obtained from
partial measures can mislead as they fail to capture substitution effects between different input
factors. In order to receive an overall picture of the airport’s performance multi-dimensional
approaches should be applied instead. Three well-documented quantitative methods are often
applied to analyze the productivity and efficiency of government and private enterprises
which are highlighted in grey in Figure 1.
(i) Total Factor Productivity (TFP)
A non-parametric, index number approach is used to measure the total factor productivity
(TFP). The application of index-number approaches is most common in measuring price and
quantity changes over time; the retail price index (RPI) is the most popular economic
indicator. An advantage of index-number approaches is the provision of meaningful results
with only two observations because the productivity is not assessed relative to other units.
The Törnqvist index is widely used in economic studies however the index is restricted to
time-series analyses. Caves, Christensen and Diewert (1982a) proposed a multilateral translog
index known as the CCD index to compare the TFP of a set of units over different years:
∑ ∑
∑ ∑
−+−−+−
−+−−+=
−−−=
i iiijiijiikiik
i iiijiijiikiik
jkjkkj
XXWWXXWW
YYRRYYRR
XXYYTFP
)ln)(ln(21)ln)(ln(
21
)ln)(ln(21)ln)(ln(
21
)ln(ln)ln(lnln
(1.01)
where Yik and Rik are the output quantity and the revenue share for output i of DMU k; iR
is the arithmetic mean of the revenue share and iY is the geometric mean of output i over the
entire sample. Xik is the input quantity and Wik is the input cost share for input i of DMU k; iW
is the arithmetic mean of cost shares and iX the geometric mean of input i over the entire
sample. However, in order to aggregate multiple inputs and outputs to an index, market prices
are required as weights. Furthermore, this measurement assumes that all units operate
efficiently, which is unlikely to be true for airports that are heavily influenced by external
Introduction
22
factors. Instead, frontier approaches as SFA and DEA are more appropriate to estimate an
efficient production or cost frontier.
(ii) Stochastic Frontier Analysis (SFA)
Parametric stochastic frontier analysis (SFA) assesses the efficiency of DMUs utilizing
econometric analysis. The parameters of a production or cost function are estimated with
regression analysis or maximum likelihood estimation. The model of the stochastic
production frontier was first introduced by Aigner, Lovell and Schmidt (1977) and Meeusen
and van den Broeck (1977):
iiii uvxy −+= β')ln( (1.02)
where the scalar ln(yi) is the observed output; xi represents a vector of inputs; β is a vector
of technology parameters to be estimated; vi is the stochastic random error and ui is a non-
negative term for managerial inefficiency.
SFA allows for a separation of the unobservable random error from technical inefficiency
and is based on assumptions regarding to the distributional forms of the efficiency function
and error term. However, prior assumptions need to be carefully made as they may heavily
affect the results as discussed by Stone (2002).
Fig. 2: Stochastic frontier analysis
Source: adapted from Coelli et al. (2005)
Input (x)
Deterministic Frontier (xi’β)
Frontier Output
Output (y)
Inefficiency effect (ui)
Observed Output
Noise effect (vi)
a
b
Introduction
23
Figure 2 illustrates the concept of SFA. The deterministic frontier production function is
described by the function ln(yi)=xi’β. If the DMU is technically efficient (ui=0), the frontier
output is given by the point a, which lies above the deterministic frontier due to a positive
noise effect (vi>0)5. The observed output b includes both the existence of technical efficiency
(ui>0) and a noise effect where a position below the deterministic frontier is given because vi
<ui (Coelli et al. 2005).
Based on the initial cross-sectional6 model by Aigner, Lovell and Schmidt, panel data
models are proposed by Battese and Coelli (1992) that allow for time varying inefficiencies.
To further capture unobserved cross-firm heterogeneity unrelated to technical inefficiency,
Greene (2005) introduces an additional model to shift time-invariant effects to unobserved
heterogeneity whereas the inefficiency term varies over time.
Observed heterogeneity either within or beyond managerial control affects the production
technology or the inefficiency. Early models by Battese and Coelli (1995) assume all firms to
operate under same conditions. To overcome this limitation, Coelli, Perelman and Romano
(1999) propose a model where the heterogeneity affects the shape of the production
technology. Hence each firm will be compared with the most favorable production frontier.
This application might be of relevance in the transportation sector as network characteristics
are likely to affect the production technology.
(iii) Data Envelopment Analysis (DEA)
Non-parametric data envelopment analysis (DEA) measures the relative efficiency of
DMUs utilizing multiple inputs and outputs and will be the major approach applied in this
PhD research. The principle idea of DEA is based on a common ratio measure for assessing
the performance of a DMU, namely dividing the output by an input. In order to consider
multiple inputs and outputs, weights are required to aggregate the variables to virtual inputs
(v1x1o+…+vmxmo) and virtual outputs (u1y1o+…+usyso). The weights of this non-parametric
approach are unknown and calculated from the data using linear programming being solved
for each DMU. The weights are chosen in order to show the specific DMU in as positive a
light as possible, under the restriction that no other DMU, analyzed under the same weights, is
more than 100% efficient (Cooper et al. 2007).
5 A negative noise effect (vi<0) results in a frontier output below the deterministic frontier. 6 Cross-sectional data is a one-dimensional set of variables including N firms in one year whereas panel data contains information on a set of firms over more than one time period.
Introduction
24
By aiming to maximize the weighted combination of outputs to inputs under the constraint
that no DMU exceed the efficiency score, θ, of one, fractional programming can be applied in
order to calculate the input weights, vi, and output weights, ur.
0,
n1,...,j , 1
..
1
1
1
1
≥
=∀≤
=
∑
∑
∑
∑
=
=
=
=
ri
m
iiji
s
rrjr
m
iioi
s
rror
uv
xv
yu
ts
xv
yuMax θ
(1.03)
When maximizing the virtual output and setting the virtual input equal to one (in order to
avoid infinite solutions) fractional programming can be transformed to a linear programming
problem. The second constraint assures that the efficiency score, θ, can not exceed a value
greater then one. The optimal weights calculated in this so-called multiplier form can be
interpreted as shadow prices, hence this form is appropriate when trade-offs between inputs
and outputs are analysed.
0
0
1
..
11
10
1
≥
≤−
=
=
∑∑
∑
∑
==
=
=
ri
m
iiji
s
rrjr
m
iii
s
rror
,uv
xvyu
xv
ts
yuMax θ
(1.04)
The dual of the multiplier form is the envelopment form and is most often applied in
empirical research because it identifies benchmarks for inefficient units.
0,1≥
=
≥
≤
∑θ λ
λy Yλθxs.t. Xλ
θMin
a
a
λ,θ
(1.05)
Introduction
25
where θ is a scalar that estimates the radial contraction of all inputs, i.e. the efficiency
score. λ is a non-negative vector of weights that are determined by the optimization process
and xa and ya are the input and output quantities of DMUo, the airport under investigation. X
and Y represent input and output matrices respectively. Adding the constraint Σλ=1 changes
the assumption of CRS to VRS (Cooper et al. 2007).
DEA was first published in Charnes et al. (1978) under the assumption of constant
returns-to-scale and was extended by Banker et al. (1984) to include variable returns-to-scale.
In contrast to parametric approaches, DEA assumes neither a specific functional form for the
production function nor the inefficiency distribution. Historically, DEA has been primarily
applied to not-for-profit organizations because they could not develop rankings with
profitability indicators. To measure the overall performance, a variety of input and output
quantities have to be covered (Ramanathan 2003). Nowadays, also profit-making
organizations have introduced performance techniques such as DEA to measure the
productivity and efficiency as they concluded that "profit per se is not a good indication of the
potential for improvement within an organization, and because other factors are necessary
for a holistic assessment of performance" (Ramanathan 2003, p.26).
A graphical illustration of DEA is given in Figure 3. With linear programming a Pareto
frontier is attained, bounded by specific DMUs on the envelope of input-output variable
space. The inefficient DMU G is compared to the Pareto frontier with C and D as
benchmarks. In order to become relative efficient and move to G’ on the frontier, G should
radially decrease its inputs.
Fig. 3: Data envelopment analysis
Source: adapted from Coelli et al. (2005)
X2/Y
FE
D
G
G’
X1/Y
C
B
A
Introduction
26
The piece-wise linear frontier which runs parallel to the axes can however cause some
problems to capture all sources of inefficiency. Consider for example DMU A in Figure 3
which was identified to be relative efficient. With the same amount of inputs the DMU could
however produce more outputs in order to operate on the level of DMU B and is known as a
slack problem. These slacks will not be identified in so-called radial models as stated above.
Non-radial models such as the additive model proposed by Charnes et al. (1985) account for
both the desired equi-proportional reductions (expansion) in all inputs (outputs) and any
remaining slacks (Coelli et al. 2005).
Over the years, the basic model is continuously developed. In order to rank efficient
airports and improve the discriminatory power of efficiency estimates, Andersen and Petersen
(1993) introduce the super-efficiency model where airports with rather unique input-output
combinations receive excessively high rankings. A sophisticated approach to reduce the curse
of dimensionality is the principal component analysis (PCA) combined with DEA. PCA-DEA
is applied to replace the original inputs and/or outputs with a smaller group of principle
components (PCs), which explain the variance structure of a matrix of data through linear
combinations of variables with minimal information loss (Adler and Golany 2001, 2002).
Panel data models assess productivity and efficiency changes over time. The most popular
tool is the Malmquist index introduced by Caves, Christensen and Diewert (1982b). Utilizing
DEA with distance functions the approach compares two adjacent time periods with each
other. Different to econometric techniques, non-parametric approaches do not allow for
statistical inference. In order to examine the sensitivity of the estimated frontier,
bootstrapping, a re-sampling technique developed by Efron (1979), is introduced to DEA by
Simar and Wilson (1998, 2000).
Unsurprisingly, numerous studies are concerned to explain efficiency differences across
airports. Amongst other factors ownership forms, hub or size effects and the location are
mostly assumed to substantially impact the airports’ efficiency. Whereas parametric
techniques integrate environmental variables in the production or cost function, DEA may
utilize a two-stage approach where the first-stage efficiency estimates are regressed against a
set of environmental variables in order to evaluate their impact. The advantage of second-
stage approaches is that environmental variables are not included in the DEA model, hence
not affecting the discriminatory power of the first stage.
Introduction
27
1.2 Contribution of this PhD research
Within academic benchmarking a number of studies emerged since the late nineteen-
nineties assessing the productivity and efficiency of airports with DEA, SFA and index
number TFP. To-date DEA proves to be the dominant application requiring neither prior
assumptions on the functional form nor price information to aggregate multiple inputs and
outputs. Common objectives of empirical studies are the examination of efficiency changes
over time or aiming to explain efficiency differences with exogenous factors. Nevertheless,
previous research indicates inconsistencies among the results thereby encouraging future
research. The following section will point out the motivation for this PhD research from the
economic and managerial perspective based on the current literature on airport benchmarking.
Furthermore, the three research articles of this cumulative thesis are summarized thereafter.
1.2.1 Discussions arising from previous research
The wave of airport privatizations in the past two decades motivates the assessment of its
empirical effects however, as in other industries the results are so far rather inconclusive
(Megginson and Netter 2001). Parker (1999) utilizes DEA on the British airports owned by
the BAA covering the periods of pre and post privatization. No evidence is found that full
privatization improves technical efficiency. In contrast, Yokomi (2005) reviews six BAA
airports from 1975 to 2001 utilizing Malmquist DEA. As opposed to Parker, Yokomi finds
that the BAA airports exhibit positive changes in efficiency and technology as a result of the
privatization. The effects of ownership on efficiency are further analyzed by comparing
different ownership forms. Barros and Dieke (2007) analyze 31 Italian airports using DEA in
the first stage and Mann-Whitney hypothesis testing in the second stage, revealing that private
airports operate more efficiently than their partially private counterparts. Lin and Hong (2006)
find no connection between ownership form and efficiency after analyzing a dataset of
worldwide airports utilizing DEA and hypothesis testing. Oum et al. (2006) assess a sample of
100 airports worldwide utilizing variable factor productivity and reach the conclusion that the
productivity of a public corporation is not statistically different from that of a major private
airport. However, airports with major public shares or multiple government involvement
appear to operate significantly less efficiently than other ownership forms. Very often,
changes in ownership form are accompanied by changes towards incentive economic
regulation as for example in Hamburg. Consequently, changes in efficiency may be
attributable to multiple explanations rather than a change in ownership per se. Furthermore
Introduction
28
the competitive environment has changed since airline deregulation thereby putting the
general usefulness of economic regulation into question. Following Vickers and Yarrow
(1991) privatization is not a universal solution and should not be separated from the
economics of competition and regulation which are all determinants of corporate incentives.
While the number of academic benchmarking studies is increasing there is also some
rising resistance especially among airport managers who criticize benchmarking as being of
little use for their business. They are often skeptical towards sophisticated overall quantitative
techniques and prefer the analysis of partial processes. Sarkis and Talluri (2004) propose a
second-stage clustering to identify benchmarks for relatively poor performing airports after
applying DEA. However, not conducting a priori clustering may lead to inappropriate
benchmarks with substantially different resource levels. Referring to Section 1.1.3
geographical constraints, product diversification or the degree of vertical integration are likely
to differ across airports.
1.2.2 Motivation for research
Building on the inconsistencies and discussions from previous research the general
objective of this thesis is twofold. Firstly, the aim is to further the understanding of the airport
industry by exploring the current literature on airport benchmarking and conducting empirical
research. Major emphasis is given on how to explain efficiency differences across airports.
Secondly, the application of airport benchmarking will be improved in order to enhance its
usefulness for managerial, political and regulatory purposes.
This cumulative doctoral thesis is divided into three research articles. The first article
provides a comprehensive survey of the empirical literature of airport benchmarking. It aims
at providing an overview for future research. The quantitative techniques and variables
selected are summarized and discussed, and empirical findings are critically compared.
Thematically, this article heads the empirical research of this thesis as depicted in Figure 4.
Introduction
29
Fig. 4: Structure of this dissertation
The second article applies airport benchmarking from an economic perspective and aims
to explain efficiency differences across Australian and European airports7. Amongst other
exogenous factors, particular emphasis is given on the unresolved questions of the role of
ownership and the necessity of regulation in competitive environments. A semi-parametric
approach is utilized with DEA in the first stage and regression analysis in the second stage.
The third article proposes improvements on the use of benchmarking to identify best-practices
in order to alleviate some of the scepticism of airport managers hold with respect to overall
measurement approaches. An a priori dynamic clustering approach integrated in DEA
7 The list of airports in the sample can be taken from Table 11 in Appendix 3.A, p.98.
Outcome of this dissertation: additional information on efficiency differences across airports and the improvement of benchmarking for managerial and regulatory purposes
Overall research objective of this dissertation: To explore the airports’ heterogeneity with benchmarking and to improve the application of benchmarking to airports
Article 1: Literature Review
Research Question: What can be learnt from existing studies? Methodology: Reviewing the literature with regard to benchmarking methods, data
and results Sample: 58 airport benchmarking studies conducting DEA, SFA and TFP Findings: Steady progress but future research needed to improve comparability
Article 2: Benchmarking from Economic Perspective
Research Question: Does competition matter more than ownership and regulation? What else explains inefficiency?
Methodology: 1st stage: additive DEA 2nd stage: robust cluster regression (OLS, censored and truncated regression)
Sample: 48 European and 3 Australian airports (1998-2007)
Findings: New insights on the role of ownership and regulation
Article 3: Benchmarking from Managerial Perspective
Research Question: Can we really benchmark airports?
Methodology: Network DEA integrating dynamic clustering and PCA
Sample: 43 European airports (1998-2007)
Findings: Improved benchmarking instrument for airport managers and regulators to ensure comparability across airports
Introduction
30
categorizes the airports into homogeneous groups thereby providing appropriate best-practice
airports for inefficient units. This model is applied to a set of European airports8.
Both empirical research articles include a large number of German airports, which have
rarely been under review to-date but offer a rich field for examinations in a European context.
German airports are highly vertically integrated including labour-intensive ground handling
operation whereas airports in most European countries leave this operation to their national
carrier or third party providers. Furthermore, whereas public ownership dominates the
German industry, the airports in Dusseldorf, Frankfurt and Hamburg belong to the first
airports in Europe, which became partly privatized, with equal shares in Dusseldorf and major
public shares otherwise. While cost-plus regulation has been kept at public airports, incentive
regulation was introduced in Hamburg and temporarily in Frankfurt and Dusseldorf with the
intention to increase cost efficiency. However, contrary to the UK where both aeronautical
and non-aeronautical revenues are constrained by the single till approach, the dual till
principle is applied and only aeronautical activities are assumed to possess market power.
Compared to other European countries, Germany is highly populated and airports may face
local competition with nearby airports and other transport modes. Consequently, German
airports appear to differ from other European examples such as non-ex-ante regulated and
fully privatized airports in British monopolistic regions and therefore encourage empirical
research in this respect.
Although this thesis is divided in three independent articles they are connected by the
general purpose to improve the application and usefulness of airport benchmarking. The third
article “Airport Benchmarking from the Managerial Perspective” has been submitted to
Omega and is currently under review. The remaining articles are intended for submission to
leading journals in the fields of regulatory economics such as the Journal of Regulatory
Economics and transportation research such as the Journal of Transport Economics and
Policy. An outline of the articles is presented in the following subsection including main
findings.
8 The list of airports in the sample can be taken from Table 17 in Appendix 4.A, p.131.
Introduction
31
1.2.3 Article 1: A survey of empirical research on the productivity and
efficiency measurement of airports
According to Kincaid and Tretheway (2006) as well as Morrison (2009) the past and
current practice of airport performance analysis lacks in consistency. They state that the clear
definition of an airport model is crucial to understand the industry. Therefore the collection of
essential inputs and outputs describing the airport technology is very important. While
Forsyth (2000) and Graham (2005) provide an overview of a number of airport benchmarking
studies and focus on methodological issues, no comprehensive literature research has been
conducted to-date.
With the publication of a substantial number of benchmarking studies, the aim of this
paper is to provide an overview of previous research that utilizes DEA, SFA and index-
number TFP to assess the productivity and efficiency of airports. In order to further our
understanding of the airport industry this paper reviews how the studies define the airport
model and compares empirical findings with respect to consistency. Further a discussion on
the use of quantitative approaches is provided to examine the progress in airport
benchmarking. A contribution to the literature is a tabular summary covering 60 academic
airport benchmarking studies and a synthesis of inputs and outputs considered in previous
research9.
The survey reveals a number of issues that remain unresolved to-date. Although the
studies aimed to capture the overall performance of an airport, difficulties in gathering
sufficient data on non-aeronautical activities often restrict the model to aeronautical outputs
(passengers, cargo and air transport movements). However, staff employed in commercial
activities is not removed from the sample accordingly. This however may bias the efficiency
results for airports that are highly involved in commercial activities. Generally, data
availability proves to be the most difficult issue in all studies as sufficient data is often not
available to the public. Especially including capital and undesirable outputs such as delay
proves to be difficult however both are crucial for consistent estimations. A comparison of
research findings indicates that whereas increasing commercialization and restructuring lead
to efficiency increases in all studies the findings on ownership and scale effects prove to be
rather inconclusive.
9 see Table 2 to 4 in Appendix 2.A, p. 57 ff. and Figure 10, p. 46.
Introduction
32
Methodologically, the studies indicate substantial progress in utilizing frontier approaches
by considering the heterogeneous character of airports which was found to be essential. DEA
remains the dominant methodology but where large datasets are available an increasing trend
in the utilization of econometric techniques is observed. Overall future research is needed to
further improve airport benchmarking for economic and managerial issues.
1.2.4 Article 2: Joint impact of competition, ownership form and economic
Regulation on airport performance
Whereas previous studies analyze the effects of ownership, regulation and competition
individually, the argument of Button and Weyman-Jones (1992) and Vickers and Yarrow
(1991) is supported in this research that all three factors should be accounted for
simultaneously as their combined impact is likely to affect airport productivity.
The second article aims to discuss whether the deregulation of the airline industry and
changes in airport ownership and management has affected the competitive situation, airport
productivity and pricing behaviour to the extent that the benefits of economic regulation are
potentially unnecessary. Furthermore, such an analysis contributes to the search for the most
desirable combinations.
The dataset covers 48 European airports between 1998 and 2007 and three Australian
airports in a bid to include a sufficiently heterogeneous sample with respect to the ownership
structure, regulatory mechanism and competitive environment. The two-stage analysis
combines DEA in the first stage and regression analysis in the second stage. A semi-
parametric approach is chosen to assess the statistical significance of the exogenous effects.
Banker and Natarajan (2008) demonstrate that two-stage procedures in which DEA is applied
in the first stage and regression analysis in the second stage provide consistent estimators and
outperform parametric one- or two-stage applications. The non-radial additive input-oriented
DEA model is chosen to identify all relative inefficiencies of the inputs. A recent debate in the
literature discusses the most appropriate second stage regression model to be applied when
investigating DEA efficiency estimates. Simar and Wilson (2007) argue that truncated
regression, combined with bootstrapping as a re-sampling technique, best overcomes the
unknown serial correlation complicating the two-stage analysis. Banker and Natarajan (2008)
conclude that simple ordinary least squares, maximum likelihood estimation or Tobit
regression dominate other alternatives. Combining the arguments of Simar and Wilson (2007)
and Banker and Natarajan (2008), robust cluster regression is applied based on ordinary least
Introduction
33
squares in order to account for the correlation across observations. Furthermore, in order to
ensure the robustness of the results, robust cluster Tobit and truncated regressions are also
applied.
The empirical results reveal that under monopolistic conditions, airports should be
regulated to encourage cost efficiency. Dual till price-cap regulation appears to be the most
effective form. Furthermore, airports are likely to exploit market power and set higher
passenger and landing charges. However, gateway or regional competition replaces the need
for economic regulation, thereby supporting the argument that competition rather than
privatization is the key driver of cost efficiency. Nevertheless, unregulated major and fully
private airports within a competitive setting remain profit-maximizers and in this regard may
still require ex-ante regulation.
The regression results prove robust since the outcomes of the robust cluster ordinary least
squares, censored and truncated regressions are very close and the general directions are clear
across all three modelling approaches.
1.2.5 Article 3: Benchmarking airports from a managerial perspective
The majority of studies to date treat the airport technology as a single production process
avoiding the complexity inherent in airport systems as depicted in Figure 5(a). Gillen and Lall
(1997) and Pels et al. (2003) are the first to argue that the airport could be analyzed as two
separate decision-making processes, one serving airside activities and the other serving
landside production (see Figure 5(b)), assuming different returns-to-scale for both sides. Since
the liberalization of the aviation industry however, many airports attempt to increase revenues
from non-aeronautical sources which are not directly related to aviation activities in order to
cross-subsidize aviation charges in turn attracting more airlines and passengers to their airport
(Zhang and Zhang 2010). It is therefore arguable to analyze airports as a single unit due to the
direct complementarities. In order to open the black box of non-parametric approaches
network DEA has been proposed by Färe (1991) and is chosen for this research in order to
separate the complexity of airports into partial production stages.
This research develops a network DEA modelling approach in order to measure the
relative cost and revenue efficiencies of 43 European airports over a ten-year period (1998-
2007) with respect to aeronautical and commercial activities simultaneously, whereby
activities are connected via passengers as the common intermediate product. As illustrated in
Figure 5(c) network DEA recognises the fact that generalized and fixed costs connected to the
Introduction
34
two sets of activities can only be split in an artificial manner and that whilst aeronautical
revenues draw from passengers, cargo and air traffic movements, the non-aeronautical
revenue is more closely tied to passenger throughput. Although airports may have limited
control over traffic volume, non-aeronautical revenues drawn from non-airport related
activities, such as airport cities, are indeed within the purview of airport management.
Fig. 5: Airport production function in DEA
In order to improve the comparability of airports, a dynamic clustering approach (Golany
and Thore 1997) is applied using integer linear programming which forms the reference sets
based on similar mixes of inputs or outputs and intermediate products. Furthermore, the
provision of ground handling is shown to severely affect efficiency estimates leading to a
separation in the comparison of those airports that undertake the process in-house compared
to those that outsource.
Finally, principal component analysis (PCA) combined with DEA (Adler and Golany
2001; Adler and Yazhemsky 2010) is applied in the input-oriented model in order to reduce
the curse of dimensionality and any resulting bias, reducing the set of cost efficient airports
from 53 to 38% in the current application.
A comparison with basic DEA results demonstrates that the additional restrictions in the
network PCA-DEA dynamic clustering formulation lead to more reasonable peer
comparisons, permitting an analysis of strategies which could potentially be adopted over
short and medium term planning horizons. By identifying each airport’s individual reference
set, unique airport outliers influence relative efficiency less severely than occurs under basic
DEA.
(a) Aggregated model
INPUTS All activities (staff, capital,
materials, outsourcing)
OUTPUTS All activities (passengers,
cargo, movements,
non-aeronautical revenues)
(b) Separate model
INPUTS Terminal side (staff, capital,
materials, outsourcing)
INPUTS Airside
(staff, capital, materials,
outsourcing)
OUTPUTS
passengers OUTPUTS
cargo, movements
(c) Network model
INPUTS Both activities
(staff, capital, materials, outsourcing)
INTER-MEDIATE passengers
INTER-MEDIATE
cargo, movements
OUTPUTS Aeronautical
revenues
OUTPUTS Non-aeronautical
revenues
Introduction
35
The model in this research allows airport managers to include their industry knowledge in
the form of limitations on airport size, operating conditions and restricted variability of
capacity encapsulated in the dynamic clustering approach. For example, the results of the
under-utilized airport in Hanover indicate that in the medium-term the airport could either
reduce operations to two of their three existing runways, instead of closing two runways as
obtained with basic DEA. Furthermore, a sufficient number of airports in the data set enable
the application of benchmarking for regulatory purposes. By using an a priori clustering
approach, airports operating under similar conditions may be clustered for analysis thereby
improving the comparison of the airports’ cost level and determining airport charges.
In summary, the methodology provides a number of tools for both exploratory data
analysis and inefficiency estimation, removing the need for additional tests of homogeneity.
1.3 Concluding remarks
With the deregulation of the aviation industry, airport benchmarking became an
important instrument for airports, customers and political institutions. In order to improve its
application a number of academic studies emerged during the last two decades. However,
airports are rather unique in its product diversification and the industry proves to be highly
affected by external heterogeneities that are, at least in the short-term, beyond managerial
control. Hence, meaningful comparison among airports proves to be a difficult task.
The aim of this dissertation was to explore the airports’ heterogeneity with benchmarking
and to improve its application to airports. The comprehensive overview of previous studies
suggests a rather unclear definition of the inputs and outputs that define the production
process to-date and encourages airport stakeholders and academics to undertake further
research. A comparison of empirical findings may give recommendations to airport managers
regarding commercialization and restructuring (in particular ground handling); both proving
to increase the airports’ efficiency.
Empirical research was conducted in this research on European and Australian airports
presenting various DEA models to assess the airports’ efficiency from an economic and a
managerial perspective, including network DEA combined with PCA and dynamic clustering
and the assessment of radial and non-radial models. Regression analyses are utilized to assess
the statistical impact of environmental factors on the DEA efficiency estimates.
Introduction
36
Following previous airport benchmarking studies data availability appears to be most
difficult. It would be extremely helpful if government organizations make data available that
they already collect. Data collection also appears to be a serious issue in this research. After
defining salient variables, the model is substantially reduced in the light of data availability
issues. Nevertheless, the results provide additional information to previous research. The
results of the model combining the impact of ownership, regulation and competition provide
additional information compared to the assessment of individual effects. The empirical results
from this research help to understand the operating behavior of airports under different
institutional settings related to cost efficiency and airport pricing. It therefore provides advice
for policy institutions on the role of ownership and the usefulness of regulation under
different competitive settings. For example the results reveal that for public airports effective
competition provides incentives to operate cost efficient and prevent market power abuse.
Hence administration costs that incur from regulatory procedures may be economized.
Furthermore this dissertation aims to improve the utilization of airport benchmarking from the
managerial perspective. Compared with basic DEA, network DEA formulations combined
with a priori dynamic clustering provide more appropriate benchmarks which enable airport
managers to improve performance in the short and medium-term. In addition, the dynamic
clustering approach might be useful for regulatory purposes in order to determine aeronautical
charges with benchmarking. Capturing external heterogeneities across the airports a priori
ensures a comparison of airports under similar operating environments.
Future research will require substantially more data. For improved managerial
benchmarking, disaggregated data with regard to non-aeronautical activities will help to
identify successful strategies on the commercial side. Further, larger number of airports may
improve the quality of results. Including additional factors, such as scheduling practices, and
political and geographical constraints improves the homogeneity in the clustering approach
and the assessment of efficiency differences across airports. Additional undesirable outputs,
including noise, airport-related delay and air pollution, will enable the development of a social
welfare analysis of airports and the trade-off across the different stakeholders.
Since the liberalization of the aviation industry airport benchmarking has become
increasingly important and will remain a key instrument for managers, political decision
makers and may improve its usefulness for regulatory purposes. Therefore, communication
between management, research and policy in the future is crucial to further improve the
application of airport benchmarking.
37
2 A SURVEY OF EMPIRICAL RESEARCH
ON THE PRODUCTIVITY AND
EFFICIENCY MEASUREMENT OF
AIRPORTS10
Subsequent to the changing nature of the airport industry, benchmarking became popular
for economic and managerial purposes. Within academic benchmarking a number of studies
emerged utilizing parametric and non-parametric approaches. This paper provides a literature
survey on the methods, data and findings of empirical research in order to gain further
understanding of the airport industry. The survey indicates substantial progress in capturing
the heterogeneous character of airports however many issues still remain unresolved. Whereas
increasing commercialization and restructuring contribute to efficiency increases findings on
ownership and scale effects are rather inconsistent. Data availability generally proved to be a
serious issue, in particular the collection of capital inputs and undesirable outputs. To improve
the use of benchmarking airport managers and academics should therefore cooperate and
share their expert knowledge.
10 The author is grateful to Prof. Dr. Hans-Martin Niemeier for helpful discussions and support. The paper is not yet submitted for publication.
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
38
2.1 Introduction
For more than three decades benchmarking serves as an important instrument for
managerial and economic purposes. Benchmarking can be utilized to assess the performance
within and across companies and to estimate productivity and efficiency changes over time.
Hensher and Waters (1993) propose three techniques that are often applied in the overall
productivity and efficiency analysis. Non-parametric index number approaches measure the
total factor productivity (TFP) by aggregating inputs and outputs with market prices.
Parametric approaches such as stochastic frontier analysis (SFA) estimate the efficiency with
regression and separate the error term into random noise and managerial inefficiency. The
non-parametric data envelopment analysis (DEA) utilizes linear programming and divides the
set into relatively efficient and inefficient units without prior knowledge of the functional
relationship between inputs and outputs.
The changing nature of the airport industry during the last decades offers an equally
challenging and interesting objective for applying performance and benchmarking techniques.
Traditionally, airports were managed and regulated as public utilities, which is still present in
many countries. However, in the late eighties a worldwide process of privatization emerged
and was often accompanied by regulatory reforms from rate of return towards incentive and
light-handed regulation. A change in the management style towards commercialization led to
substantial investments in non-aeronautical activities. Furthermore vertical and horizontal
boundaries have changed over time. While some airports outsourced their labour-intensive
activities such as ground handling others have remained highly integrated either as a separate
airport or within an airport system organized as a civil aviation authority. Moreover,
competition between airports began to arise. Some airports like Manchester or London-
Stansted face effective competition from airports in their catchment area. With the
deregulation of the airline industry the airports in Pittsburgh or Brussels lost their hub carrier
and experienced competition with other hub airports. The liberalization of bilateral air service
agreements offered the opportunity to attract international traffic to gateway airports and
secondary hubs. In order to meet future demands, congested airports needed to expand their
capacity and introduce new technologies to increase runway and terminal capacities. However
at many major airports the excess demand was rationed rather inefficiently through queuing
and slot allocation mechanisms increasing the number of non-weather related delays. For
these reasons, airports offer a rich field for performance and benchmarking analyses.
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
39
The application of airport benchmarking is manifold (Kincaid and Tretheway 2006).
Airport managers utilize benchmarking to identify best-practice standards and to develop new
concepts for performance improvements. Customers, shareholders and investors are interested
in using benchmarking as decision-making instrument. National and regional governments
aim to promote their region or assess the effects of political decisions such as privatization.
As proposed by Shleifer (1985) benchmarking may serve for regulatory purposes, widely
known as yardstick competition. This form of regulation implies virtual competition amongst
regulated firms by comparing their cost levels and determining the permitted price based on
an average level. Whereas yardstick competition evolved to a standardized approach in the
British water and railway industry it has rarely been applied to airports to-date11. The Civil
Aviation Authority (CAA) in the UK explains this reluctance with the heterogeneous
character of airports and the challenge to find appropriate data (CAA 2000).
In order to improve benchmarking for practical use academic research continuously aims
to refine quantitative techniques. Since the late Nineties, a number of academic research
studies emerged utilizing quantitative approaches to assess the productivity and efficiency of
airports. Kincaid and Tretheway (2006) as well as Morrison (2009) however challenge the
past and current practice of airport performance analysis for inconsistencies and its limited
value to managers. They state that the clear definition of an airport model is crucial to
understand the industry. Therefore the collection of consistent inputs and outputs is very
important. In addition, different techniques and the airports’ heterogeneous character may
lead to different results across the studies. The last argument is discussed in a response to
Morrison (2009) by Adler et al. (2009) as especially econometric approaches have
substantially been developed to account for observed heterogeneity across decision making
units (DMU). From our point of view, the debate has not resolved all issues and will continue;
also because airport managers still prefer to use simple partial measures while academics
utilize sophisticated overall productivity and efficiency approaches.
In other transport sectors that are traditionally under public ownership we find
comprehensive surveys by Oum et al. (1999) on the rail sector, de Borger et al. (2002) on
public transport and Gonzalez and Trujillo (2009) on seaports. While Forsyth (2000) provides
an overview on a few airport benchmarking studies and Graham (2005) focuses on
11 To the best of our knowledge, the Dublin Airport Authority (DAA) is the only European example that has been subject to regulation with yardstick competition by the Commission Aviation for Regulation in 2001. However, it was highly criticized by the airport for inappropriate peer airports that have been identified (Reinhold et al. 2010).
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
40
methodological issues, comprehensive research has not been undertaken to-date that
summarizes the empirical studies and discusses the findings. With the publication of a
substantial number of academic benchmarking studies, the aim of this paper is therefore to
provide an overview of the current literature that utilizes DEA, SFA and TFP to assess the
productivity and efficiency of airports. In order to further our understanding of the airport
industry this paper reviews the variable selection in previous studies and compares empirical
findings. Furthermore, a discussion on the use of quantitative approaches is provided to
examine the progress in airport benchmarking. A contribution to the literature is a tabular
summary covering 60 academic airport benchmarking studies and a synthesis of inputs and
outputs considered in empirical research. The survey reveals a number of issues that remain
unresolved to-date. Whereas increasing commercialization and restructuring consistently lead
to efficiency increases, findings on ownership and scale effects prove to be rather
inconclusive. Furthermore, consistent capital measures and undesirable outputs are crucial for
analyses to prevent an overestimation of the efficiency results but are often problematic to
obtain. Although the studies indicate substantial progress in utilizing frontier approaches by
considering the heterogeneous character of airports, more research is needed to improve
airport benchmarking for regulatory purposes and to alleviate some of the scepticism of
airport managers.
The paper is organized as follows; Section 2.2 introduces the methodology of performance
measurement and reviews their application in airport benchmarking studies. Section 2.3
discusses the inputs and outputs that have been used. Section 2.4 continues with a comparison
of findings from research studies. Special attention will be drawn on the results of
productivity and efficiency changes over time, ownership and scale effects. Concluding
remarks are given in Section 2.5 on the issues that still remained unresolved.
2.2 Productivity and efficiency measurement concepts
In the media productivity and efficiency are often used interchangeably. However, this is
somewhat misleading as both terminologies are defined differently. The following section
introduces the meaning of productivity and efficiency and provides an overview of
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
41
measurement techniques. We then continue with the concepts of price index-based
approaches, DEA and SFA as the dominant approaches in the field of airport benchmarking12.
2.2.1 Productivity and efficiency
The productivity of an airport may be calculated as the ratio of output(s) per input(s).
Partial productivity measures divide for example the number of passengers by the number of
employees. In order to include the capital productivity and assessing total factor productivity,
multiple inputs and outputs are aggregated to an index. Technical efficiency on the other hand
defines the comparison of the observed outputs (inputs) to its optimal values while holding
the inputs (outputs) constant. The seminal paper on efficiency by Farrell (1957) defines three
different forms of efficiency. A firm is technically efficient if it operates on the production
frontier. According to Figure 6, airport P is technically inefficient and ought to change its
input/output combination in order to reach the frontier at Q. An airport that chooses the input
mix which produces a given output quantity at minimum costs is said to be allocative efficient
and operates at R. Technical and allocative efficiency in combination define the economic
efficiency where the firm is perfectly efficient and operates at Q’ (Coelli et al. 2005).
Fig. 6: Technical, allocative and economic efficiency
Source: adapted from Coelli et al. (1998)
Over the years a number of quantitative techniques emerged which assess the productivity
and efficiency of decision making units (DMU) such as airports (see Figure 7). The one-
dimensional approach is the simplest form to assess the productivity by dividing one output
by one input. However this measure should be treated with caution. As discussed by Forsyth
12 This paper does not explain the methodological approaches in detail. For further information we suggest the following literature: Coelli et al. (2005) and Fried et al. (2008) for an overview of productivity and efficiency analysis and Kumbhakar and Lovell (2000) and Cooper et al. (2007) for an advanced application of SFA and DEA respectively.
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
42
et al. (1986) partial measures should only be applied if data for overall measures are not
available. Results obtained from partial measures can mislead as they fail to capture
substitution effects between different inputs. For this reason academic research (Graham and
Holvad 2000; Oum et al. 2003, 2004) use partial productivity measures only in addition to
overall approaches.
Fig. 7: Quantitative methods in productivity and efficiency analysis
Productivity and Efficiency Analysis
One-dimensional
Multi-dimensional
Partial Performance
Average Approaches
Frontier Approaches
Non-Parametric(index numbers)
Parametric (Deterministic)
Parametric (Stochastic)
Non-Parametric (Deterministic)
Total Factor Productivity
(TFP)
Ordinary Least Squares (OLS)
Stochastic Frontier Analysis
(SFA)
Data Envelopment
Analysis (DEA)
Source: adapted from von Hirschhausen and Cullmann (2005)
In order to gain an overall measure of the airport’s performance multi-dimensional
approaches should be applied instead which can be distinguished in frontier and average
approaches. The most popular average approach is linear regression with ordinary least square
(OLS). The availability of price information enables the utilization of price index-based
number (PIN) approaches to measure the total factor productivity. Anyhow average
techniques assume that all DMUs operate efficiently. Frontier approaches in contrast estimate
the efficient production or cost function where an airport that deviates from the frontier
appears to be inefficient. The pioneering work on efficiency by Farrell (1957) was taken up
by Charnes, Cooper and Rhodes (1978) who developed data envelopment analysis based on
linear programming and inspired the econometricians Aigner, Lovell and Schmidt (1977) to
assess the technical efficiency with stochastic frontier analysis.
2.2.2 Price-based index number approaches
The application of index-number approaches is most common in measuring price and
quantity changes over time. A popular economic indicator is the consumer price index (CPI),
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
43
measuring price changes of goods and services over time. In order to measure total factor
productivity with index numbers, market prices are required to weight and aggregate multiple
input and output quantities. However, information on market prices is often not available, and
instead cost and revenue shares are used. An advantage of index-number approaches is the
provision of meaningful results with only two observations because the productivity is not
assessed relative to other units. The Törnqvist index is widely used in economic studies
however the index is restricted to time-series analyses. Caves, Christensen and Diewert
(1982a) proposed a multilateral translog index known as the CCD index to compare the TFP
of a set of airports over different years:
∑ ∑
∑ ∑
−+−−+−
−+−−+=
−−−=
i iiijiijiikiik
i iiijiijiikiik
jkjkkj
XXWWXXWW
YYRRYYRR
XXYYTFP
)ln)(ln(21)ln)(ln(
21
)ln)(ln(21)ln)(ln(
21
)ln(ln)ln(lnln
(2.01)
where Yik and Rik are the output quantity and the revenue share for output i of DMU k; iR
is the arithmetic mean of the revenue share and iY is the geometric mean of output i over the
entire sample. Xik is the input quantity and Wik is the input cost share for input i of DMU k; iW
is the arithmetic mean of cost shares and iX the geometric mean of input i over the entire
sample.
To-date a limited number of airport benchmarking studies apply index number
approaches. Nyshadham and Rao (2000), Hooper and Hensher (1997) and Vasigh and
Gorjidooz (2006) utilize the CCD index to measure the airport’s TFP, while Oum and Yu
(2004) and Oum et al. (2006) assessed the variable factor productivity (VFP) with the same
index but without capital input. Instead of market prices all studies included cost and revenues
shares as weighting factors.
As aforesaid, index numbers assume technical efficiency for all observations which is
unlikely to be true for airports. Instead environmental constraints, ownership forms, the
regulatory procedure or the market structure which are all exogenous to the airport
management are expected to affect the efficiency. Alternatively non-parametric frontier
approaches may be applied to assess TFP changes. The most popular model in airport
benchmarking is the DEA-based Malmquist index decomposing TFP changes into pure
technical efficiency, scale efficiency and the technological changes. Airport studies utilizing
Malmquist DEA will be introduced in the following section.
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
44
2.2.3 Data envelopment analysis
The majority of airport benchmarking studies apply data envelopment analysis. An
advantage of this approach is that multiple inputs and outputs are aggregated without price
information. Furthermore, DEA neither require the specification of a production or cost
frontier nor assumptions of the distribution of the error term. The mathematical framework of
DEA has first been proposed by Charnes, Cooper and Rhodes (1978) to assume constant
returns-to-scale (CRS) and has been extended by Banker, Charnes and Cooper (1984) to
include variable returns-to-scale (VRS). With linear programming, DEA compares each DMU
to the efficient set of observations, with similar input and output ratios. This non-parametric
approach solves the linear programming formulation for each DMU and the weights assigned
to each linear aggregation are the results of the corresponding linear program. The weights are
chosen in order to show the specific DMU in as positive a light as possible, under the
restriction that no other DMU, analyzed under the same weights, is more than 100% efficient.
Variable returns-to-scale have mostly been assumed as the size of an airport may substantially
differ within the sample set (Abbott and Wu 2002; Barros 2008a; Bazargan and Vasigh 2003;
Fernandes and Pacheco 2002). Gillen and Lall (1997) who divided the airport into terminal
and airside activities suggest VRS for passengers and CRS for movements. Furthermore, a
number of studies compared the results of CRS and VRS assumptions in order to obtain the
level of scale efficiency (Abbott and Wu 2002; Barros and Dieke 2007; de la Cruz 1999; Lam
et al. 2009; Murillo-Melchor 1999). Formulation (2.02) presents an input-oriented model
assuming variable returns-to-scale.
0,1≥
=
≥
≤
∑θ λ
λy Yλθxs.t. Xλ
θMin
a
a
λ,θ
(2.02)
where θ is a scalar that estimates the radial contraction of all inputs. λ is a non-negative
vector of weights that are determined by the optimization process and xa and ya are the input
and output quantities of DMUo, the airport under investigation. X and Y represent input and
output matrices respectively. Adding the constraint Σλ=1 changes the assumption of CRS to
VRS (Cooper et al. 2007).
Over the years, the basic model has been continuously refined. It includes the formulation
of non-radial models which reflect all inefficiencies (including slacks) identified in the inputs
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
45
and outputs, the estimation of efficiency changes over time, the improvement of the
discriminatory power of efficiency estimates, the formulation of weight restrictions, the
introduction of network DEA to open up the black box of DEA, the identification of outliers
and the introduction of statistical inference to DEA (Cooper et al. 2007). Figure 8 summarizes
DEA applications which are utilized in airport efficiency studies.
Fig. 8: Models in data envelopment analysis
Data EnvelopmentAnalysis
Cross-section,pooled
PanelData
Additionalapplications
Basic DEA(CCR, BCC)
Other models(e.g. SBM, FDH,
additive, network DEA)
Malmquistindex
WindowAnalysis
Statistical inference
(e.g. Bootstrapping)
Ranking methods(super-, cross-
efficiency)
Improve Discrmination(e.g. PCA-DEA)
Abbott and Wu (2002), Barros and Sampaio (2004),
Bazargan and Vasigh (2003), de la Cruz (1999), Fernandes
and Pacheco (2002), Gillen and Lall
(1997), Martín and Román (2001), Pacheco and
Fernandes (2003), Pacheco et al. (2006), Parker
(1999), Pels et al. (2003), Pels et al.
(2001), Vogel (2006), Yoshida and
Fujimoto (2004)
Holvad and Graham (2000), Lam et
al. (2009), Adler and Liebert (2010), Adler
et al. (2010)
Abbott and Wu (2002), Barros and
Assaf (2009), Barros and Weber (2009), Chi-Lok and Zhang (2008), Fung et al. (2008), Gillen and
Lall (2001), Murillo-Melchor (1999),
Yokomi (2005)
Yu (2004) Assaf (2010a), Barros (2008a),
Barros and Dieke (2008),
Barros and Dieke (2007), Lin and Hong
(2006), Martín and Román (2008),
Martín and Román (2006), Sarkis
(2000), Sarkis and Talluri (2004)
Adler et al. (2010)
From the beginning, data collection appeared to be a serious issue and the sample size was
mostly rather small. However, a low ratio of observations to the number of inputs and outputs
weakens the discriminatory power of DEA. For example, Parker (1999) assesses the technical
efficiency of the BAA as a single unit over a period of seventeen years with three inputs and
two outputs. He reveals an average efficiency score of 96%, where nine years ought to be
efficient. In order to rank efficient airports, Andersen and Petersen (1993) introduce the super-
efficiency model where airports with unique input-output combinations receive excessively
high rankings and are identified as outliers. Several airport studies apply models to rank
efficient airports in order to improve the discrimination (Adler and Berechman 2001; Sarkis
and Talluri 2004; Barros and Dieke 2007). In order to avoid too many efficient airports
Bazargan and Vasigh (2003), Lam et al. (2009) and Martín and Román (2006) include a
virtual efficient airport which possesses the lowest input and highest output values of the
sample. A sophisticated approach is principal component analysis (PCA) integrated in DEA.
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
46
PCA-DEA is applied to replace the original inputs and/or outputs with a smaller group of
principle components (PCs), which explain the variance structure of a matrix of data through
linear combinations of variables with minimal information loss (Adler and Golany 2001,
2002). Adler et al. (2010) utilizes PCA-DEA to reduce four inputs to two PCs which explain
more than 85% of the variance in the original data.
Panel data models allow the assessment of productivity and efficiency changes over time.
The most popular tool within DEA is the Malmquist index which was introduced by Caves,
Christensen and Diewert (1982b). Distance functions compare two adjacent time periods and
TFP changes are decomposed into pure technical efficiency change, scale efficiency change
and technological change. Malmquist DEA is applied in a number of studies in order to
disentangle technical and efficiency changes (see Figure 8). An alternative to account for
efficiency changes over time is window analysis where different sets of pooled data with
overlapping time periods are assessed to observe a trend in efficiency changes. This model is
a trade-off between solving one aggregate model with pooled data and estimating each time
period separately. Window analysis is utilized by Yu (2004) with a two-year window.
A number of studies are concerned with explaining efficiency differences across airports.
Amongst others ownership forms, hub or size effects and the location are assumed to
substantially impact the relative efficiency results. DEA offers various models to assess the
impact of exogenous effects on the efficiency estimates. The majority of DEA studies utilize a
second-stage regression where the first-stage DEA efficiency estimates are regressed against a
set of environmental variables in order to evaluate its significance. The advantage of second-
stage approaches is that environmental variables are not included in the DEA model, hence
not affecting the discriminatory power of the first-stage. A number of studies adopted
(censored) Tobit regression where the DEA estimate is censored at the value of one (Abbott
and Wu 2002; Barros and Sampaio 2004; Chi-Lok and Zhang 2008; Gillen and Lall 1997).
Arguing that DMUs only appear to be relative efficient due to biased estimations Simar and
Wilson (2007) proposed truncated regression which drops the efficient units from the sample.
This recent approach has been adopted by Barros (2008a) and Barros and Dieke (2008).
Non-parametric Mann-Whitney and Kruskal-Wallis tests assess the significance of
efficiency differences among various groups. Bazarghan and Vasigh (2003) apply both tests
to assess efficiency differences between public and private airports and Graham and Holvad
(2000) conduct Mann-Whitney tests on Australian and European airports. Adler et al. (2010)
utilize the four-step program evaluation procedure by Brockett and Golany (1996) and
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
47
Sueyoshi and Aoki (2001) which incorporates a Kruskal-Walis test in the last stage to assess
efficiency differences between ground handling providing and non-providing airports.
DEA normally treats the production technology as a black box where no functional
relationship is assumed between input and output quantities. To overcome this shortcoming,
Färe (1991) and Färe and Grosskopf (1996; 2000) propose network-DEA which allows an
analysis of the optimal production structure of DMUs, to determine both efficient subsystems
and overall efficiency. Adler et al. (2010) extended the idea of Gillen and Lall (1997) to
separate both operational sides for the assumption of different scale effects. The efficiency of
aeronautical and commercial activities is estimated separately in a single model and both
sides are connected via intermediate products.
Furthermore, DEA does not allow for statistical tests. The estimated technical efficiency
only depends on the observed sample. It may be a matter of concern to examine the sensitivity
of the estimated frontier and how the efficiency estimates correspond to changes in the
sample. The introduction of bootstrapping, a re-sampling technique developed by Efron
(1979), offers statistical inference and hypothesis testing into DEA as proposed by Simar and
Wilson (1998, 2000). In airport studies bootstrapping is still in its early stages. Assaf (2010a)
applied DEA combined with bootstrapping to evaluate and test scale efficiency among UK
airports and found differences between original and bootstrapped results.
2.2.4 Stochastic frontier analysis
An advantage of SFA over deterministic approaches is that it does not purely explain
inefficiency as mismanagement rather than considering a stochastic random error which
accounts for a not observable relation between the inputs and the output. The parametric
frontier approach was first independently proposed by Aigner, Lovell and Schmidt (1977) and
Meeusen and van den Broeck (1977):
iiii uvxy −+= β')ln( (2.03)
where the scalar ln(yi) is the observed output; xi represents a vector of inputs; β is a vector
of technology parameters to be estimated; vi is the stochastic random error and ui is the term
for managerial inefficiency. Over the years the basic model has been continuously refined. It
includes the introduction of panel data models that in addition may capture unobserved and
observed heterogeneity. Furthermore, distance functions rather than traditional Cobb-Douglas
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
48
or translog production functions allow multiple outputs in a single production model. Figure 9
illustrates the most important models that are applied to airports.
Fig. 9: Models in stochastic frontier analysis
Stochastic FrontierAnalysis
Cross-section,pooled
PanelData
Time-invariant
Time-varying
Chow and Fung 2009 (M;T;P) Pels et al. (2001)
Homogeneous function
Homogeneous function,
unobserved heterogeneity
Heterogeneous function
Heterogeneous function,
unobserved heterogeneity
Barros (2008c), Oum et al. (2008)
Assaf (2009b), Barros (2008b),
Martín et al. (2009), Martín and Voltes-
Dorta (2009), Pels et al.
(2003),Tovar and Martín-
Cejas (2010), Tovar and Martín-
Cejas (2009)
Assaf (2008) Barros (2009), Barros and Marques
(2008)
In early stages of airport benchmarking SFA has rarely been utilized due to its sensitivity
to small sample sizes and the requirement to specify a functional relationship (Assaf 2010a).
Today, datasets become larger and its application has been established. Initial studies in SFA
estimated a production function with physical input data of the airport infrastructure such as
the number of gates or the terminal size in order to assess the efficient use of an airport (Pels
et al. 2001, 2003). Nowadays, it is often emphasized to collect operating and capital cost data
in order to assess the cost efficiency (Assaf 2010b; Martín et al. 2009; Martín and Voltes-
Dorta 2007). Allowing for more flexibility, translog functions are preferred over Cobb-
Douglas functions (Barros 2009; Chow et al. 2009; Martín et al. 2009; Oum et al. 2008; Tovar
and Martín-Cejas 2009). Input distance functions are utilized by Tovar and Martín-Cejas
(2009, 2010) however output distance functions are not considered to-date which we find an
important issue for an industry producing multiple aviation and commercial outputs. Instead
the outputs are aggregated to operational income (Assaf 2008) or separated into different
output models (Gillen and Lall 1997; Pels et al. 2003).
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
49
Whereas Chow et al. (2009) examine cross-sectional data with 46 observations the
remaining studies utilize panel data models in order to obtain an increasing number of
observations. Panel data models are initially developed by Pitt and Lee (1981) and Schmidt
and Sickles (1984) where the inefficiency is assumed to be time-invariant. To relax this
restriction, Cornwell, Schmidt and Sickles (1990) and Battese and Coelli (1992) introduce
time-varying inefficiency models where the latter model is dominantly applied in airport
benchmarking (Assaf 2010b; Chow et al. 2009; Pels et al. 2003; Tovar and Martín-Cejas
2009, 2010). To capture cross-firm heterogeneity which is not related to technical inefficiency
Greene (2005) further refines the model of Schmidt and Sickles (1984) and Battese and Coelli
(1992). Time-invariant firm-specific effects are allocated to a parameter explaining
unobserved heterogeneity whereas the inefficiency term allows a variation over time.
Unobserved heterogeneity has been considered in airport studies by Barros (2008c) and Oum
et al. (2008). Based on the formulations of Battese and Coelli (1992) the latent class model by
Orea and Kumbhakar (2004) conducts a clustering of the data set into different classes where
the class membership remains unknown to the analyst. Barros (2009) applies this formulation
on UK airports by defining his classes according to the market share based on passenger
volume.
In contrast to unobservable heterogeneity, observed heterogeneity is reflected in
exogenous effects that contribute to inefficiency. An important issue in parametric research is
whether exogenous factors impact the production technology or the inefficiency term. Early
studies in this area assume a homogenous production technology for all airports as proposed
by Battese and Coelli (1995) where environmental variables affect the inefficiency (Chow et
al. 2009; Pels et al. 2003; Tovar and Martín-Cejas 2009). Heterogeneous models that account
for different production technologies across airports were for instance applied by Oum et al.
(2008). They assume ownership to affect the technical efficiency whereas the shares of
international traffic and cargo are likely to shift the cost function. Assaf (2008) utilizes a
meta-frontier approach by O’Donnell et al. (2007) accounting for technological differences
between small and large airport in UK. The airport’s efficiency is estimated against its own
technology and a meta-frontier which envelopes the technology of both small and large
airports.
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
50
2.2.5 Comparison of techniques
Reviewing the different methodologies introduced in Section 2.2.2 to 2.2.4 it appears that
the decision for a technique remains difficult (see Table 1). DEA and SFA offer various
formulations and indicate substantial progress since their introduction in the late Seventies. In
general, frontier approaches appear to be appropriate for airport benchmarking because they
do not assume that all airports operate efficiently. Further, they do not necessarily require
market prices to aggregate multiple inputs and outputs. If utilizing benchmarking as a
management instrument in order to identify best practice standards, DEA provides
information on composite benchmarks which help to reach the target inputs or outputs. A
major drawback of DEA and SFA however is the requirement of large datasets in order to
obtain fairly robust efficiency estimates. In addition, DEA is very sensitive to outliers as the
efficient frontier is constructed by the data points. Although SFA benefits from the separation
of random noise from managerial inefficiency it requires prior information how to disentangle
the stochastic error and how to specify the functional relationship between inputs and outputs.
Both assumptions may heavily affect and bias the results as argued by Stone (2002).
Tab. 1: Comparison of DEA, SFA and PIN properties
Category DEA SFA PIN Method and specifications
Parametric approach No Yes No Assumes that all airports are efficient No No Yes Accounts for random noise No Yes No Types of measurement:
- Technical efficiency Yes Yes No - Allocative efficiency Yes Yes No - Technical change Yes Yes No - Scale effects Yes Yes No - TFP change Yes Yes Yes
Data requirements Type of data:
- Cross-sectional Yes Yes Yes - Time series No No Yes - Panel Yes Yes Yes
Robust estimates with small number of observations No No Yes Variable requirement:
- Input and output quantities Yes Yes Yes - Input and output prices No No Yes
Can include exogenous variables Yes Yes No Results
Provides full ranking No Yes Yes Provides composite benchmarks Yes No No Conventional hypothesis testing No Yes No
Source: own compilation adapted from Coelli et al. (1998; 2003; 2005)
In price index number approaches in contrast two observations already provide
meaningful productivity measures. However, it requires information on market prices to form
an input and output index which are often not available for airports. To-date DEA has proven
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
51
to be mostly been applied in airport benchmarking which may be explained with the
requirement of fewer assumptions than SFA and the use of multiple inputs and outputs
without price information.
In order to ensure the reliability and verifiability of efficiency estimations the German
regulator for the electricity sector Bundesnetzagentur applies both DEA and SFA (Reinhold et
al. 2008). However, based on empirical studies that compare the efficiency estimates of DEA
and SFA, the results are rather inconsistent (see Ferrier and Lovell 1990 and Bauer et al. 1998
for the banking sector).
In short, the choice of the technique heavily depends on various criteria including the
availability of data, the object of research and prior knowledge of the airport technology that
may be included in the estimations. The following section will continue with a review on the
choice of inputs and outputs in airport benchmarking studies.
2.3 Variable selection in airport studies
Following Coelli et al. (2003, p.83) “Irrespective of which methodology […] -PIN, SFA or
DEA- it cannot avoid the first rule for empirical economics: garbage in = garbage out”. The
data included in a formulation should clearly explain the underlying airport technology. We
find a broad consensus that capital, labour, materials and other external services are
theoretically necessary to handle traffic volume and selling non-aeronautical products. The
following section discusses the inputs and outputs that are included in academic airport
benchmarking studies so far (see Figure 10).
2.3.1 Outputs
Prior airline deregulation airports were often considered as public utilities focussing on
the provision and operation of the infrastructure that handles passengers, cargo and aircrafts
as common outputs13. Hence the objective of an airport was to offer a good level of service
irrespective of commercial and financial purposes. Early studies therefore often limited their
evaluation to an efficient use of the airside infrastructure (Abbott and Wu 2002; Fernandes
and Pacheco 2002; Gillen and Lall 1997; Martín and Román 2001; Murillo-Melchor 1999;
13 Martín et al. (2009) and Martín-Cejas (1999) for example further combine passengers and cargo to work load units (WLU) however, given that both are different in their charging, use of resources and yield in revenues they should not be seen as an equal output to the airport.
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
52
Pels et al. 2001) thereby ignoring non-aviation activities. The deregulation affected the airport
industry by increasing airport competition, changes in ownership forms and the shift towards
incentive and light-handed regulation. Airports began to operate as commercial enterprises
where financial targets became more important. A regulated non-congested airport in a
competitive environment may seek to maximize commercial revenues in order to cross-
subsidize aeronautical charges and thereby attract passengers and airlines to their airport
(Zhang and Zhang 2010). Monopolistic and unregulated private airports may be interested in
optimizing overall income. Unless a quantity index is designed, non-aeronautical products are
expressed in revenues (Barros 2008b; Bazargan and Vasigh 2003; de la Cruz 1999; Hooper
and Hensher 1997; Oum et al. 2006; Pacheco et al. 2006; Yokomi 2005) and are an inherent
part of an airport to-date. Therefore it is crucial to consider this activity in overall airport
analyses unless the input side can be clearly separated into aeronautical and non-aeronautical
activities.
Similar to manufacturing companies airport generate undesirable by-products. Noise and
pollution affect the environment and delays decrease the service quality. Consequently, they
contribute to a decrease in the airport’s performance. Pathomsiri et al. (2008) analyse US
airports and include delays as a negative (undesirable) output. Their results clearly indicate
that ignoring the quality of airport services would otherwise overestimate technical efficiency
gains of airports with higher utilization. To the best of our knowledge information on delays
such as the average delay per movement are not publicly available for a large sample of
European airports. However, the costs of delays to an airline have been estimated by Cook et
al. (2004) and may be considered as a negative output to airports that cause heavy delays.
Furthermore, Yu (2004) integrates aircraft noise as an undesirable output and again concluded
that undesirable outputs severely affect the technical efficiency of airports.
2.3.2 Inputs
Input quantity measures on the airport infrastructure were already available in early stages
of airport benchmarking where financial targets remained secondary. The number of
employees is collected as an operating input, and the numbers of gates, runways or the
terminal size are collected to measure capital. Materials and other outsourced services such as
ground handling are usually expressed as other operating expenditures unless designing a
quantity index (Assaf 2008; Bazargan and Vasigh 2003; Hooper and Hensher 1997; Lam et
al. 2009; Martín et al. 2009; Oum et al. 2003; Pacheco and Fernandes 2003; Sarkis 2000).
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
53
Fig. 10: Inputs and outputs in previous airport benchmarking studies
The consideration of staff varies among the studies. Arguing that the airport industry is
rather heterogeneous, Pels et al. (2001, 2003), Yoshida (2004) and Fung et al. (2008) ignore
labour in their model to avoid an inappropriate comparison between vertically integrated
airports and airports that outsource labour-intensive operations. The number of employees is
mostly reported in annual reports or other public sources and therefore used by a number of
Inputs
Labour • No. of employees
o Heads o Ordered by qualification
• No of. full time equivalents • Staff costs
Capital • Physical infrastructure
o Airport area o Car parking area, no. of car
parking spots o Passenger terminal area, no. of
passenger terminals o No. of check-in desks, area
departure lounge, no. of gates o Baggage claim area, no. of
baggage collection belts o Cargo handling facility area o Runway area, runway length, no.
of runways o Apron area, no. of terminal
bridges, no. of remote stands • Monetary measure
o Capital stock (PIM) o Capital value (book value) o Capital invested o Interests on net assets o Amortization o Rate-of-return on net capital stock
• Capacity measure o Terminal capacity (no. of
passengers per hour) o Declared runway capacity (no. of
movements per hour)
Other • Average access costs • Distance to city centre
Outputs
Desirable outputs Aeronautical side
• Physical measure o No. of passengers
Domestic International
o No. of air transport movements (ATM) Commercial Commuter, general aviation,
military o Work load units (WLU)
• Monetary measure o Aeronautical revenues
Landing revenues Passenger revenues Parking revenues Handling revenues
• % of on-time operations Non-aeronautical side
• Non-aeronautical revenues
Undesirable outputs Delays:
• Delayed air transport movements • Time delays
Noise:
• Aircraft noise (surcharge)
Materials and outsourcing • Other operating costs
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
54
studies (Abbott and Wu 2002; Lin and Hong 2006; Murillo-Melchor 1999; Sarkis 2000;
Tovar and Martín-Cejas 2009; Yokomi 2005; Yoshida and Fujimoto 2004). This figure
however does not distinguish between full-time and part-time employees and may bias results
among airports with different vertical structures. Other studies collected information on full-
time equivalents but this figure is often publicly not available (Assaf 2010a; Martín and
Voltes-Dorta 2007; Oum et al. 2008). Information on staff costs were utilized either in
combination with staff quantities to design an index (Assaf 2010b; Barros 2008b; Martín et al.
2009; Oum et al. 2008) or to differentiate between unskilled, skilled and management staff
(Hooper and Hensher 1997; Martín and Román 2001). Cross-country comparisons however
require adjustments for different staff costs levels.
Airports are a typical example for an industry with lumpy investments such as the runway
system (Golaszewski 2003). Consequently, capital plays a major role in airport
benchmarking. For example, capital requires consideration in the examination of economies
of scale, in order to assess an efficient use of airport infrastructure, for regulatory purposes
and to decide investment projects. The measurement of capital input however appears to be a
serious issue in airport benchmarking. Due to lack of data Oum and Yu (2004) and Oum et al.
(2006) ignore capital and assess the variable factor productivity instead. A number of studies
collect physical information of the infrastructure such as the number of runways, gates and
check-in counters (Barros 2008a; Gillen and Lall 1997; Pels et al. 2003; Oum et al. 2003).
This capital measure seems to be appropriate in order to assess the technical efficiency and for
cross-border studies thereby avoiding adjustments of financial data by different national
accounting procedures. However a simple linear aggregation of these variables is problematic,
for example because the number of runways does not include information on the
configuration, weather impacts or environmental restrictions. As an example the runway
system substantially varies in their configuration thereby making a comparison of the number
of runways problematic. To overcome this problem the terminal and declared runway
capacity can be used as a proxy of capital for estimating the technical efficiency (Adler et al.
2010). The advantage is that a capacity measure considers the entire infrastructure
configuration in one measure and improves comparability. Studies aiming to examine
financial issues measure the monetary value of physical capital. Various studies included
amortization and interests or the book value of fixed assets as a measure of capital costs
(Barros and Weber 2009; Martín and Voltes-Dorta 2007) however different depreciation
procedures and expected useful lives across the countries may affect the results. For example,
the airports of the British Airports Authority depreciate their runways over 100 years whereas
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
55
the airports operated by the Aéroports de Paris apply a 10 to 20 year depreciation rule
(Graham 2005). To improve comparability among the monetary value of physical capital,
Abbott and Wu (2002) and Hooper and Hensher (1997) apply the perpetual inventory method
(PIM) proposed by Christensen and Jorgenson (1969) for Australian airports to estimate the
gross fixed capital stock. The approach estimates the costs of historical capital investments
which are not fully depreciated yet and adjusts the value by inflation. Yet collecting data on
past capital investments and information on expected useful life and depreciation methods are
very time-consuming, especially for cross-border studies.
In summary, supporting Kincaid and Tretheway (2006) and Morrison (2009) the selection
of inputs and outputs needs to be carefully considered. It should be attempted to include
inputs that are connected to all outputs defined in the model. An imbalance would otherwise
distort the results. Ground-handling for example is a labour intensive operation where the
output is measured in revenues. DEA-studies that restrict the output side to airside activities
(i.e. passengers, cargo and movements) will automatically obtain substantially lower
efficiency estimates for airports providing ground handling unless staff employed in this
activity is removed from the input side. According to Adler et al. (2010) ground-handling
providing airports appear 10% less cost efficient than its outsourced counterparts if revenues
generated from ground handling are not included but staff is not adjusted accordingly.
So far, we discover that the empirical studies on airport benchmarking apply various
parametric and non-parametric approaches with different underlying assumptions of the
airport production technology. Furthermore the variables to describe the airport substantially
vary and are often subject to data availability. Consequently, we are concerned that different
approaches and variables may reveal inconclusive results. The following section continues
with a comparison of empirical findings from airport studies.
2.4 Empirical results of productivity and efficiency studies
The object of research varies among the research studies. In order to evaluate an efficient
use of the airport infrastructure the technical efficiency is assessed with physical input and
output quantities. Productivity and efficiency changes over time are estimated in order to
capture technological progress. Furthermore, it is often aimed to explain efficiency
differences across airports with factors that are at least in the short-term beyond managerial
control (e.g. geographical and environmental constraints or political restrictions). Consistent
findings are revealed on the impact of increasing commercialization and outsourcing both
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
56
affecting the airport in a positive direction (Adler et al. 2010; Oum et al. 2006; Tovar and
Martín-Cejas 2009). Other scopes of analysis appeared to be inconclusive as we illustrate in
the following. This section compares the findings on productivity and efficiency changes and
the effects of ownership and size, all appearing to be frequently considered in efficiency
analyses.
2.4.1 Productivity and efficiency changes over time
Productivity and efficiency increases may be explained with an efficient use of the airport
infrastructure, innovations or changes in political decisions which aim to encourage
performance increases such as privatization and incentive regulation. The majority of studies
find positive changes mostly explained with technological improvements due to airport
investment programmes.
Assaf (2010b) utilizing SFA discovers cost efficiency increases in Australia since their
complete privatization in 2002 which he explains with the introduction of a light-handed price
monitoring that encourages investments and innovations at airports. Abbott and Wu (2002)
reveal technical progress prior privatization (1990-2000) due to advanced computer and air
traffic systems. The results from the Malmquist-DEA model point out that capacity expansion
is unneeded in the near future. Inconsistent outcomes were received for the British airport
market. Whereas Yokomi (2005) reveals an improvement post BAA privatization (1975-
2001), Barros and Weber (2009) find an average efficiency decrease between 2000 and 2004;
both utilizing Malmquist-DEA. However, given their period under review the declining trend
is not surprising where staff and other operating costs may have increased disproportionately
high to passengers, cargo and movements in the aftermath of the terror attacks in New York.
Furthermore, non-aeronautical activities have been ignored by Barros and Weber. Positive
changes are assessed for Chinese airports by Chi-Lok and Zhang (2008) and Fung et al.
(2008) conducting similar estimations between 1995 and 2004 with DEA. Fung et al. report
an annual productivity growth of 3% which is mainly explained with technical progress.
Gillen and Lall (2001) utilizing Malmquist-DEA find high productivity growth for terminal
side operations at US airports which however do not necessarily imply high growth rates for
the airside production and vice versa. On the terminal side Gillen and Lall identify hub and
gateway hubs as innovative airports by introducing fully automated baggage handling
systems. Technological progress is further revealed in the SFA estimation by Tovar and
Martín-Cejas (2009) for Spanish airports between 1993 and 1999. They explain
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
57
improvements as an outcome of the massive investment programme by the airport operator
AENA. Murillo-Melchor (1999) in turn concludes negative changes in Spain from 1992 to
1994 utilizing Malmquist-DEA. However positive technical changes coincide with efficiency
decrease in the first year comparison and vice versa in the year after.
2.4.2 Empirical effects of ownership
With the intention to reduce government involvement, to minimize costs and to maximize
productivity, a wave of airport privatizations began in the late Eighties in UK. Most European
countries followed to partially privatize their airports in the mid Nineties (Gillen and
Niemeier 2008). Reviewing the theoretical literature on privatization, its effects seem to be
somewhat controversial. Sappington and Stiglitz (1987) and Shapiro and Willig (1990)
support public ownership due to lower transaction costs and less asymmetric information.
Opponents of this point-of-view sought evidence to demonstrate that state intervention leads
to inefficiency as discussed by Shleifer and Vishny (1994) arguing with incomplete contracts.
Empirical studies attempting to assess the effects of ownership on the efficiency of
airports are so far rather inconclusive. Two different opportunities occur to consider
ownership as efficiency driver. The first study in this field is an analysis of privatization
effects of BAA airports. Parker (1999) estimates the technical efficiency prior and after
privatization (1979-1996) with basic DEA. He finds no evidence that privatization has
improved the airport’s technical efficiency and concludes that the golden share which is kept
by the government does not induce enough capital market pressures. Further he argues that
BAA is still subject to economic regulation and it may be argued whether incentives to
operate more efficiently can be distorted by government regulation. In contrast, Yokomi
(2005) reviews the technical and efficiency change of 6 BAA airports from 1975 to 2001
utilizing Malmquist-DEA. Different to Parker he finds that BAA airports have improved after
their privatization exhibiting positive changes in technical efficiency and technology. In
particular on the non-aeronautical side, the growth after privatization is substantial; this
activity is not considered in Parker’s analysis. However, before and after comparisons are
often problematic as privatizations are often accompanied with changes in the regulation or
restructuring processes such as outsourcing.
Other studies assess the effects of ownership by comparing the efficiency of public and
private airports. Again the results do not reach a clear conclusion. Lin and Hong (2006), Oum
et al. (2003) and Vasigh and Gorjidooz (2006) measure the effects of ownership on a
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
58
worldwide set and reveal no significant relationship for financial and operational efficiencies.
Oum et al. (2003) argue that the extent of managerial autonomy dominates the effect of
ownership. Furthermore, Vasigh and Haririan (2003) argue that privatized airports intend to
maximize their revenues whereas public airports aim to optimize traffic. Barros and Marques
(2009) find in their SFA estimation that private airports operate more cost efficiently than its
partial private counterparts. Furthermore, Oum et al. (2006) and Oum et al. (2008) are in
favour of privatization conducting index-number VFP and SFA respectively. In contrast to
previous studies they separate airports owned by one public shareholder from airports with
multilevel government involvement. Referring to Charkham (1995) they argue that different
ownership and governance structures can affect the quality of managerial performance. Oum
et al. (2006) reach the conclusion that public corporation are not statistically different from
major private airports. However, airports that are major publicly owned or have multiple
government involvement seem to operate significantly less efficient from the other ownership
forms. Oum et al. (2008) conclude that airports with major private shareholders are more
efficient than public airports or airports with major public influence. The results by Vogel
(2006) on a European set of airports reveal that privatized airports operate more cost efficient
and receive higher returns on total assets and revenues. Public airports in turn enjoy the
advantage of higher gearing and financial leverage.
According to Vickers and Yarrow (1991), privatization can not be seen as a universal
solution and should not be separated from the economics of competition and regulation which
are all determinants of corporate incentives. Airport benchmarking studies on the effects of
regulation and local competition are rare to-date. Barros and Marques (2008) utilizing SFA
reveal that regulatory procedures contribute to cost savings worldwide. Oum et al. (2004)
study alternative forms of regulation to assess the TFP including differences between single
till and dual till approaches and are in favour of a dual till price-cap regulation. Chi-Lok and
Zhang (2009) consider the impact of local competition in China which is likely to contribute
to efficiency improvements. In order to search for the most efficient ownership and regulation
form given the level of local and hub competition Liebert and Adler (2010) combine all
factors in an Australian-European semi-parametric two-stage research. The study concludes
that under monopolistic conditions, airports of any ownership form should be subject to
economic regulation. However, regulation can be replaced by effective competition in order
to ensure cost efficiency. Furthermore, public and major or fully private airports appear to
operate equally cost efficient. Hence, no clear answer reveals on the ownership form in
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
59
competitive situations, thereby indirectly supporting the inconsistent impact of ownership in
studies assessing ownership individually.
2.4.3 Scale effects
Benchmarking airports of different sizes can raise the question of how to eliminate the
effects of size for a multi- product ‘airport’ firm. According to Graham (2004) after the
airport reaches the size of about 3 to 5 million passengers, economies of scale effects flatten
out, so that for benchmarking of medium and large sized airports the size does not matter.
However, various benchmarking studies lead to different conclusions on the scale effects at
airports14.
The British market has been assessed numerous times. Doganis and Thompson (1973)
include UK airports in a regression and find decreasing average costs up to three million
WLU. Tolofari et al. (1990) estimate a long-run translog cost function of seven UK airports
owned by the BAA and find that scale economies exhaust at a level of 20.3 WLUs. However,
this result needs to be treated carefully as London-Heathrow is the only large airport in their
sample. Main et al. (2003) utilizing OLS conclude sharp decreasing costs up to 4 million
passengers and 5 million WLU and weak decreasing costs up to 64 million and 80 million
passengers and WLU respectively. On a worldwide sample they reveal that economies of
scale exhaust at a level of 90 million WLU. Jeong (2005) finds economies of scales at US
airports to exhaust at three million passengers. Different to all other studies under review,
capital is not considered due to lack of data. However non-aeronautical activities were
included thereby capturing a more complete picture of an airport. Keeler (1970) further
assesses US airports with OLS and concluded that scale is not the main source of inefficiency.
Furthermore, Doganis et al. (1995) conduct regression analysis on European airports and find
decreasing average costs up to five million WLU.
Martín and Voltes-Dorta (2007) applied SFA on a worldwide sample and conclude that
even Atlanta and Chicago as the two largest airports in the world operate under increasing
returns-to-scale. Pels et al. (2003) research the European market and reveal in their SFA
estimation for an average airport decreasing returns-to-scale from 12.5 million passengers for
the airside (i.e. movements) but increasing returns on the terminal side (i.e. passengers).
Similar conclusions are reached by de la Cruz (1999) utilizing DEA on Spanish airports in an
14 The studies under review in this section assessed both economies of scale and returns to scale.
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
60
aggregated model considering aeronautical and non-aeronautical activities. He finds constant
returns-to-scale between 3.5 and 12.5 million passengers and decreasing returns-to-scale from
then. In contrast, Martín et al. (2009), utilizing SFA find that all Spanish airports operate at
increasing returns-to-scale. However, different to de la Cruz, they restrict the airport model to
airside activities.
In summary, reviewing the rather inconsistent empirical findings from previous studies, it
becomes clear that different underlying assumptions from the methodology as well as
different inputs and outputs may lead to mixed results. Public and private airports may pursue
different targets and the object of research may influence the results of ownership impacts.
Furthermore, ownership should not be separated from the impact of economic regulation or
competition. In addition, commercial activities may contribute to efficiency improvements
and should implicitly be included. In order to assess the level of economies of scale and
returns-to-scale, regression analyses only permit single outputs unless combining passengers
and cargo to work load units (WLU). Hence the majority of studies failed to capture a
complete picture of an airport. In addition, non-aeronautical activities are mostly ignored
thereby inducing an imbalance on the input and output side where staff is not adjusted
accordingly.
2.5 Conclusion
Since the late Nineties a number of empirical research studies emerged on the productivity
and efficiency analysis of airports with overall quantitative methods such as DEA, SFA or
PIN. Studies proved to mostly assess economic rather than managerial issues. With the
publication of a meanwhile substantial number of benchmarking studies in the academic
literature, this paper aimed to provide an overview of the methodology applied, the variables
selected and the findings on productivity and efficiency changes, scale and ownership effects
in order to learn from previous research.
Frontier approaches have been preferred over productivity measures with index numbers
in order to account for efficiency differences across airports. Both efficiency techniques, DEA
and SFA, have substantially improved over the years and its usefulness for the airport
industry has steadily increased. This regards in particular the consideration of the airports’
heterogeneous character. DEA as a non-parametric technique proved to be mostly utilized
requiring fewer assumptions than SFA however its application has increased since recently.
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
61
Comparing the findings of different studies proved to be difficult. The application of
different methods and data are likely to affect the results on the relative efficiency.
Nevertheless, efficiency improvements from increasing commercialization, restructuring and
technological progress appeared to be fairly consistent. Ownership and scale effects proved to
be inconclusive showing that further research is required.
The collection of appropriate data proved to be a serious issue in airport benchmarking.
Especially early studies ignored non-aeronautical activities in order to assess the airside’s
efficiency thereby obtaining biased efficiency estimates where the input side was not adjusted
accordingly. Another aspect is the measurement of capital which appears to be crucial.
Although the airport industry is typical for being capital intensive with lumpy investments
various studies failed to include capital appropriately or ignored this figure due to lack of
data. Undesirable outputs such as delay or noise proved to be important in efficiency analysis
to prevent an overestimation of efficiency results.
In order to conduct managerial benchmarking we suggest to assess airports with similar
input/output combinations in order to ensure comparability. Management strategies as the
degree of vertical integration or airport characteristics like the traffic structure and size need
to be carefully considered. Furthermore, airports operating under different scheduling
practices may have different declared capacities. Hence a comparison of slot-coordinated
airports with uncoordinated counterparts needs to be treated with caution. Due to lack of
detailed information, all studies considered the overall airport system rather than focussing on
partial processes which airport managers often find more informative. In order to improve the
modelling and its application for political and managerial purposes we argue that airport
managers should contribute with their industry knowledge. In summary, the area of
benchmarking appears to be a valuable instrument for airport managers, governments and
regulators but future research is needed to improve its utilization.
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
62
2.A Appendix15
Tab. 2: Studies using non-parametric approaches
15 For abbreviations see legend at end of the tables.
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
63
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
64
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
65
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
66
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
67
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
68
Tab. 3: Studies using parametric approaches
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
69
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
70
A Survey of Empirical Research on the Productivity and Efficiency Measurement of Airports
71
Tab. 4: Studies using price-based index approaches
72
3 JOINT IMPACT OF COMPETITION,
OWNERSHIP FORM AND ECONOMIC
REGULATION ON AIRPORT
PERFORMANCE16
The combined impact of ownership form, economic regulation and local and gateway
competition on airport performance is analyzed using data envelopment analysis in a first
stage efficiency measurement and regression analysis in a second stage environmental study.
The results of an analysis of European and Australian airports over a ten-year period prove to
be stable across different robust cluster regression models and show that airports not facing
regional or hub competition should be regulated to increase cost efficiency. However, in a
competitive setting, economic regulation inhibits airports of any ownership form from
operating efficiently. On the other hand, unregulated major and fully private airports act as
profit-maximizers even within a competitive setting by charging higher aeronautical revenues
than those that are regulated.
16 The author is grateful to Dr. Nicole Adler for helpful discussions and support. Paper is not yet submitted for publication.
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
73
3.1 Introduction
Historically, airports were mostly deemed state-owned entities with the objective to
provide and operate infrastructure for airlines. Being often viewed as natural monopolies with
large economies of scale airports were subject to economic regulation in order to prevent
abuse of market power. However, the nature of the airport industry has changed over the last
two decades. Moving away from viewing the airport as a public utility, airports have begun to
operate as modern enterprises pursuing commercial objectives. A number of privatization
processes have been actively promoted by governments with the proclaimed intention of
reducing government involvement and increasing airport productivity and innovation.
However, given the assumed profit-maximizing behaviour of private companies working in a
natural monopolistic environment, the majority of privatized airports in Europe remain
subject to economic regulation (Gillen 2010).
Whilst some studies have analyzed the impact of ownership form, regulatory regime and
level of competition from nearby airports on efficiency and airport pricing, none have
examined their joint impact. In other words, the literature has yet to discuss whether the
deregulation of the airline industry and changes in airport ownership and management has
affected the competitive situation, airport pricing and efficiency to the extent that the benefits
of economic regulation are potentially unnecessary. For example, deregulation has led to
increased competition between gateway hubs (e.g. Frankfurt and Amsterdam) and former
military airports have opened to serve low cost carriers within the catchment area of existing
airports (e.g. Hahn in Germany), in turn substantially changing the downstream airline market
and potentially impacting the airport market too. Furthermore, as a result of increasing
commercialization, many airports have augmented their revenues from non-aeronautical
sources in order to cross-subsidize aviation charges and attract additional airlines and
passengers to their airport (Zhang and Zhang 2010). The aim of this research is therefore to
analyze the impact of the structural changes in the aviation markets on airport efficiency and
pricing in order to further our understanding of the most appropriate ownership form and
regulatory regime given the level of regional and hub competition at a specific airport.
Performance measurement may serve multiple purposes, as outlined by Oum et al. (1992).
It may assess the productivity or efficiency of units within or across companies or industries
and identify best-practice standards. Furthermore, the availability of panel data permits the
measurement of changing levels of productivity over time. Although the instrument of
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
74
performance measurement was applied in other transport sectors and regulated utilities in the
nineteen seventies, it only became of primary importance in the airport industry twenty years
later. Graham (2005) argues that the increasing interest in airport benchmarking is a result of
the changes in ownership that began in 1987 with the privatization of BAA and the
liberalization, commercialization and globalization trends which have influenced airport
business growth, complexity and competitiveness.
Three well-documented quantitative methods have been applied to analyze the
productivity and efficiency of government and private enterprises. A non-parametric, index
number approach has been used to measure total factor productivity (Caves, Christensen and
Diewert 1982a), however this approach requires input and output prices and quantities which
are not always available. Parametric stochastic frontier analysis (SFA) assesses efficiency
utilizing regression analysis and disentangles unobservable random error from technical
inefficiency (Aigner, Lovell and Schmidt 1977; Meeusen and van den Broeck 1977) based on
assumptions as to the distributional forms of the efficiency function and error term. Non-
parametric data envelopment analysis (DEA), based on linear programming, categorizes data
into efficient and inefficient groups hence produces weaker results than those of SFA, but
does not require assumptions with respect to a functional form therefore is chosen for the
purposes of this study. Airport studies of efficiency utilizing all three approaches are reviewed
in Liebert (2010).
Various environmental variables that, at least in the short-term, are beyond managerial
control may affect the DEA efficiency estimates. Previous research argues that airport
characteristics such as hub status or traffic structure, outsourcing policies, regulatory
procedures and ownership structure all may contribute to airport efficiency (Gillen and Lall
1997; Oum et al. 2006). Assessing the importance of the environment on the efficiency
estimates may be undertaken utilizing either non-parametric Mann-Whitney and Kruskal-
Wallis tests or parametric regression. Banker and Natarajan (2008) demonstrate that two-stage
procedures in which DEA is applied in the first stage and regression analysis in the second
stage provide consistent estimators and outperform parametric one- or two-stage applications.
Published airport studies apply simple ordinary least squares (Chi-Lok and Zhang 2009),
Tobit regression (e.g. Gillen and Lall 1997; Abbott and Wu 2002) and truncated regression
(Barros 2008a) for this purpose. A recent debate in the literature discusses the most
appropriate second stage regression model to be applied when investigating DEA efficiency
estimates. Simar and Wilson (2007) argue that truncated regression, combined with
bootstrapping as a re-sampling technique, best overcomes the unknown serial correlation
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
75
complicating the two-stage analysis. Banker and Natarajan (2008) conclude that simple
ordinary least squares, maximum likelihood estimation or Tobit regression dominate other
alternatives. Combining the arguments of Simar and Wilson (2007) and Banker and Natarajan
(2008), we apply robust cluster regression based on ordinary least squares in order to account
for the correlation across observations. Furthermore, in order to ensure the robustness of the
results, we also apply robust cluster truncated and censored regressions.
The second stage analysis of this research considers the impact of ownership form,
economic regulation and levels of local and hub competition amongst other factors. Several
empirical studies to date have assessed the effects of privatization on airport efficiency.
Parker (1999), utilizing data envelopment analysis, argues that the privatization of BAA had
no effect on subsequent efficiency. Oum et al. (2006), applying variable factor productivity
argues that private majority ownership and pure government ownership are equally efficient
and both are strictly preferable to government majority ownership or multi-tiered government
ownership. Oum et al. (2004) analyze different regulatory regimes and conclude that total
factor productivity is maximized under dual till price-caps rather than single till price-caps or
rate of return regulation. Chi-Lok and Zhang (2009), utilizing data envelopment analysis on a
Chinese airport dataset, reach the conclusion that the intensity of airport competition at the
level of the local catchment area likely encourages greater productivity.
Whereas previous studies analyze the effects of ownership, regulation and competition
individually, we support the argument of Button and Weyman-Jones (1992) that all three
factors should be accounted for simultaneously as their combined impact is likely to affect
airport efficiency. Such an analysis may contribute to the search for the more desirable
combinations and may indicate whether effective competition from nearby airports or
gateway competition replaces the need for economic regulation. In addition, we examine the
combined impact on aeronautical revenues generated from passengers and movements in
order to understand the pricing behaviour of airports under different institutional settings. Bel
and Fageda (2010), who assessed 100 large airports in Europe, argue that local competition
decreases the abuse of market power whereas private unregulated airports tend to charge
higher prices than public and regulated airports.
The dataset in this research consists of European and Australian airports in order to
include a sufficiently heterogeneous sample with respect to the ownership structure,
regulatory mechanism and competitive environment. The empirical results reveal that under
rather monopolistic conditions, airports should be regulated to encourage cost efficiency and
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
76
dual till price-cap regulation appears to be the most effective regulatory form. Furthermore,
airports of any ownership form under monopolistic conditions are likely to abuse market
power and set higher aeronautical charges. However, gateway or regional competition
replaces the need for any form of economic regulation, thereby supporting the argument of
Vickers and Yarrow (1991) that competition rather than privatization is the key driver of
efficiency. Nevertheless, unregulated major and fully private airports within a competitive
setting remain profit-maximizers and in this regard may still require ex-ante regulation.
The paper is organized as follows: the theoretical and empirical literature discussing
ownership form, economic regulation and competition is presented in Section 3.2; Section 3.3
introduces the methodology and model specifications, Section 3.4 discusses the dataset for the
two stages of analysis, the results are presented in Section 3.5 and conclusions and directions
for future research are suggested in Section 3.6.
3.2 Literature on competition, regulation and ownership
The neoclassical theory of the firm states that competition leads to increased productive
and allocative efficiency as a result of lower prices and higher outputs. In the case of
indivisibilities, as typically occurs in the provision of infrastructure based services and
utilities, one large firm might be able to produce at lower costs leading to monopolistic
conditions. In this case, in order to encourage efficiency and avoid abuse of market power, the
natural monopolist should be subject to economic regulation (Lipczynski et al. 2009).
In Europe17, airport charges have traditionally been regulated according to a rate of return
or cost-plus principle (Reinhold et al. 2010). Such regulation permits airports to generate
sufficient revenue to cover total expenditures, including the depreciation of capital and an
expected rate of return on capital. However, according to Averch and Johnson (1962), this
form of regulation may lead to overcapitalization which does not engender productive
efficiency. To solve the problem of overinvestment, Littlechild (1983) proposes an incentive
based price-cap regulation. Price-caps are generally set over a regulatory period of five years
according to the RPI-X formula where RPI represents the retail price index and X is the
efficiency improvement that the regulators consider reasonable within the timeframe. If the
airport management achieves greater cost reductions over the five year period, the gains are
enjoyed by the company. In the case of airports, the single till principle is applied in the UK,
17 Gillen and Niemeier (2008) provide a comprehensive overview of the current economic regulation at European airports.
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
77
in which case both aeronautical and non-aeronautical revenues are constrained. Over the
years, price-cap regulation has been emulated by other European authorities however, unlike
the UK model, a dual till approach is applied whereby aeronautical revenues alone are subject
to regulation (Gillen and Niemeier 2008). Compared to traditional rate of return regulation, a
price-cap creates incentives for cost savings hence encourages efficiency, however it equally
may lead to underinvestment on the part of firms with heavy infrastructure sunk costs.
Consequently, it may be necessary to regulate in order to ensure a reasonable level of quality
with respect to the products or services offered. Another approach to stimulate efficiency is
yardstick competition originally proposed by Shleifer (1985). This form of regulation implies
virtual competition amongst regulated firms by comparing their cost levels and determining
the permitted price based on an average level. Common approaches utilized to assess
appropriate cost levels for regulated firms include frontier techniques such as DEA and SFA.
In addition, the cost function should be corrected to take into account external heterogeneities.
Factors, such as geographical constraints, may affect airport costs but are considered to be
beyond the control of the airport management. Whereas yardstick competition evolved to a
standardized approach in the British water and railway industries, it has rarely been applied to
airports so far. To the best of our knowledge, the Dublin Airport Authority (DAA) is the only
European example that attempted to implement yardstick competition in 2001. However, it
was highly criticized by airport management for identifying inappropriate peer airports
(Reinhold et al. 2010) and was discontinued. The British CAA argues that the heterogeneous
character of airports and the challenge to obtain appropriate data contribute to their reluctance
to apply this type of economic regulation (CAA 2000).
In the theoretical literature, the debate as to the necessity for and type of airport regulation
seems to be rather controversial. Gillen and Niemeier (2008) argue in favour of price-cap
regulation but also that commercial and ground handling activities may be disciplined to some
extent by potential competition, hence the dual till price-cap approach is preferable. Czerny
(2006) argues that market power exists in both the aeronautical and commercial spheres of
activity. For non-congested airports, he suggests that the single till outperforms dual till price-
cap regulation in maximizing social welfare. For large, congested airports, Beesley (1999)
argues that the single till is inappropriate because increasing concession profits would lead to
lower airport charges over time. In addition, Starkie (2002) finds no evidence of economies of
scale for airports with large throughput and argues that demand complementarities across
aeronautical and terminal activities will prevent airports from abusing market power,
obviating the need for any regulation. In particular, airports generating additional revenues
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
78
from non-aeronautical activities are likely to lower their charges and cross-subsidize using
commercial revenues in order to attract both passengers and airlines (Zhang and Zhang 2010).
To the best of our knowledge, the impact of regulation on efficiency and airport pricing
has only been empirically assessed in few papers. Barros and Marques (2008) incorporate a
dummy variable defining cost-plus or price-cap regulation in order to assess a worldwide set
of airports from 2003 to 2004, estimating a heterogeneous cost frontier utilizing stochastic
frontier analysis. They conclude that regulatory procedures contribute to cost savings. The
study by Oum et al. (2004) collected data on worldwide airports for the years 1999 and 2000
and applying gross endogenous-weight total factor productivity. They carefully study various
forms of regulation including differences between single till and dual till concepts. The results
indicate that airports under dual till price-cap regulation tend to have higher levels of gross
total factor productivity than those with a single till price-cap or those that operate under the
single till rate of return regulation. Furthermore, dual till approaches together with rate of
return regulation appear to provide incentives to improve efficiency but are very complex to
estimate. Bel and Fageda (2010) examine the impact of privatization, regulation and regional
and intermodal competition on airport charges at European airports in 2007. Utilizing
regression analysis, they reveal that competition with nearby airports and other transport
modes is likely to decrease the potential to abuse market power. Furthermore, private
unregulated airports charge higher prices than public and regulated airports thereby
supporting the analytical findings of Oum et al. (2004). Van Dender (2007) assessed the US
market between 1998 and 2002 utilizing an econometric approach and similarly concluded
that airports under regional competition charge lower fees. He also argued that slot-
constrained airports are likely to charge higher aeronautical fees which are explained by the
airport management’s ability to capture scarcity rents.
With the stated aim of reducing government involvement, minimizing costs and
maximizing productivity, a wave of airport privatizations began in the late Eighties in the UK.
Due to successful initial public offerings and increasing share prices, many European
countries began to partially privatize their airports in the mid Nineties (Gillen and Niemeier
2008). Reviewing the theoretical literature on privatization, its effects seem to be somewhat
controversial. Sappington and Stiglitz (1987) argue that the transaction costs of government
intervention are lower under public ownership. In a similar vein, Shapiro and Willig (1990)
argue that the government is better informed and more capable of regulating state-owned
firms. Opponents of this point-of-view sought evidence to demonstrate that state intervention
leads to inefficiency. Shleifer and Vishny (1994), for example, argue that the relationship
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
79
between politicians and managers is governed by incomplete contracts leading to inefficient
incentives. In addition, the emergence of partially privatized models complicates the debate as
to the effects of ownership on productivity. Boardman and Vining (1989) review the effects of
mixed ownership structures based on theoretical arguments and empirical studies. They
conclude that large, industrial, partly privatized and state-owned companies perform in a less
productive and profitable manner than their fully private counterparts, which may be caused
by the public and private shareholders’ differing objectives. Considering the issue to be more
complex, Vickers and Yarrow (1991) argue that privatization is not a universal solution to the
agency problem in the public sector and should not be separated from the economics of
competition and regulation which are all determinants of corporate incentives.
Empirical studies that attempt to assess the effects of ownership on the efficiency of
airports are so far rather inconclusive. Parker (1999) utilizes DEA to estimate the technical
efficiency of the BAA airports between 1979 and 1996 covering the period pre and post
privatization. No evidence is found that complete privatization leads to improved technical
efficiency and he concludes that the UK government’s golden share limits the impact of
capital market pressures. Furthermore, he argues that BAA remained subject to economic
regulation hence incentives to operate more efficiently are distorted as a result of government
intervention. In contrast, Yokomi (2005) reviews the technical and efficiency change of six
BAA airports from 1975 to 2001 utilizing Malmquist DEA. As opposed to Parker, Yokomi
finds that the BAA airports exhibit positive changes in efficiency and technology as a result of
the privatization. It should be noted that commercial growth after privatization was
substantial; however this activity is not considered in Parker’s analysis.
The effects of different ownership forms on efficiency were also analyzed but again the
results have not reached clear conclusions. Barros and Dieke (2007) analyze 31 Italian
airports from 2001 to 2003 using DEA in the first stage and Mann-Whitney hypothesis testing
in the second stage, to reveal that private airports operate more efficiently than their partially
private counterparts. However, Lin and Hong (2006) find no connection between ownership
form and efficiency after analyzing a dataset of worldwide airports for the years 2001 and
2002 utilizing DEA and hypothesis testing. Oum et al. (2006, 2008) distinguish between
public airports owned by public corporations and those owned by more than one public
shareholder (multilevel). Referring to Charkham (1995), they argue that different ownership
and governance structures affect the quality of managerial performance. Oum et al. (2006)
assess a sample of 100 airports worldwide covering the years 2001 to 2003 utilizing variable
factor productivity. They reach the conclusion that the productivity of a public corporation is
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
80
not statistically different from that of a major private airport. However, airports with major
public shares or multiple government involvement operate significantly less efficiently than
other ownership forms. Oum et al. (2008) estimate a heterogeneous translog cost function
with stochastic frontier analysis on a similar set of airports as that of Oum et al. (2006),
measuring cost efficiency between the years 2001 and 2004. The authors conclude that
airports with major private shareholders are more efficient than public airports, particularly
those with a major public ownership structure.
The traditional perspective of airports behaving as monopolists has changed as a result of
the deregulation of the downstream aviation industry according to Tretheway and Kincaid
(2010). Today, competition for airport services covers a multiplicity of markets including (1)
a shared local catchment area, (2) connecting traffic through regional hubs and international
gateways, (3) cargo traffic, (4) destination competition, (5) non-aeronautical services, (6)
competing ground handling companies and off-site car parks and (7) alternative modes of
transport such as high speed rail in the medium distance markets. Amongst the empirical
literature, only Chi-Lok and Zhang (2009) examine the effects of regional competition
utilizing a Chinese airport dataset for the years 1995 to 2006. After applying DEA in the first
stage and ordinary least squares in the second-stage, they conclude that airports operating in a
locally competitive environment tend towards efficiency. However, the outcome of a Tobit
regression found competition intensity to be insignificant.
In summary, whereas research to date has analyzed the individual effects of ownership,
regulation and competition on efficiency, the joint impacts may be of great interest as argued
in Button and Weyman-Jones (1992, p.440) that “[t]he degree of competitiveness in a firm's
market, the extent to which it is incorporated as part of a public-sector bureaucracy, and the
nature of the regulatory regime under which a firm operates are all primary sources of
possible X-inefficiency”. Consequently, our intention is to assess the combined impact of
ownership structure and economic regulation (or lack thereof) given relevant levels of local
and hub competition. We argue that the choice of ownership form and the regulatory
procedures instituted are clearly within the bounds of public policy initiatives over the
medium term, whereas the competitive environment remains more costly to change.
3.3 Methodology and model specification
The following section presents the weighted additive DEA model (Lovell and Pastor
1995) which we apply in the first-stage analysis in order to account for both the desired equi-
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
81
proportional reductions in all inputs and any remaining slacks. We then discuss the second-
stage regression specifications in which the DEA efficiency estimates are regressed against
environmental factors.
3.3.1 Data envelopment analysis
DEA is a non-parametric method of frontier estimation that measures the relative
efficiency of decision-making units (DMUs) utilizing multiple inputs and outputs. DEA
accounts for multiple objectives simultaneously without attaching ex-ante weights to each
indicator and compares each DMU to the efficient set of observations, with similar input and
output ratios, and assumes neither a specific functional form for the production function nor
the inefficiency distribution. DEA was first published in Charnes et al. (1978) under the
assumption of constant returns-to-scale and was extended by Banker et al. (1984) to include
variable returns-to-scale. This non-parametric approach solves a linear programming
formulation per DMU and the weights assigned to each linear aggregation are the results of
the corresponding linear program. The weights are chosen in order to show the specific DMU
in as positive a light as possible, under the restriction that no other DMU, analyzed under the
same weights, is more than 100% efficient. Consequently, a Pareto frontier is attained,
marked by specific DMUs on the boundary envelope of input-output variable space. Charnes
et al. (1981, p.668) described DEA as a “mathematical programming model applied to
observational data [which] provides a new way of obtaining empirical estimates of extremal
relations – such as the production functions and/or efficient production possibility surfaces
that are a cornerstone of modern economics”.
The weighted additive model (Charnes et al. 1985; Lovell and Pastor 1995), chosen for its
units and translation invariance properties, reflects all inefficiencies identified in the inputs.
The input oriented model is chosen because we assume that airport managers control
operational costs and to a lesser extent airport capacities, but have less control over traffic
volume. By comparing n units with q outputs denoted by Y and r inputs denoted by X, the
efficiency measure for airport a is expressed as in model (3.01).
0,, 1
-- ..
,
≥=
−=
=−
σλλ
σλ
λσ
s e
XXYsYts
swMax
a
a
t
s
(3.01)
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
82
In (3.01), λ represents a vector of DMU weights chosen by the linear program, wt a
transposed vector of the reciprocals of the sample standard deviations, e a vector of ones, σ
and s vectors of input and output slacks respectively and Xa and Ya the input and output
column vectors for DMUa respectively. Hence DMUa, the airport under investigation, is
efficient if and only if all input slacks equal zero. Variable returns-to-scale is assumed (eλ=1)
because the sample dataset consists of airports of substantially different sizes, ranging from
0.5 million passengers at Southampton to more than 50 million per annum at London-
Heathrow and Frankfurt. It should be noted that the objective function value of equation
(3.01) lies between zero and infinity with a DMU deemed efficient when the sum of slacks
equal zero. In order to interpret the coefficients obtained from the second-stage regression as
percentages, the efficiency scores were normalized to a range from zero to one, where one
depicts a relatively efficient airport.
3.3.2 Second-stage regression
The inefficiency scores estimated in the first stage may be explained by factors beyond
managerial control. In order to conduct hypothesis testing, regression analyses is often applied
in a second stage in which the DEA efficiency estimate is regressed against a set of potential
environmental variables. Banker and Natarajan (2008) and Simar and Wilson (2007)
independently review appropriate forms to conduct second-stage regressions of DEA
estimates which led them to different conclusions. Based on Monte Carlo simulations, Banker
and Natarajan (2008) argue that ordinary least squares, Tobit (censored) regression and
maximum likelihood estimation in the second-stage outperform one-stage and two-stage
parametric methods. Simar and Wilson (2007) argue that the majority of empirical two-stage
studies do not properly define the data generating process. The efficiency estimates are likely
to be serially correlated via the efficiency frontier hence the error term, εi, will also be serially
correlated thereby violating the common assumption that the errors are identically and
independently distributed. They also state that any bias in the efficiency estimate is ignored
and will be automatically included in the error term. Consequently, Simar and Wilson
advocate truncated regression in the second stage, which removes the efficient units from the
sample. The problem with this approach is that we would then ignore all airports deemed to
be lying on the efficient frontier, yet we are searching for the most appropriate form of
ownership and regulation given the competitive environment. Drawing from both papers, we
apply ordinary least squares thereby following the Banker and Natarajan (2008) approach. To
handle the issues identified in Simar and Wilson (2007), we utilize a robust cluster approach.
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
83
Accounting for heteroscedastic robust standard errors as proposed by White (1980), we
construct unbiased t-tests and confidence intervals hence attempt to solve the limitation that
the error term is not identically distributed. Furthermore, the sample will be clustered at the
airport level in order to overcome the limitation that the error term is not independently
distributed.
For purposes of sensitivity analysis, we present the results of three models; the truncated
regression proposed by Simar and Wilson (2007) and ordinary least squares (OLS) and
(censored) Tobit regression as proposed by Banker and Natarajan (2008). The Tobit
regression is censored at one whereas the truncated regression will remove all observations
whose dependent variables equals one, as presented in Table 5.
Tab. 5: Regression analysis
Ordinary Least Squares Regression Tobit Regression Truncated Regression
εβ += Xy y presents a vector of DEA efficiency estimates; X a matrix of environmental variables, β the parameters to be estimated and ε the error term.
otherwise y
1y if 1y
ˆ
⎩⎨⎧ ≥
=
+= εβXy
ŷ is the true but unobservable efficiency.
εβ += Xy ŷ is truncated if y= ŷ for all ŷ ≥1
In addition to examining the joint impact of ownership form, economic regulation and
regional and gateway competition on airport efficiency, we assess their effects on pricing
behaviour utilizing robust cluster OLS. Unfortunately, we could not obtain sufficient
disaggregated data with regard to departing passenger or landing charges for our sample.
Consequently, we approximate the price via the total revenues obtained from aeronautical
activities18. In two separate regressions, revenues per passenger and revenues per movement19
are regressed over the environmental variables amongst other factors. Although we are aware
of different pricing strategies across airports, for simplicity we assume that passengers and
landing charges represent an equal share of aeronautical revenues.
3.4 Dataset
In this section we present the airports to be analyzed, the variables collected for the
efficiency analysis and then the environmental variables included in the second stage
regression. Table 11 in Appendix 3.A lists the complete set of airports in the sample, which
18 Revenues from ground handling services are not considered. 19 For simplicity the role of cargo has been ignored.
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
84
include 48 European airports of which half are located in Germany and the United Kingdom.
To ensure a heterogeneous dataset with respect to the form of ownership, economic regulation
and level of local and hub competition, we further include three fully privatized Australian
airports, Melbourne, Perth and Sydney, which are neither ex-ante regulated20 nor exist in a
competitive environment due to the great distances across this continent. The pooled data
consists of an unbalanced set of 398 observations covering the time period between 1998 and
2007. The size of the airports under review varies considerably between an annual passenger
volume of half a million passengers at regional airports such as Southampton to more than 50
million passengers at international gateways such as London-Heathrow and Frankfurt.
3.4.1 Variables in the first-stage efficiency analysis
For the first-stage efficiency analysis, three inputs and four outputs are collected as
summarized in Table 6. The operating inputs consist of staff costs and other operating costs,
including materials and outsourcing. Despite being a smaller airport than London-Heathrow
in terms of air traffic movements, Frankfurt spends the most on staff costs because it is a
highly integrated airport that operates most airport services in-house or through wholly-
owned subsidiaries. Consequently, Heathrow spends the most in the other operating costs
category, reflecting the high levels of outsourcing undertaken.
Tab. 6: Variables in analysis (DEA)
Variable Description Average Standard Deviation Maximum Minimum Source
Staff costs Wages and salaries, other staff costs (2000=1 and US$=1)
63,654,765 120,554,070 1,080,756,267 3,655,825 Annual Reports
Other operating costs
Costs of materials, outsourcing and other (2000=1 and US$=1)
84,811,284 117,603,464 725,987,196 3,631,353 Annual Reports
Declared runway capacity
Number of movements per hour 46 20 110 15 IATA (2003)
Airport Coordinator
Passengers Annual passenger volume (only terminal passengers) 11,091,246 12,761,170 67,673,000 480,011 Annual Reports
Cargo Metric tons (trucking excluded) 172,922 366,881 2,190,461 0 Annual Reports
Air transport movements
Number of commercial movements 132,482 109,190 492,569 19,397 Annual Reports
Non-aeronautical revenues
Revenues from concessions own retail and restaurants, rents, utilities and ground handling activities (2000=1 and US$=1)
124,647,578 186,748,031 1,167,377,411 6,194,408 Annual Reports
20 According to the Trade Practices Act 1974, the Australian Competition & Consumer Commission (ACCC) is responsible for the monitoring of prices, costs and profits related to aeronautical and airport car parking services and facilities in Adelaide, Brisbane, Melbourne, Perth and Sydney. Hence the airports experience some form of ex-post regulation (Forsyth 2004).
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
85
It is also necessary to consider capital however this variable is extremely problematic
since it is often unreported. If the dataset covers more than one country, the monetary
measurement of physical capital creates difficulties due to different national accounting
standards and depreciation methods or periods across countries. For example, the airports of
the British Airports Authority depreciate their runways over 100 years whereas the airports
operated by the Aéroports de Paris apply a 10 to 20 year depreciation rule (Graham 2005).
Consequently, physical data such as the number of runways, gates or check-in-counters and
terminal size are often collected for cross-border studies as a proxy for capital (Gillen and Lall
1997; Pels et al. 2003). However a simple linear aggregation of these variables is problematic,
for example because the number of runways does not include information on the
configuration, weather impacts or environmental restrictions. In this study, we include
declared runway capacity as a proxy for capital, which is defined as the capacity constraint on
the number of departure and arrival movements per hour. Declared runway capacity is
negotiated twice a year in agreement with airport stakeholders and is primarily used to avoid
congestion at schedule facilitated airports and to allocate slots at coordinated airports.
Compared to the theoretical capacity, declared runway capacity is not only limited by
physical runway constraints but also by the air traffic control system, weather impacts and
noise and emissions restrictions (IATA 2010). Compared to pure physical information, the
capacity measure allows for greater variability since it accounts for bottlenecks that may be
solvable in the short to medium term. In the dataset, Amsterdam possesses the highest runway
capacity at 110 movements per hour, due in part to the geographical location near the coast
which requires additional runways and a special configuration to handle operations
consistently irrespective of weather conditions. The smallest airport with respect to runway
capacity is Ljubljana with a maximum hourly rate of fifteen movements. Consistent terminal
data proved very difficult to collect hence has been excluded in this study, however runway
capacity is highly correlated to terminal capacity therefore this omission should not greatly
impact the results.
On the output side, the annual traffic volume is represented by the number of passengers,
commercial air transport movements and tons of cargo (trucking was excluded). Freight
handling is of differing importance across the airports in the sample set since Dortmund and
London-City do not serve cargo operations whereas Leipzig and Cologne-Bonn are the
European hubs for DHL and UPS respectively. The fourth output variable captures revenues
from the non-aeronautical activities, including concessions, car parking and rent. In addition
to the traditional income sources, non-aeronautical revenues also include revenues from
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
86
labour-intensive ground handling activities. Ignoring this operation on the output side would
otherwise bias the results since the input data could not be adjusted to exclude this service. At
least theoretically we should be able to compare all three airport models, namely airports who
produce ground handling services in-house and have relatively higher employee costs and
requisite revenues, those who outsource which appear in the other costs category and their
respective revenues and the third case in which airports do not provide the service nor earn
revenue beyond perhaps a nominal fee from third party contractors. All financial data is
deflated to the year 2000 and adjusted by the purchasing power parity according to the United
States dollar in order to ensure comparability across countries. In addition, the data has been
normalized by the standard deviation to ensure that all inputs are considered equally within
the additive model.
3.4.2 Variables in the second-stage regression
Variables describing ownership structure, economic regulation and the level of hub and
local competition have been collected for this study in addition to specific airport
characteristics and information on managerial strategies. All factors are at least in the short-
term beyond managerial control yet may contribute to the inefficiency measurement process.
All data is expressed in the form of categorical variables. To further assess efficiency changes
over the review period and remove time-related effects, categorical variables on the financial
years21 have been included.
Airports frequently attempt to increase revenues from non-aeronautical sources that are
not directly related to aviation activities in order to cross-subsidize aviation charges in turn
attracting more airlines and passengers to their airport (Zhang and Zhang 2010).
Consequently, revenue source diversification that exploits demand complementarities across
aeronautical and non-aeronautical services may improve airport efficiency. Oum et al. (2006)
find a positive and highly significant relationship between the share of non-aeronautical
revenues and the level of efficiency. To compare our results with that of Oum et al. (2006),
we compute the percentage share of non-aeronautical revenue ignoring ground handling
activities. Airports are split between those that earn less than 50% of their revenues from non-
aeronautical activities and those that exceed this share. The threshold of 50% was chosen such
that a rich set of airports exist in the two categories and sensitivity analysis show no change in
the results when reducing the threshold to 40% or increasing it to 60%.
21 Note that financial data has been adjusted by different reporting periods.
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
87
Oum and Yu (2004) conclude that a higher share of intercontinental traffic leads to
efficiency decreases due to additional service requirements that generate higher costs
compared to domestic traffic. We define a threshold of 15% or more of intercontinental traffic
to separate the set between hubs such as Frankfurt, Amsterdam and London-Heathrow and
airports with predominantly domestic and European destinations.
Pathomsiri et al. (2008) consider the impact of delay on productivity by incorporating the
number of delayed flights and time delays in minutes as a negative output in their non-
parametric model analyzing a sample of US airports. Perhaps unsurprisingly, they conclude
that ignoring delays leads to an overestimation of efficiency at congested airports. For the
European market we were not able to collect airport related delays per movement for the first
stage analysis hence have collected a categorical variable based on the ranking of the most
delayed airports (departure and arrival) as reported by the European air traffic control22. Over
the review period, Amsterdam, the London airports Heathrow, Gatwick and Luton and
Manchester are consistently ranked as the European airports with the greatest levels of delay,
with Zurich listed up until 2005 (Eurocontrol 1999-2008). An airport appearing on the list of
the Top 50 delayed airports in Europe was categorized accordingly23. In addition, we consider
runway utilization in order to assess the effects of congestion which are expected to positively
impact efficiency. The variable was calculated based on annual air transport movements
divided by estimated annual declared capacity24. We define three categories including (1)
airports with less than 50% runway utilization indicating under-utilization, (2) airports
between 51% and 90% runway utilization and (3) airports achieving more than 90%25 runway
utilization indicating congestion.
According to Kamp et al. (2007), analyzing airports without considering the degree of
outsourcing is likely to bias the efficiency results particularly with regard to labour-intensive
ground handling services. In our dataset, airports located in Austria, Germany and Italy
traditionally operate ground handling activities in-house whereas in the UK and Switzerland
these operations are provided by the airlines themselves or via independent third party
providers. Munich, a major German hub, announced in 2009 that their ground handling
services department has suffered losses ever since this activity underwent liberalization in
22 However, the measure reported by Eurocontrol does not capture airport-related delays rather than delays caused by airlines, the air traffic control system and weather, which are beyond control of the airport manager.
23 Australian airports are included in the group of non-delayed airports for lack of further information. 24 The annual capacity has been estimated from the declared hourly capacity obtained from the airport coordinator. 25 Note that the theoretical runway capacity normally exceeds the declared runway capacity hence the runway utilization with
respect to the theoretical value is somewhat lower than 90%.
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
88
1996. Munich airport management argue that salaries are paid based on public tariffs which
are on average 20% higher than the private sector and strong labour unions in Germany make
it difficult to either adjust the compensation or to outsource this segment (Hutter 2009). It
should be noted, however, that it may be in the interests of the airport to cross-subsidize the
ground handling operations with alternative revenues in order to survive in the competitive
market (Barbot 2010). In order to test whether ground handling operations in-house affects
the efficiency estimates of airports despite considering all salient variables, we include a
categorical variable with regard to the provision of this activity. To consider competition with
independent ground handling providers since the liberalization in 1996 we further separated
the group of ground handling providing airports into (1) ground handling providing airports
without competing independent providers and (2) ground handling providing airports with at
least one competitor26. Unfortunately, ground handling provision must be analyzed separately
because of the high correlation with regulation (0.65), ownership (0.55) and the share of non-
aeronautical revenues (0.40).
Ownership form is defined according to (1) fully public airports, (2) public-private
airports with minor private shares (less than 50%), (3) public-private airports with major
private shares (above 50%) and (4) fully private airports.
The form of economic regulation has been categorized according to (1) no ex-ante
regulation27, (2) single till cost-plus regulated, (3) dual till cost-plus regulated, (4) single till
price-cap regulated and (5) dual till price-cap regulated airports. Unfortunately, this level of
refinement is not possible in the combined model due to an insufficient level of data hence we
aggregated the classification to ex-ante unregulated and regulated airports in the combined
environmental modelling approach.
Regional competition has been defined as the number of operating commercial airports
with at least 150,000 passengers per annum within a catchment area of 100 km around the
airport. The radius of the catchment area has been defined in line with Bel and Fageda (2010).
In addition, competition between gateway airports is considered. Consequently, weak
competition is defined as competition at the regional level of no more than a single airport and
strong competition as a location with at least two nearby competitors or possessing hub status.
Due to lack of information for the whole sample, we were not able to capture different
product diversification strategies, such as low cost carrier traffic, which may limit the level of
26 We refer to baggage and ramp handling which have been ground handling activities protected from competition prior to liberalization.
27 For simplicity we refer to airports subject to ex-post standard anti-trust regulation as unregulated airports.
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
89
competition amongst nearby airports. Hence, our conservative measure indicates the
maximum level of competition between airports. Table 7 presents the combinations and the
number of airports and observations that belong to the different groups in the combined
model.
Tab. 7: Combination of environmental variables analyzed
Weak local competition
Heavy local and hub
competition no. of
airports no. of obs.
no. of airports
no. of obs.
No ex-ante regulation 6 37 5 36 Public Ex-ante regulation 9 59 7 60 No ex-ante regulation 0 0 2 8 Minor private Ex-ante regulation 2 13 3 24 No ex-ante regulation 1 2 2 19 Major private Ex-ante regulation 2 9 2 15 No ex-ante regulation 5 33 6 53 Fully private Ex-ante regulation 2 6 3 24
3.5 Empirical results
In the following section we first discuss the DEA efficiency results. The complete set of
normalized DEA efficiency scores is listed in Table 12 in Appendix 3.A. The second part of
this section discusses the results of the regression analyses on the DEA efficiency estimates,
initially discussing the individual impacts and followed by the combined environmental
regression approach. The third part examines the combined impact as a function of revenues
per passengers and aircraft movements respectively in order to approximate the pricing
behaviour of airports under different institutional and market conditions.
3.5.1 Efficiency scores from data envelopment analysis
The average efficiency score of the dataset obtained from the input-oriented additive DEA
model is 0.62 (after normalization) with 14% of all airports categorized as relatively efficient.
The majority of airports exhibit an efficiency decrease over time which proved significant
according to the paired sample Wilcoxon sign rank test (p-value = 0.000). As a result of the
general economic downturn and the attacks on the world trade centre in New York in 2001,
the majority of airports experienced stable or declining traffic rates with disproportional
increases in staff and other operating costs. Increased security measures for baggage screening
require additional training and the recruitment of specialized workers, expenses which have
been covered at least partially by the airports (Vienna International Airport 2004). Hence one
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
90
could argue that the additional costs provide an increased quality with respect to safety related
to passenger traffic.
Consistently efficient airports include Ljubljana and Malta that represent the smaller
airports in the dataset, many of the Australian airports, as well as the largest operators,
Frankfurt and London-Heathrow. In Australia, domestic terminals are often operated by
incumbent airlines under long-term leases, thereby lowering maintenance and staff costs
(Hooper et al. 2000). In 2002, both Melbourne and Perth experienced efficiency drops of 20%
and 30% respectively. After the collapse of the Australian airline Ansett in 2001, their
dedicated domestic terminals were sold back to the airport owner in 2002 thereby increasing
staff and other operating costs (ACCC 2003). Costs remained relatively consistent thereafter
enabling Melbourne to achieve relative efficiency by 2004 and Perth two years later, as a
result of both traffic and revenue increases. Frankfurt and London-Heathrow obtain
reasonably high cost efficiency estimates over time. It should be noted that both airports are
severely congested and require airside capacity expansions. Whereas Frankfurt has long been
fighting for the construction of a fourth runway which is now expected to open in 2011,
Heathrow was denied the right to construct a third runway in May 2010 by the new UK
government (The Guardian 2010; Fraport 2009). Both airports place great emphasis on cost
efficiency with Heathrow attempting to minimize staff costs and Frankfurt tending to reduce
other operating costs. The diverse strategies are not surprising given the different levels of
outsourcing including ground handling provision. Whereas Frankfurt provides ground
handling services in-house, this operation has long been operated by airlines and third-party
providers at Heathrow.
The airports in Amsterdam, Brussels, Copenhagen, Dortmund, Dusseldorf, Leeds-
Bradford, London-Gatwick and Nice operated on the Pareto frontier at the beginning of their
respective review periods but all experience substantial decreases over time. Basel-Mulhouse,
Bratislava, Marseille, Tallinn and Zurich were inefficient throughout the timeframe and show
further efficiency declines over time. Many of these airports both increased their costs and
served lower traffic throughput which explains the decreasing efficiency scores. In addition,
Basel-Mulhouse and Bratislava suffer from heavy reductions in their cargo operations which
are not fully compensated by passenger growth rates. An increase in declared runway capacity
at Zurich decreased their relative efficiency score from 2005 by 14%. On the other hand, the
average delay per movement dropped from 10.35 to 5.75 minutes between 2005 and 200728,
28 Eurocontrol (2006-2008).
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
91
mainly due to a reduction in the length of the departure queues, which we have been unable to
consider in this analysis due to a lack of comparable data.
At Brussels and Dortmund, efficiency estimates dropped from 1.00 to 0.42 over time.
Brussels suffered heavily from the bankruptcy of the home carrier Sabena in 2001, with a
substantial decrease in traffic whilst concurrently increasing staff and other operating costs.
Compared to their benchmark in Nice, Brussels ought to lower costs in order to achieve
relative efficiency. Dortmund completed large and expensive capacity expansions on the
terminal and airside yet is located in a highly competitive corridor with Dusseldorf, Munster-
Osnabrueck and Paderborn airports within a 90 km radius as well as alternative transport
modes, hence may find it difficult to fill excess capacity even in the medium term.
Furthermore, Dortmund airport report operating losses in all years under review.
Except for Sydney, no airport consistently improved their relative efficiency scores over
time. Between 2003 and 2007, Sydney increased its score from 0.56 to 1.00 which is mainly
attributable to a large increase in non-aeronautical revenues with fairly constant cost inputs. In
contrast to Melbourne and Perth, a cost increase from the sale of the Ansett terminal back to
Sydney’s airport management is not reported as the review period begins in 2003. However,
the ACCC responsible for the price monitoring of aeronautical charges and car parking fees at
the top five Australian airports accused Sydney airport of abusing market power. In March
2010, the ACCC reported that the airport had substantially increased passenger charges and
that the car parking fees had almost doubled from 2008 to 2009 (ACCC 2010). Southampton
airport reached an efficiency peak in 2003 after a substantial increase in cargo operations.
However, a reduction in non-aeronautical revenues in 2004 decreased their efficiency
estimates compared to that of Leeds-Bradford and Ljubljana, their reference peers.
Average efficiency scores are achieved by many of the small to medium sized airports
with less than 10 million passengers per year and this proved to be reasonably consistent over
the review period. The airports of Budapest, Cologne-Bonn, Hanover, Leipzig, Lyon,
Manchester, Munich and Vienna appear to be the least relatively cost efficient airports in the
sample. Vienna, for example, has higher staff and other operating costs compared to its
benchmarks, including London-Gatwick, Nice and Sydney. Manchester airport is also more
expensive than its benchmarks, including Gatwick, Ljubljana and Melbourne. Athens
underwent substantial capacity expansions and a new green-field location hence capacity
utilization is low in comparison to reference airports such as Gatwick and Nice. Furthermore,
the German airports of Cologne-Bonn and Leipzig suffer from excess airside capacities
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
92
despite the extensive cargo operations resulting from their positions as the European hubs for
UPS and DHL respectively. Hanover also suffers from excessively low capacity utilization
and exhibits relatively high operating costs compared to its benchmarks. However, as service
quality indicators such as congestion and delay are not included in the first stage analysis, the
inefficiency may be somewhat overestimated.
3.5.2 Regression results explaining cost efficiency
In this section we analyze the impact of environmental variables on the DEA cost
efficiency scores and also include time indicators as identified in Section 3.5.1. Tables 8 and 9
present the results obtained from the three regression models introduced in Section 3.3 with
the former presenting the results of the modelling approach in which the environmental
impacts are analyzed individually and the latter presenting the joint effects. All three models,
although based on substantially different underlying assumptions, clearly highlight general
trends, despite the fact that the truncated regression removes all efficient observations from
the analysis and Tobit regression censors the score of efficient units at one. The base case for
Table 8 is defined as a monopolistic unregulated public airport with less than 50% non-
aeronautical revenues, no heavy delays, intercontinental traffic of less than 15% and capacity
utilization below 50%. Due to the high correlation between the ground handling dummy and
ownership form, it was not possible to include both sets of variables in a single model hence
we report both sets of results per regression.
All time trend data prove increasingly negative and statistically significant; hence
coefficient estimates of the explanatory variables are adjusted by time-related effects. The
dummy variable defining airports that earn more than 50% of their revenues from non-
aeronautical sources29 prove weakly positively significant, supporting the results of Oum et al.
(2006) and indicating a marginal contribution to cost efficiency of approximately 8% as
occurred at Sydney airport after privatization. A substantial relative share of intercontinental
traffic proves statistically insignificant across all regressions. On the other hand, delay and
congestion have a statistically significant impact on airport cost efficiency, as discussed in
Pathomsiri et al. (2008). Delay impacts cost efficiency negatively in the region of 10 to 20%.
In contrast, efficiency significantly increases with runway capacity utilization. Congestion
(proxied by capacity utilization above 90% which includes gateway airports Frankfurt and
29 Note that revenues from ground handling services were excluded from the figure to compare the results with previous studies.
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
93
London-Heathrow) has a large positive impact ranging from 37% to 64% increases in airport
efficiency compared to underutilized airports. Given that the positive impact of congestion is
higher than the negative effect of delays, this would indicate the need for service quality
indicators written into contracts between airlines and airports or internalization through
compensation to airlines and passengers for airport related delays.
Across the board, airports with less than 50% of shares traded privately prove to be
significantly less efficient than other ownership forms, a group which includes Athens,
Hanover and Vienna. Fully private airports do not prove to be statistically significantly
different in terms of cost efficiency than their fully public counterparts, in line with Oum et al.
(2006). Unregulated airports generally dominate their regulated counterparts, such as
Ljubljana and a number of Australian and British airports. Cost-plus regulation appears to be
the least appropriate form of economic regulation whether single or dual till, reducing
efficiency by 14% to 19%. Single till price-caps also appear to be dominated by dual till
price-caps and standard ex-post anti-trust monitoring. The importance of local and gateway
competition is not statistically significant in the models, in line with Chi-Lok and Zhang
(2009).
Our empirical results on ground handling reveal that airports operating this service under
monopolistic conditions are not significantly less cost efficient than non-ground handling
providing airports since only the truncated regression model indicates weakly negative
significance. According to Barbot (2010), airports operating as the sole provider may charge
higher prices for ground handling services if passenger and landing charges are capped. Given
that airports within this category are mostly subject to cost-plus regulation (Bremen,
Dortmund, Dresden, Leipzig, Nuremberg and Salzburg) it would appear that airports
generally charge higher ground handling fees given their level of market power. If airports are
in competition with at least one independent ground handling provider, it appears that they
operate on average 15% less cost efficiently. In line with the analytical result of Barbot
(2010), it would appear that aeronautical charges are not sufficiently capped at the airports in
Cologne-Bonn, Dusseldorf, Hanover, Munich and Stuttgart, which are mostly cost-plus
regulated, thereby permitting cross-subsidization leading to lower ground handling revenues
in order to remain ‘competitive’ with independent providers.
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
94
Tab. 8: Second-stage regression results from the individual cost efficiency model
Robust Cluster OLS Regression
Robust Cluster Tobit Regression
Robust Cluster Truncated Regression
Dependent Variable DEA Efficiency Scores DEA Efficiency Scores DEA Efficiency Scores
a) Airport Characteristics and Management Strategies Monopolistic ground handling operation - -0.089 (0.059) - -0.097 (0.067) - -0.081 (0.044)*
Competitive ground handling operation - -0.149 (0.047)*** - -0.162 (0.047)*** - -0.134 (0.047)***
Share of non-aviation revenues >50% 0.080 (0.040)** - 0.094 (0.043)*** - 0.045 (0.035) -
Share of intercontinental traffic > 15% 0.021 (0.040) - 0.032 (0.042) - -0.035 (0.039) -
Heavy delays -0.164 (0.039)*** -0.190 (0.045)*** -0.184 (0.045)*** -0.211 (0.052)*** -0.109 (0.037)*** -0.116 (0.038)***
Runway capacity utilization 50-90% 0.088 (0.034)*** 0.058 (0.039) 0.095 (0.036)*** 0.069 (0.042) 0.059 (0.031)** 0.008 (0.034)
Runway capacity utilization > 90% 0.452 (0.063)*** 0.450 (0.070)*** 0.591 (0.068)*** 0.642 (0.087)*** 0.407 (0.050)*** 0.368 (0.033)***
b) Ownership, Regulation and Competition Minor private airport -0.124 (0.040)*** - -0.127 (0.043)*** - -0.116 (0.036)*** -
Major private airport 0.138 (0.062)*** - 0.173 (0.082)*** - 0.024 (0.041) -
Fully private airport 0.004 (0.047) - 0.001 (0.051) - 0.028 (0.045) -
Heavy competition 0.029 (0.035) 0.052 (0.044) 0.030 (0.037) 0.056 (0.049) 0.015 (0.034) 0.039 (0.037)
Cost-plus regulation, single till -0.171 (0.050)*** - -0.187 (0.059)*** - -0.145 (0.042)*** -
Cost-plus regulation, dual till -0.149 (0.039)*** - -0.155 (0.041)*** - -0.140 (0.039)*** -
Price-cap regulation, single till -0.129 (0.049)*** - -0.133 (0.052)*** - -0.084 (0.047)* -
Price-cap regulation, dual till 0.032 (0.062) - 0.045 (0.079) - -0.047 (0.043) -
C) Time Trend Year 1999 -0.015 (0.020) 0.001 (0.024) -0.022 (0.026) -0.005 (0.031) 0.011 (0.018) 0.019 (0.018)
Year 2000 -0.033 (0.028) -0.002 (0.034) -0.036 (0.032) -0.003 (0.041) -0.011 (0.023) 0.014 (0.028)
Year 2001 -0.089 (0.033)*** -0.054 (0.040) -0.113 (0.040)*** -0.080 (0.049) -0.008 (0.027) 0.007 (0.032)
Year 2002 -0.106 (0.037)*** -0.071 (0.044) -0.127 (0.044)*** -0.092 (0.052) -0.027 (0.027) -0.009 (0.034)
Year 2003 -0.142 (0.039)*** -0.107 (0.045)*** -0.168 (0.046)*** -0.135 (0.054)*** -0.062 (0.026)*** -0.040 (0.035)
Year 2004 -0.122 (0.039)*** -0.088 (0.046)** -0.143 (0.046)*** -0.109 (0.054)*** -0.051 (0.026)** -0.028 (0.035)
Year 2005 -0.147 (0.039)*** -0.112 (0.045)*** -0.167 (0.046)*** -0.132 (0.053)*** -0.085 (0.025)*** -0.064 (0.033)**
Year 2006 -0.168 (0.041)*** -0.134 (0.046)*** -0.195 (0.049)*** -0.160 (0.054)*** -0.090 (0.028)*** -0.080 (0.036)***
Year 2007 -0.173 (0.049)*** -0.138 (0.051)*** -0.197 (0.057)*** -0.159 (0.059)*** -0.109 (0.035)*** -0.099 (0.040)***
Intercept 0.794 (0.052)*** 0.786 (0.068)*** 0.824 (0.061)*** 0.824 (0.080)*** 0.706 (0.045)*** 0.678 (0.053)
R2 0.5678 0.3823 0.561 0.374 0.4859 0.3572
Observations (n) 398 398 398 (342 uncensored)
398 (342 uncensored) 342 342
Note: *, **, *** indicate the level of significance at 10%, 5% and 1% respectively. Robust standard errors in parentheses; clustered at airport level. R2 in Tobit and truncated regression calculated as a rough estimate of the degree of association by correlating the dependent variable with the predicted value and squaring the result.
Table 9 presents the results of the combined model in which the monopolistic, minor
private, regulated airport defines the base case. The time trend dummies prove to be
statistically significant across all regressions from 2002. Hence, after accounting for the
efficiency decreases over time, we conclude that ownership form, competition and regulation
play an important role in explaining efficiency differences across airports both individually
and in combination.
Under weak competitive conditions, defined as at most one airport within the catchment
area, privatized airports with at least 50% of the shares in private hands are the most efficient
ownership form. In comparison to minor private airports (the base case), the major or fully
privatized counterparts are on average 30% more efficient when regulated and 15% more
efficient when unregulated suggesting that economic regulation is desirable. Referring back to
the individual regression model results presented in Table 8, dual till price-cap regulation
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
95
would appear to be the most preferable instrument. The unregulated major and fully private
airports Aberdeen, Bratislava, Melbourne (from 2002), Perth (from 2002) and Sydney
perform on average 25% less efficiently than their regulated counterparts Copenhagen,
Melbourne (until 2002), Malta and Perth (until 2002), after accounting for the time trend.
Purely public airports are also strictly preferable to their minor private counterparts.
Furthermore, publicly owned and regulated airports (Dresden, Dublin, Leipzig, Nuremberg,
Oslo, Salzburg, Stuttgart and Tallinn) perform on average 10% less efficiently than their
unregulated counterparts (Basel-Mulhouse, Bratislava, Budapest, Geneva, Lyon, Riga). This
indicates that managers of public airports behave as welfare maximizers and additional
economic regulation decreases relative cost efficiency. However, it should be noted that the
group of regulated public airports are predominantly regulated according to a cost-plus regime
which according to the individual regression model is significantly less efficient than no ex-
ante regulation. Hence this may explain why the unregulated public airports in a non-
competitive setting appear to be more cost efficient that their regulated counterparts.
In a competitive environment, unregulated purely public airports (Leeds-Bradford,
Marseille and Nice) and major or fully privatised airports (Edinburgh, Glasgow and London-
Stansted) are equally cost efficient. It is clear that such airports do not require economic
regulation to maintain cost efficiency as compared to airports operating in a weakly
competitive environment. The public regulated airports, namely Amsterdam, Bremen,
Cologne-Bonn, Dortmund, Munich and Manchester, perform on average 30% less cost
efficiently than their public unregulated counterparts. Major and fully private unregulated
airports perform to the order of 10% to 15% more cost efficiently than their regulated
counterparts. Apart from minor private ownership, unregulated airports located in a
competitive environment generally operate more efficiently than those operating in weakly
competitive surroundings. For example, the unregulated fully private airports, Aberdeen and
Belfast, are located in uncompetitive environments and operate significantly less efficiently
than the more competitive Edinburgh, Glasgow, London-City and Southampton examples.
Among the regulated public airports, Nuremberg, Stuttgart and Salzburg located in weakly
competitive environments are substantially less cost efficient than the competitive Cologne-
Bonn, Dortmund and Munich examples. Consequently, government intervention would
appear to incur high transaction costs and is required to emulate the competitive environment
when missing but is very expensive when such conditions already exist in the market. Hence,
the results clearly show that competition replaces the need for economic regulation
irrespective of ownership form in order to encourage cost efficiency.
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
96
Tab. 9: Second-stage regression results from the combined cost efficiency model
Robust ClusterOLS
Regression
Robust Cluster Tobit Regression
Robust Cluster Truncated Regression
Dependent Variable DEA Efficiency Scores
a) Airport Characteristics and Management Strategies Share of non-aviation revenues >50% 0.068 (0.037)* 0.078 (0.040)** 0.029 (0.032)
Heavy delays -0.182 (0.042)*** -0.200 (0.047)*** -0.122 (0.036)***
Runway capacity utilization 50-90% 0.098 (0.033)*** 0.106 (0.036)*** 0.053 (0.028)*
Runway capacity utilization > 90% 0.502 (0.087)*** 0.660 (0.102)*** 0.361 (0.056)***
b) Ownership, Regulation and Competition No regulation 0.188 (0.048)*** 0,183 (0,047)*** 0.205 (0.055)***
Public Regulation 0.093 (0.057)* 0,083 (0,058) 0.130 (0.053)***
No regulation 0.127 (0.045)*** 0,116 (0,047)*** 0.168 (0.046)*** Major private Regulation 0.438 (0.091)*** 0,501 (0,138)*** 0.228 (0.041)***
No regulation 0.147 (0.084)* 0,139 (0,089) 0.138 (0.065)** Low
com
petit
ion
Fully private Regulation 0.313 (0.091)*** 0,379 (0,105)*** 0.445 (0.146)***
No regulation 0.330 (0.033)*** 0,335 (0,032)*** 0.332 (0.041)*** Public
Regulation 0.028 (0.040) 0,021 (0,040) 0.027 (0.045)
No regulation 0.110 (0.044)*** 0,112 (0,043)*** 0.118 (0.045)*** Minor private Regulation 0.048 (0.050) 0,029 (0,044) 0.025 (0.039)
No regulation 0.351 (0.107)*** 0,386 (0,144)*** 0.213 (0.032)*** Major private Regulation 0.247 (0.037)*** 0,259 (0,036)*** 0.169 (0.051)***
No regulation 0.273 (0.052)*** 0,269 (0,053)*** 0.284 (0.051)***
Hea
vy c
ompe
titio
n
Fully private Regulation 0.158 (0.066)*** 0,165 (0,065)*** 0.257 (0.050)***
C) Time Trend Year 1999 -0.018 (0.021) -0.026 (0.026) 0.005 (0.019)
Year 2000 -0.039 (0.028) -0.042 (0.032) -0.017 (0.023)
Year 2001 -0.094 (0.033)*** -0.120 (0.040)*** -0.030 (0.025)
Year 2002 -0.106 (0.037)*** -0.127 (0.043)*** -0.039 (0.025)
Year 2003 -0.139 (0.039)*** -0.166 (0.046)*** -0.070 (0.025)***
Year 2004 -0.120 (0.039)*** -0.141 (0.045)*** -0.059 (0.025)***
Year 2005 -0.139 (0.038)*** -0.158 (0.044)*** -0.091 (0.024)***
Year 2006 -0.155 (0.039)*** -0.178 (0.046)*** -0.096 (0.026)***
Year 2007 -0.143 (0.044)*** -0.161 (0.050)*** -0.100 (0.032)***
Intercept 0.591 (0.055)*** 0.624 (0.063)*** 0.500 (0.053)***
R2 0.6254 0.6165 0.5421
Observations (n) 398 398 (342 uncensored) 342
Note: *, **, *** indicate the level of significance at 10%, 5% and 1% respectively. Robust standard errors in parentheses; clustered at airport level. R2 in Tobit and truncated regression calculated as a rough estimate of the degree of association by correlating the dependent variable with the predicted value and squaring the result
In summary, minor private airports appear to be the least efficient ownership form. Under
weakly competitive conditions, dual till price caps appears to be the most appropriate form of
economic regulation. However, under competitive market conditions with respect to
catchment area and hub status, regulation is not effective irrespective of ownership form.
According to this empirical analysis, there would not appear to be a most efficient ownership
structure, supporting the theoretical arguments of Vickers and Yarrow (1991) that competition
is more important than ownership form with respect to efficiency. Finally, since the level of
competition is an exogenous factor at least in the short term, economic regulation is an
effective tool to engender cost efficiency when market conditions are poor.
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
97
3.5.3 Regression results explaining airport charges
In this section we analyze the same set of variables as described in Table 9 against the
estimated (logged) aeronautical revenues per passenger and aircraft movement in order to
approximate the pricing behaviour under different institutional settings. Additionally, we have
included the (logged) average aircraft size in order to capture the substantial differences in
fleet mix that exist within the sample. Table 10 presents the results obtained from the robust
cluster OLS regression introduced in Section 3.3. From our sample, the Australian airports in
Melbourne and Perth earned the lowest revenues per passenger and aircraft movement whilst
subject to price-cap regulation (until 2002) hence the base case for Table 10 is defined as a
monopolistic, regulated fully private airport with less than 50% non-aeronautical revenues, no
heavy delays and capacity utilization below 50%. Similar to Section 3.5.2, time dummies
were include in the analysis and show a consistent increase in revenue over time except for a
drop in 2002 as a result of the traffic decline following the terror attacks in New York in
2001.
The categorical variable representing airports that generate more than 50% of their
revenues from non-aeronautical activities indicates that such airports earn 19% less on
average from the aeronautical activities than otherwise. In line with the analytical findings of
Zhang and Zhang (2010), these airports may cross-subsidize their aeronautical costs from
additional sources in order to further attract both airlines and passengers. Airports suffering
from heavy delays appear to charge significantly lower aeronautical revenues (9%). We
observe that the airports in our sample that are not listed as heavily delayed are most
frequently either unregulated or subject to cost-plus regulation. We assume that these airports
are charging higher aeronautical fees than the average hence the negative sign in the OLS
regression. Finally, aeronautical revenues at congested airports are significantly higher (23%),
thereby supporting the empirical outcome of van Dender (2007) that congested airports are in
a position to exploit scarcity rents. The average aircraft size was found to have a positive
impact on the revenues per movement due to the fact that such charges are weight-based. An
increase in the average aircraft size of 10% leads to an 8% increase in revenues per
movement. Revenues per passenger have a significantly negative impact because transit
passenger charges are generally significantly lower than that of originating passengers; hence
a 10% increase in the average aircraft size leads to a 2% decrease in revenues per passenger.
The results for the weakly competitive environment reveal that unregulated airports
irrespective of ownership form earn higher aeronautical revenues per passenger and
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
98
movement than their regulated counterparts, suggesting that regulation is necessary from the
revenue perspective too in order to prevent exploitation of market power. The unregulated
fully private airports, Aberdeen, Belfast, Melbourne (from 2002), Perth (from 2002) and
Sydney are paid on average 27% more per passenger and movement than Melbourne and
Perth prior to the introduction of a monitoring process. The difference between public
regulated and unregulated airports, in contrast, is relatively small (5%) suggesting that public
airports are less likely to abuse market power. However, it is worth noting that the majority of
public regulated airports in this category are subject to cost-plus regulation hence it may be
true that the regulated charges are higher than would be expected under price-caps.
Tab. 10: Second-stage regression results from the combined revenue model
Robust Cluster OLS Regression
Dependent Variable revenues per passenger (log)
revenues per movement (log)
a) Airport Characteristics and Management Strategies Share of non-aviation revenues >50% -0.185 (0.058)*** -0.185 (0.058)***
Heavy delays -0.090 (0.039)*** -0.090 (0.039)***
Runway capacity utilization between 50-90% 0.028 (0.035)*** 0.028 (0.036)***
Runway capacity utilization > 90% 0.226 (0.050)*** 0.227 (0.050)***
Average aircraft size (log) -0.190 (0.150)*** 0.809 (0.151)***
b) Ownership. Regulation and Competition No regulation 0.450 (0.103)*** 0.452 (0.103)***
Public Regulation 0.397 (0.056)*** 0.397 (0.056)***
Minor private Regulation 0.474 (0.110)*** 0.473 (0.112)***
No regulation 0.656 (0.051)*** 0.656 (0.051)*** Major private Regulation 0.459 (0.110)*** 0.460 (0.109)*** Lo
w c
ompe
titio
n
Fully private No regulation 0.274 (0.035)*** 0.276 (0.035)***
No regulation 0.298 (0.055)*** 0.299 (0.055)*** Public
Regulation 0.387 (0.095)*** 0.388 (0.095)***
No regulation 0.414 (0.072)*** 0.416 (0.072)*** Minor private Regulation 0.470 (0.079)*** 0.471 (0.079)***
No regulation 0.489 (0.053)*** 0.491 (0.054)*** Major private Regulation 0.374 (0.081)*** 0.375 (0.082)***
No regulation 0.437 (0.051)*** 0.439 (0.051)***
Com
petit
ion
Fully private Regulation 0.419 (0.054)*** 0.419 (0.055)***
C) Time Trend Year 1999 0.033 (0.010) 0.032 (0.010)
Year 2000 0.052 (0.020) 0.052 (0.020)
Year 2001 0.093 (0.021)*** 0.092 (0.021)***
Year 2002 0.083 (0.020)*** 0.082 (0.020)***
Year 2003 0.106 (0.023)*** 0.104 (0.023)***
Year 2004 0.12 (0.030)*** 0.118 (0.030)***
Year 2005 0.159 (0.029)*** 0.157 (0.029)***
Year 2006 0.147 (0.036)*** 0.144 (0.037)***
Year 2007 0.173 (0.039)*** 0.171 (0.039)***
Intercept 0.628 (0.256)*** 0.629 (0.257)***
R2 0.5838 0.6697
Observations (n) 398 398
Note: *, **, *** indicate the level of significance at 10%, 5% and 1% respectively. Standard errors in parentheses. Robust standard errors in parentheses; clustered at airport level.
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
99
When operating under local or hub competition, unregulated public and minor private
airports obtain on average 8% lower aeronautical revenues per passenger and movement,
indicating that they are effectively disciplined by the market structure. Unregulated major
and fully private airports on the other hand charge higher fees than their regulated
counterparts, indicating that irrespective of the competitive setting revenues are maximized.
However, the difference in revenues between unregulated and regulated airports is smaller
than under monopolistic conditions. The results support the findings of Bel and Fageda (2010)
and Oum et al. (2004) that unregulated private airports charge higher aeronautical prices.
Furthermore, the results indicate that in contrast to cost efficiency, ownership plays a major
role with respect to the pricing behaviour of airports.
In summary, without local or hub competition, airports of any ownership form ought to be
regulated in order to encourage cost efficiency and to prevent the abuse of market power. In
competitive environments, ex-ante regulation appears to engender cost inefficiency. However,
unregulated major and fully private airports charge higher prices than their regulated
counterparts suggesting that dual till price-cap regulation or possibly yardstick competition
may still need to play a role. Purely public airports in a competitive setting, on the other hand,
could dispense with regulation since they appear to be relatively cost efficient without
charging excessively.
3.6 Conclusions
The inefficiency of airports may be explained not only by input excess and output
shortfalls but also by exogenous factors over which management have little to no control. A
number of empirical studies have assessed the impact of ownership structure, economic
regulation and levels of competition on efficiency however the effects were always
considered separately hence significance was sometimes an issue. The aim of this research
has been to assess the combined impact of the environmental variables in order to gain
understanding as to the most efficient ownership form and regulatory framework whilst
accounting for levels of regional and hub competition as well as other managerial choices.
The two-stage analysis combined DEA in the first stage and regression analysis in the
second stage. The non-radial additive input-oriented DEA model has been chosen to identify
all relative inefficiencies of the input. Following the recent debate on the most appropriate
second-stage regression model, Banker and Natarajan (2008) and Simar and Wilson (2007)
propose standard OLS and truncated regression respectively. Due to issues of
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
100
heteroscedasticity in the error terms, we applied robust clustering to each of these forms and
also included a Tobit regression to ensure that the results are robust. The regression results
proved robust since the outcomes were similar and the general directions were clear across all
three modelling approaches.
Data availability remains a serious issue and the attempt to include all combinations of
institutional settings proved difficult. However, the results of the joint analysis provide
additional information beyond that of the individual regression models. The results suggest
that ex-ante regulation at all airports located in a competitive environment is unnecessary and
generates x-inefficiency of the order of 15%, which rises substantially at purely public
airports. However, unregulated major and fully private airports located in a competitive
setting still pursue profit maximization and charge higher aeronautical fees than regulated
airports of the same ownership structure. Herein lies the trade-off between the x-inefficiency
generated as a result of regulation and the reduction in the abuse of market power as a result
of price caps. On the other hand, non-hub airports with weak local competition generally
require economic regulation in order to prevent an exploitation of market power and to
encourage cost efficiency. Dual till price-cap regulation would appear to be the most efficient
form. Combining the results from the efficiency and revenue model reveals that public
ownership appeared to be the preferable welfare maximizing combination, whereby regulation
is only necessary under monopolistic conditions.
Additional results reveal that ground handling providing airports operating in competition
with independent providers tend to charge lower ground handling fees in order to be
competitive. Public airports particularly appear to cross-subsidize from aeronautical revenues
in order to cover their higher operating costs. Whilst heavy delays impact cost efficiency
negatively, high runway utilization increases efficiency by more than double the negative
impact of delay. This would suggest that airlines may require contracts with service quality
specifications or penalties in order to encourage congested airports to internalize the delay
externalities.
Having defined competition on the regional and hub level in a relatively simple manner,
we may have assumed excessive rates of competition having ignored product diversification
strategies such as low cost carrier traffic or market destination separation. In other words, we
categorized airports as competitive that may be avoiding direct competition. Nevertheless, the
rather conservative measure achieves consistent results and would only prove more significant
had the data been more refined.
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
101
Future research would require substantially more data to permit an improved analysis of
all the categories described here. Finer distinctions with respect to ownership form and
regulation might better highlight the most efficient institutional setting given alternative levels
of competition. Additional environmental variables, including airport-related delays, noise
and air pollution, would enable the development of a social welfare analysis of airports and
the trade-off across the different stakeholders. Finally in order to benchmark, a more accurate
measure of the capital required to build, maintain and expand or reduce airports would only
be possible if the ACI, ICAO or equivalent organization were to develop standardized data
collection procedures that airports globally reported annually.
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
102
3.A Appendix
Tab. 11: List of airports
Code Airport Country Time Period Ownership Regulation Competition Ground
Handling ABZ Aberdeen UK 99-05 fully private no ex-ante regulation weak not provided
98-06 cost-plus AMS Amsterdam NL 2007
public incentive
heavy not provided
ATH Athens GR 05-07 minor private cost-plus weak not provided BFS Belfast UK 99-07 fully private no ex-ante regulation weak not provided BHX Birmingham UK 98-07 major private no ex-ante regulation heavy not provided BLQ Bologna IT 00-05 public no ex-ante regulation heavy provided BRE Bremen DE 98-07 public cost-plus heavy provided BRU Brussels BE 99-04 major private cost-plus heavy not provided
03-05 public BTS Bratislava SK 06-07 major private
no ex-ante regulation weak provided
BUD Budapest HU 00-01 public no ex-ante regulation weak not provided CGN Cologne Bonn DE 98-07 public cost-plus heavy provided CPH Copenhagen DK 01-04 major private incentive weak not provided DRS Dresden DE 98-06 public cost-plus weak provided DTM Dortmund DE 98-07 public cost-plus heavy provided DUB Dublin IE 06-07 public incentive weak not provided DUS Dusseldorf DE 99-07 major private cost-plus heavy provided EDI Edinburgh UK 98-07 fully private no ex-ante regulation heavy not provided EMA East Midlands UK 98-06 fully private no ex-ante regulation heavy not provided FRA Frankfurt DE 02-07 minor private incentive heavy provided GLA Glasgow UK 98-06 fully private no ex-ante regulation heavy not provided GVA Geneva CH 98-07 public no ex-ante regulation weak not provided HAJ Hanover DE 98-07 minor private cost-plus weak provided
98-99 public cost-plus HAM Hamburg DE 00-07 minor private incentive
heavy provided
LBA Leeds Bradford UK 98-02, 06-07 public no ex-ante regulation heavy not provided
LCY London-City UK 99-07 fully private no ex-ante regulation heavy not provided LEJ Leipzig DE 98-06 public cost-plus weak provided LGW London-Gatwick UK 98-05 fully private incentive heavy not provided LHR London-Heathrow UK 98-05 fully private incentive heavy not provided LJU Ljubljana SI 98-06 major private no ex-ante regulation heavy provided LTN London-Luton UK 00-07 fully private no ex-ante regulation heavy not provided LYS Lyon FR 98-06 public no ex-ante regulation weak not provided MAN Manchester UK 98-07 public incentive heavy not provided
99-01 incentive MEL Melbourne AU 02-07
fully private no ex-ante regulation
weak not provided
MLA Malta MT 02-06 major private incentive weak not provided MLH Basel Mulhouse FR 98-07 public no ex-ante regulation weak not provided MRS Marseille FR 98-06 public no ex-ante regulation heavy not provided MUC Munich DE 98-05 public cost-plus heavy provided NCE Nice FR 98-06 public no ex-ante regulation heavy not provided NUE Nuremberg DE 98-07 public cost-plus weak provided
99-03 cost-plus OSL Oslo NO 04-07
public incentive
weak not provided
99-01 incentive PER Perth AU 02-07
fully private no ex-ante regulation
weak not provided
RIX Riga LV 04-06 public no ex-ante regulation weak provided SOU Southampton UK 99-05 fully private no ex-ante regulation heavy not provided STN London-Stansted UK 98-06 fully private incentive heavy not provided STR Stuttgart DE 98-07 public cost-plus weak provided SYD Sydney AU 03-07 fully private no ex-ante regulation weak not provided
2004 cost-plus SZG Salzburg AT 05-07
public incentive
weak provided
TLL Tallinn EE 02-07 public cost-plus weak provided 00-04 public VCE Venice IT 2005 minor private
no ex-ante regulation heavy not provided
VIE Vienna AT 98-07 minor private incentive heavy provided ZRH Zurich CH 01-07 minor private no ex-ante regulation heavy not provided
Joint Impact of Competition, Ownership Form and Economic Regulation on Airport Performance
103
Tab. 12: DEA efficiency scores
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 ABZ - 0,579 0,599 0,546 0,522 0,505 0,515 0,536 - - AMS 1,000 1,000 1,000 0,496 0,309 0,249 0,253 0,297 0,259 0,280 ATH - - - - - - - 0,421 0,432 0,450 BFS - 0,512 0,489 0,487 0,480 0,485 0,575 0,502 0,500 0,502 BHX 0,549 0,577 0,598 0,582 0,568 0,583 0,535 0,530 0,514 0,481 BLQ - - 0,755 0,715 0,710 0,733 0,698 0,740 - - BRE 0,587 0,648 0,602 0,589 0,585 0,568 0,575 0,568 0,562 0,562 BRU - 1,000 1,000 0,653 0,455 0,417 0,431 - - - BTS - - - - - 0,854 0,732 0,575 0,579 0,563 BUD - - 0,357 0,336 - - - - - - CGN 0,390 0,410 0,390 0,417 0,409 0,388 0,361 0,350 0,337 0,323 CPH - - - 1,000 0,617 0,562 0,635 - - - DRS 0,639 0,588 0,575 0,547 0,559 0,574 0,548 0,538 0,532 - DTM 1,000 0,843 0,732 0,472 0,433 0,418 0,408 0,410 0,410 0,408 DUB - - - - - - - - 0,504 0,475 DUS - 1,000 1,000 0,605 0,682 0,472 0,588 0,531 0,508 0,574 EDI 0,664 0,712 0,714 0,702 0,713 0,718 0,738 0,711 0,620 0,661 EMA 0,637 0,628 0,629 0,606 0,629 0,640 0,637 0,584 0,558 - FRA - - - - 1,000 1,000 1,000 1,000 0,732 1,000 GLA 0,702 0,712 0,714 0,705 0,693 0,685 0,697 0,703 0,692 - GVA 0,635 0,660 0,791 0,720 0,709 0,700 0,676 0,638 0,645 0,687 HAJ 0,313 0,319 0,331 0,316 0,314 0,310 0,308 0,301 0,302 0,297 HAM 0,535 0,524 0,426 0,394 0,399 0,381 0,420 0,404 0,406 0,414 LBA 1,000 1,000 0,901 0,830 0,842 0,881 0,888 - - - LCY - 1,000 1,000 0,933 0,887 0,867 0,799 0,744 0,888 1,000 LEJ 0,571 0,570 0,392 0,382 0,374 0,369 0,367 0,337 0,359 - LGW 1,000 0,861 1,000 0,744 0,682 0,619 0,658 0,537 - - LHR 1,000 1,000 1,000 0,907 1,000 0,852 1,000 1,000 - - LJU 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 - LTN - 0,584 0,518 0,525 0,513 0,520 0,564 0,511 0,529 0,473 LYS 0,487 0,504 0,525 0,509 0,459 0,443 0,448 0,414 0,417 - MAN 0,477 0,425 0,407 0,334 0,352 0,434 0,504 0,442 0,434 0,397 MEL - 0,862 1,000 1,000 0,815 0,760 0,903 1,000 1,000 1,000 MLA - - - - 1,000 1,000 1,000 1,000 1,000 - MLH 0,887 0,821 0,670 0,542 0,502 0,439 0,432 0,438 0,414 0,413 MRS 0,832 0,844 0,932 0,808 0,621 0,586 0,597 0,600 0,588 - MUC 0,335 0,338 0,355 0,353 0,332 0,263 0,264 0,234 - - NCE 1,000 1,000 1,000 0,857 0,684 0,612 0,543 0,526 0,538 - NUE 0,628 0,613 0,636 0,619 0,596 0,574 0,578 0,570 0,568 0,564 OSL - 0,555 0,517 0,515 0,440 0,479 0,441 0,458 0,494 0,543 PER - 1,000 1,000 0,972 0,790 0,689 0,678 0,673 0,703 1,000 RIX - - - - - - 0,547 0,558 0,563 - SOU 0,847 0,919 0,943 0,720 0,617 1,000 0,751 - - - STN 0,670 0,770 0,651 0,595 0,723 0,707 0,661 0,635 - - STR 0,561 0,564 0,559 0,534 0,513 0,505 0,462 0,516 0,513 0,519 SYD - - - - - 0,563 0,748 1,000 1,000 1,000 SZG - - - - - - 0,733 0,716 0,722 0,723 TLL - - - - 0,904 0,690 0,628 0,522 0,516 0,513 VCE - - 0,625 0,599 0,627 0,671 0,682 0,545 - - VIE 0,428 0,456 0,471 0,396 0,365 0,338 0,338 0,325 0,320 0,310 ZRH - - - 0,656 0,532 0,516 0,481 0,342 0,315 0,306
104
4 AIRPORT BENCHMARKING FROM A
MANAGERIAL PERSPECTIVE30
Benchmarking airports is currently popular both in the academic literature and in practice
but has proved rather problematic due to the heterogeneity inherent in any reasonably sized
dataset. Most studies either treat the airport production technology as a black box or separate
terminal and airside activities, assessing them individually. In this paper we analyze airports
as a single unit due to the direct complementarities, avoiding the artificial separation of inputs
between the terminal and airside but opening the black box through network data
envelopment analysis (DEA). To further improve the benchmarking process, we identify
appropriate peers for 43 European airports over 10 years through a dynamic clustering
mechanism according to pre-defined characteristics and restrict the integer linear program
with respect to potential reductions in capital inputs. Compared to basic DEA models, the
results of the network DEA structure provide more meaningful benchmarks with comparable
peer units and target values that are achievable in the medium term. By identifying each
airport’s individual reference set, unique airport outliers influence relative efficiency less
severely than occurs under basic DEA. In addition, the formulation is shown to be suitable in
assessing different strategies with respect to aeronautical and commercial activities not only
separately but also in combination.
30 The research of this chapter is obtained in collaboration with Ekaterina Yazhemsky, Hebrew University of Jerusalem. We are grateful to Dr. Nicole Adler for helpful discussions and support. The paper has been submitted to the Journal of Productivity Analysis in June 2010 and is currently under review.
Airport Benchmarking from a Managerial Perspective
105
4.1 Introduction
According to the Princeton dictionary, an airport is defined as “an airfield equipped with
control tower and hangars as well as accommodations for passengers and cargo”. Airports
can be defined as an important basic infrastructure to a society in which aviation is one of the
drivers of a modern economy. An alternative approach defines an airport as a private
production system in which society maximizes social welfare by encouraging airport
management to maximise profits whilst at the same time, considering consumer surplus via
some form of airport regulation if deemed necessary. Consequently, it is unclear whether
airports should be considered as a not-for-profit, public good, the general approach in the
United States, or as a private enterprise maximizing shareholder value. Since it would appear
to be true that large regions of the world are gradually adopting the privatized form (Zhang
and Zhang 2003) and that independent authorities running public airports in the United States
do not behave differently to their private counterparts with respect to productivity (Oum et al.
2006), in this paper we develop an airport benchmarking methodology from an airport
manager’s perspective in which we assume that the airport intends to maximize revenues or
minimize costs.
Oum et al. (2006) and Barros and Dieke (2007) review airport benchmarking studies
applied to a diverse range of activities using various methodologies. The most popular
methods include price index total factor productivity (Hooper and Hensher 1997; Oum and
Yu 2004; Vasigh and Gorjidooz 2006), parametric stochastic frontier analysis (Pels et al.
2003; Oum et al. 2008) and non-parametric data envelopment analysis (DEA). DEA has been
used to compare the performance of airports within national boundaries, U.S. (Gillen and Lall
1997; Sarkis 2000), U.K. (Parker 1999), Spain (Martín and Román 2008; Murillo-Melchor
1999), Australia (Abbott and Wu 2002), Taiwan (Yu 2004), Portugal (Barros and Sampaio
2004) as well as airports around the world (Adler and Berechman 2001; Lin and Hong 2006).
It is rather difficult to draw general inferences since many of these papers arrive at directly
opposing conclusions. For example, Murillo-Melchor (1999) show that Spanish airports in
their dataset suffer from decreasing returns-to-scale whereas Martín et al. (2009) concluded
increasing returns-to-scale for the same set of airports. Abbott and Wu (2002) found most
Australian airports enjoy increasing returns-to-scale, Pels et al. (2003) argue that European
airports operate under constant returns-to-scale in air traffic movements and increasing
returns-to-scale on the terminal side and Lin and Hong (2006) argue that most airports are not
operating at an optimal scale. Graham and Holvad (2000) and Abbott and Wu (2002) argue
Airport Benchmarking from a Managerial Perspective
106
that Australian airports are more efficient than their European counterparts, Lin and Hong
(2006) argue that the US and European airports are more efficient than their Asian and
Australian counterparts and Pels et al. (2003) conclude that widespread European airport
inefficiency is not specific to a country or region. Consequently, Morrison (2009) has called
for a balanced approach and dialogue between airport managers and researchers.
The majority of studies to date treat airport technology as a single production process
avoiding the complexity inherent in airport systems. Gillen and Lall (1997) and Pels et al.
(2003) were the first to argue that the airport could be analyzed as two separate decision-
making processes, one serving airside activities and the other serving landside production.
The approach developed in this research connects the two sides of the production function
whilst opening the black box via network DEA (Färe 1991). We argue that a single black box
approach would be insufficient to capture the rich picture underlying this approximation, as
demonstrated in Figure 11. Since the liberalization of the aviation industry in Europe in the
late eighties, airports have focused on both aeronautical and commercial landside activities.
The network DEA approach recognises the fact that generalized and fixed costs connected to
the two sets of activities can only be split in an artificial manner and that whilst aeronautical
revenues draw from passengers, cargo and air traffic movements, the non-aeronautical
revenue is more closely tied to passenger throughput. Although airports may have limited
control over traffic volume, non-aeronautical revenues drawn from non-airport related
activities, such as airport cities, are indeed within the purview of airport management. As
argued in Oum et al. (2003), the omission of outputs such as commercial services is likely to
bias efficiency results as it underestimates the productivity of airports whose managers focus
on generating additional revenue sources. Many airports attempt to increase revenues from
non-aeronautical sources which are not directly related to aviation activities in order to cross-
subsidize aviation charges in turn attracting more airlines and passengers to their airport
(Zhang and Zhang 2010). Revenue source diversification that exploits demand
complementarities across aeronautical and non-aeronautical services appears to improve
airport productive efficiency (Oum et al. 2006). We would argue that it is more reasonable to
analyze the airports as a single unit due to the direct complementarities yet avoid the need to
separate inputs between the terminal and airside. In general, the airport technology may be
defined as a network that consists of multi-production processes and stages as described in
Figure 11. Consequently, in this paper we develop a network DEA modelling approach in
order to measure the relative cost and revenue efficiencies of airports with respect to
Airport Benchmarking from a Managerial Perspective
107
aeronautical and commercial activities simultaneously, whereby activities are connected via
passengers as the common intermediate product.
Fig. 11: Airport network technology
Another issue that appears in the airport benchmarking literature is the problem of
comparability. A base assumption within the DEA context that has been questioned is the
homogeneity of the decision-making unit under analysis and the appropriateness of this
assumption with respect to airports (Morrison 2009). The aim of the formulation presented
here is to broach the direct question of airport benchmarking in light of the reasonable level of
heterogeneity found in a multiple airport study, necessary to generate sufficient data points for
purposes of analysis. In order to ensure comparability, we apply a dynamic clustering
approach (Golany and Thore 1997) using integer linear programming which forms reference
sets based on similar mixes of inputs or outputs and intermediate products. Certain inputs may
be beyond managerial control in the short to medium term yet affect airport efficiency (Adler
and Berechman 2001). In general, capital is frequently treated as a non-discretionary variable
over which management has little to no control (Banker and Morey 1986). In this research,
capital has been defined in terms of declared runway and terminal capacity. Declared runway
capacity is agreed upon within a multiple stakeholder setting and result in a number that
accounts for the airport system configuration. For example, some airports consist of a
Labour: Full-time equivalent employees Capital: Terminal capacity Runway capacity Apron capacity (terminal/remote) Security capacity Baggage handling capacity Airport area Gates Public parking spots Materials and supplies: Outsourcing costs Snow removal equipment Fire truck & stations Hangers Maintenance costs
Business & leisure passengers: International-transfer International non-transfer Domestic-transfer Domestic non-transfer
Air Transport Movements Cargo (tons)
Non-traveling customers
Undesirable outputs: Non-weather related delays Aircraft noise Air pollution
Concession revenues: Duty free and retail Catering Car parking Rental Banking Entertainment Departing passenger services Airport cities
Aeronautical revenues: Aircraft landing fees Passenger charges & fees Aircraft parking fees Ground handling fees Cargo fees Centralized infrastructure fees Noise surcharges Security charges
Airport Benchmarking from a Managerial Perspective
108
reasonably large number of runways however for reasons of weather and/or geographical
layout, only a smaller portion may be in use in a given timeframe which declared capacity
takes into account as compared to simply counting the number of runways. On the terminal
side, check-in counters, security and passport control and gates together produce a throughput
level per hour that is otherwise assumed to be linear within a standard DEA framework. We
argue that the capacity of an airport, as a proxy for capital, may be adjusted to a certain extent
in the medium term, hence the model restrictions permit terminal and declared runway
capacity to change up to a pre-determined level. Pure capital investment is not an appropriate
measure even within a specific country because the accounting processes differ, rendering the
information incompatible. Finally, principal component analysis (PCA) combined with DEA
(Adler and Golany 2001; Adler and Yazhemsky 2010) is applied in the input-oriented model
in order to reduce the curse of dimensionality and resulting bias, reducing the set of cost
efficient airports from 53% to 38% in the current application.
The aim of this research is to develop a comprehensive methodology tailored to airport
benchmarking from the managerial perspective. A comparison with basic DEA results
demonstrates that the additional restrictions in the network PCA-DEA dynamic clustering
formulation lead to more reasonable peer comparisons, permitting an analysis of strategies
which could potentially be adopted over short and medium term planning horizons. The
model in this research allows airport managers to include their industry knowledge in the
form of limitations on airport size, operating conditions and restricted variability of capacity
encapsulated in the dynamic clustering approach. For example, the results of the under-
utilized airport in Hanover indicate that in the medium-term the airport could either reduce
operations to two of their three existing runways, instead of closing two runways as obtained
with basic DEA, or alternatively attempt to increase cargo throughput as occurred at their two
medium-term benchmark airports located in Venice and Hamburg. The formulations
developed are suitable for assessing appropriate strategies with respect to aeronautical and
commercial activities not only separately but also in combination, assuming cross-
subsidization is an acceptable policy. According to the combined network DEA dynamic
cluster revenue maximization approach, Lyon airport has achieved a sustainable level of
aeronautical revenues and ought to search for appropriate commercial revenue opportunities
as opposed to the basic DEA results which suggest a further increase in aeronautical revenues
of 40%. The methodology provides an airport manager with the tools for both exploratory
data analysis and inefficiency estimation, removing the need for additional tests of
homogeneity. Furthermore, utilizing an hourly capacity measure as both a terminal and airside
Airport Benchmarking from a Managerial Perspective
109
proxy of physical capital appears to be new in airport benchmarking studies. Compared to the
standard quantity measures such as the number of runways or gates, this proxy provides an
improved managerial measure of the airport infrastructure as a system and allows us to
consider bottlenecks at an airport.
The paper is organized as follows: Section 4.2 presents individual modelling formulations
that have been combined in Section 4.3 in order to produce airport benchmarks based on a
network PCA-DEA dynamic clustering approach. Section 4.4 provides a description of the
public data available for analysis and Section 4.5 compares the results of the combined
formulations to those of basic DEA models and benchmarks a select subset of airports in
order to demonstrate the utility of the approach developed in this research. Finally, Section
4.6 concludes and presents recommendations for further research.
4.2 Methodology
DEA is a non-parametric method of frontier estimation that measures the relative
efficiency of decision-making units utilizing multiple inputs and outputs. DEA was first
published in Charnes et al. (1978) under the assumption of constant returns-to-scale31 and was
extended by Banker et al. (1984) to include variable returns-to-scale. The DEA model
categorizes decision-making units into two groups, those that are deemed efficient and define
the Pareto frontier and those that lie within the envelope and are deemed inefficient, for which
benchmarks are clearly defined.
This section discusses the dynamic clustering mechanism that ensures comparable
benchmarks are chosen from a dataset given exogenous parameter values and the network
DEA model first designed to disaggregate the process of decision-making within a unit.
Subsequently we discuss the combination of principal component analysis and data
envelopment analysis, which reduces efficiency over-estimation bias and a multi-dimensional
scaling approach that produces a graphical representation of the data. Finally, we discuss a
non-parametric statistical procedure that measures efficiency variation across different groups
within the dataset in order to estimate the potential impact of environmental variables on the
relative Pareto efficient frontier.
31 Constant returns-to-scale means that the producers are able to linearly scale the inputs and outputs without increasing or decreasing efficiency.
Airport Benchmarking from a Managerial Perspective
110
4.2.1 Dynamic clustering
Basic DEA benchmarking may lead to inappropriate targets for improvement in a dataset
in which there are substantial differences in size among the decision-making units (DMUs)
under analysis. Sarkis and Talluri (2004) propose second-stage clustering to identify
benchmarks for poor performers after applying DEA to determine the relative efficiencies of
airports. This study applies a dynamic clustering approach first proposed by Golany and
Thore (1997) that restricts the selection of best practice DMUs according to predefined
boundaries within the DEA framework in a single stage process. The boundaries of the cluster
are defined in relative terms, limiting the efficient reference set32 to those DMUs whose input-
output values are within the distance defined by the proportions.
Fig. 12: Benchmark clustering
In Figure 12 we demonstrate the impact of the cluster restrictions for a simplified model
with two outputs and a single input. DMUa is compared to the Pareto frontier (blue line
defined by DMUs 3 to 6) in a standard DEA formulation, with DMUs 4 and 5 acting as
benchmarks. In our proposed approach, each inefficient airport may refer to a set of
benchmarks that do not lie directly on the Pareto frontier rather within the dotted radius. If
DMUa lies far enough away from the Pareto frontier as shown in Figure 12, all potential
benchmarks will lie in the interior of the envelope, resulting in DMUs 1 and 2 acting as
benchmarks for DMUa. The assumptions of this approach lead to the conclusion that DMUa’’,
the hypothetical observation lying on the interior frontier, represents a relevant target which is
32 An efficient reference set, or peer group, is defined by a subset of efficient units "closest" to the unit under evaluation i.e. with similar mixes of inputs and outputs.
y1/x
DMU3
DMU2
DMU1
DMUa
DMU4
DMU5
DMU6
DMUa’
DMUa’’
y2/x
Airport Benchmarking from a Managerial Perspective
111
more accessible than DMUa’ in the short to medium term. Dynamic clustering improves on
the Sarkis and Talluri two-stage procedure since additional information, such as the
importance of each target DMU, can be drawn from the one step procedure.
4.2.2 Network DEA
Network DEA models were first introduced by Färe (1991) and Färe and Grosskopf
(1996, 2000) and subsequently extended by Lewis and Sexton (2004), Emrouznejad and
Thanassoulis (2005), Chen (2009), Kao (2009) and Tone and Tsutsui (2009). Opening the
black box permits an analysis of the optimal production structure of DMUs and their
priorities, to determine both efficient subsystems and overall efficiency. In transportation,
network DEA has been applied by Yu and Lin (2008) in order to simultaneously estimate
passenger and freight technical efficiency, service effectiveness and technical effectiveness
for 20 selected railways.
This research develops a network model that defines a multi-product airport in which
capital, labour, materials and outsourcing produce traffic volume, in the form of aircraft
movements, passenger and cargo. This throughput then generates revenues from aeronautical
charges paid mostly by airlines and from commercial terminal-side services serving
passengers. The overall profits of this system are driven by services provided by outside
parties including airlines and third party contractors as well as the airport processes
themselves. Airport management retain reasonable control over labour, materials and levels of
outsourcing but limited control over capital investments. In addition, management control the
variety and the pricing policies offered on the non-aeronautical side and partially control
aeronautical tariffs, dependent on the regulatory regime of the relevant country. Network
DEA lends itself to a more accurate description of this process than standard performance
analyses.
4.2.3 Principal component analysis integrated with DEA
Dependent on the nature of the dataset, the results of the DEA model may not sufficiently
distinguish between the efficient and inefficient DMUs due to an overestimation bias caused
by the curse of dimensionality (Adler and Yazhemsky 2010). PCA-DEA is one of the
methodologies that has been developed to reduce the number of inefficient DMUs incorrectly
classified as efficient (Adler and Golany 2001, 2002). The original variables are replaced with
a smaller group of principle components (PCs), which explain the variance structure of a
Airport Benchmarking from a Managerial Perspective
112
matrix of data through linear combinations of variables. The principal components are
uncorrelated linear combinations ranked by their variances in descending order and those that
explain little of the variance of the original data may be removed thus reducing the
dimensions in the DEA linear program. In order to use principal components instead of the
original data, the DEA model needs to be transformed to take into account the linear
aggregation.
A rule-of-thumb computed in Adler and Yazhemsky (2010) suggests that at least 76-80%
of the information should be retained in the model in order to minimize the overestimation
bias33. Clearly, if we use less than full information, we will lose some of the explanatory
powers of the data but we will improve the discriminatory power of the model. It should be
noted that as a result of the free sign in principal component analysis and the transformed
constraints in the PCA-DEA model, the targets and efficient peers obtained could reflect a
change in the current mix of input-output levels of the inefficient DMUs, along the lines of
weight constrained DEA.
4.2.4 Visualizing multiple dimensions
Co-Plot, a variant of multi-dimensional scaling, aids both in exploring the raw data and in
visualizing the results of DEA (Adler and Raveh 2008). Co-Plot positions each decision-
making unit in a two-dimensional space in which the location of each observation is
determined by all variables simultaneously according to a correlation analysis. The graphical
display technique plots observations as points and variables as arrows, relative to the same
arbitrary center-of-gravity. Observations are mapped such that similar DMUs are closely
located on the plot, signifying that they belong to a group possessing comparable
characteristics and behavior. A general rule-of-thumb states that the picture is statistically
significant if the coefficient of alienation is less than 0.15 and the average of correlations is at
least 0.7534. We apply Co-Plot to the set of variable ratios (each output divided by each input),
in order to align the technique to the idea of efficiency as defined in DEA, such that Co-Plot
graphically displays the DEA results in two dimensions. In general, the efficient DMUs
33 The rule-of-thumb defines the percentage of retained information required to balance the trade-off between the two incorrect definitions of (in)efficiency, namely efficient decision-making units defined as inefficient (under-estimation) and inefficient DMUs defined as efficient (over-estimation).
34 The coefficient of alienation is a single measure of goodness-of-fit for the configuration of n observations obtained from a smallest space analysis (Guttman 1968). The higher the correlation, the better the common direction and order of the projections of the n points along the arrow. The length of the arrow is proportional to the correlation.
Airport Benchmarking from a Managerial Perspective
113
appear in the outer circle of the plot signifying their relative achievements and we
exogenously determine the color of the DMUs in order to clarify the results of the DEA.
4.2.5 Measuring efficiency variation across groups
In order to determine whether there are distinct efficiency differences between groups of
airports, we apply the program evaluation procedure outlined in Brockett and Golany (1996)
and Sueyoshi and Aoki (2001). Four steps are required to implement the procedure. In the
first step, the complete set of DMUs (j=1,…,n) are split into k sub-groups and the model is
run separately over each of the k groups. Then, for each of the k individual groups, the
inefficient DMUs are moved to their hypothetical efficient level by projecting them onto the
efficient frontier of their relevant group. In the third step, a pooled DEA is run with all n
DMUs based on their adjusted variables. Finally, a Kruskal-Wallis test is applied to determine
if the k groups possess the same distribution of efficiency values within the pooled set. If the
null hypothesis is correct, we expect to see most of the DMUs rated as efficient in step three.
Note that in order to avoid inaccuracy in the nonparametric rank test, the number of
observations in each of the k subgroups should be of similar size. If this is not the case, the
size of the smallest subgroup is calculated and simple random sampling without replacement
is applied to form subgroups of equally small sized samples. In order to test whether the
findings are robust, Banker’s F-test (1993) may be applied in the last stage of the procedure.
4.3 Model formulations
In this section we describe three network PCA-DEA approaches with dynamic clustering
that are then applied in Section 4.5. The application of network DEA to airports is new and to
the best of our knowledge, we are aware of one working paper in the field, Lozano et al.
(2009), in which capital utilization rather than managerial efficiency is analyzed based on
network-DEA. Given the public data available for the study, Figure 13 presents the airport
network technology that we analyze based on a subset of variables described in Figure 11. X
represent inputs, Y outputs and I intermediate products. The number in brackets represents a
node index in the network.
Airport Benchmarking from a Managerial Perspective
114
Fig. 13: Two-stage airport network technology
Model (4.01) assumes that airport management is interested in maximizing revenues,
drawing from aeronautical activities and concessions given airport throughput on the terminal
and airsides, which are in turn limited by the physical infrastructure and associated costs
available to support the system. Drawing on discussions with airport managers and Pels et al.
(2003), we assume constant returns-to-scale with respect to revenues in that a doubling of the
intermediate inputs, namely passengers, air traffic movements (ATM) and cargo, should
increase revenues at an equivalent rate. The network DEA formulation for the radial, output-
oriented, constant returns-to-scale, mixed integer linear program applied in this research is
presented in model (4.01), where superscript a is the index of DMUa, the unit under
investigation; Xa represents the input values of DMUa; Ya and Ia are the output and
intermediate values of DMUa respectively and the subscript of intensities for DMUa, nijλ ,
denotes the link leading from node i to node j in the network presented in Figure 13. θ1 and θ2
represent the relative efficiency scores for the commercial and airside activities respectively,
where a value of 1 indicates efficiency in generating revenues given the airport’s resources
and a value greater than 1 indicates by how much the relevant revenues ought to be increased
in order for DMUa to be deemed relatively efficient. It should be noted that the first four rows
of model (4.01) are not summed over n, in order to restrict envelopment intensities nλ24 and
nλ235 , thus comparing airports that possess input levels that lie within a boundary of 10% to
300% of DMUa inputs and between 20% and 200% of DMUa intermediate outputs. Parameter
values, αl=0.1, αu=3, βl=0.2, βu=2, were chosen such that a sufficiently rich set of airports
exist in the cluster. Sensitivity analyses of our current dataset suggest that smaller bounds
Inputs (1):
Staff costs X1
Other operating costs X2
Runway capacity X3
Intermediate goods (2): Number of passengers:
International I1
Domestic I2
Intermediate goods (3):
Tons of cargo I3
ATM I4
Output (4):
Non-aeronautical revenues Y1
Output (5):
Aeronautical revenues Y2
Airport Benchmarking from a Managerial Perspective
115
result in excessive limitations and pure self-comparisons over time whereas a wider set lead to
unreasonable benchmarks whereby London-Heathrow and Tallinn, representing the largest
and smallest airports in the dataset, are considered directly comparable.
{ } { }0
1010
,,
4321
21
1 , 43
1 ,4321
1 , 21
1 ,4321
2352421
1312
13235122351224
222351
2
2351
11241
1
241
131313
131313
121212
121212
21
≥
∈∈
≤≤≤
≥
=∀≤
≥
=∀≤
==∀≤≤
==∀≤≤
==∀≤≤
==∀≤≤
+
∑
∑
∑
∑
=
=
=
=
nn
nn
nnnnnn
anN
n
n
aj
nN
n
nj
anN
n
n
ak
nN
n
nk
namu
nnm
naml
naju
nnj
najl
naku
nnk
nakl
naju
nnj
najl
λ,θ
,λ, λ, θ θ
binary , λ, λ
λ λ λ λ λ λ
YθλY
,,, j IλI
YθλY
, k IλI
,...,Nn , m λIβλIλI β
,...,Nn ,,, j λXαλXλX α
,...,Nn , k λIβλIλI β
,...,Nn ,,, j λXαλXλXs.t. α
θ θMax
(4.01)
In order to connect Figure 13 and the clustering approach, nλ12 and nλ13 are binary
variables and nλ24 and nλ235 are non-negative continuous variables. If nλ12 =1 then DMUn could
be included in the peer group for DMUa on the non-aeronautical side and if nλ12 = nλ13 =1 then
DMUn could be included in the peer group for DMUa on the aeronautical side. nλ12
consequently connects costs to the number of passengers produced and nλ13 connects costs to
cargo and ATM production such that DMUs of similar size and cost structure represent
potential benchmarks35. nλ24 connects the number of passengers to non-aeronautical revenue
derived and nλ235 connects passengers, cargo and ATM to aeronautical revenue. Since no
trade-off between aeronautical and non-aeronautical activities is introduced in the model,
benchmarks on each side of the airport activity are determined independently.
Alternatively, whilst it may be assumed that a private, unregulated airport pursues profit
maximization, airports that are subject to economic regulation may behave as social welfare
maximizers. Hence, maximizing aeronautical revenues may not be the target of airport
management due to regulatory constraints. Furthermore, even profit maximizers may consider
35 The effect of this approach is depicted in Figure 12 resulting in DMUs 1 and 2 acting as benchmarks for DMUa: 2,1
12λ and 2,113λ are binary variables equal to 1 since they lie within the boundaries of the first four equations
in model (4.01) whereas 5,412λ and 5,4
13λ equal 0.
Airport Benchmarking from a Managerial Perspective
116
lower aeronautical charges as an opportunity to expand non-aeronautical activities and
generate additional revenues by attracting airlines through lower airport charges.
Consequently, the network DEA formulation for the radial, output-oriented, constant returns-
to-scale, mixed integer linear program combining both aeronautical and concession activities
is presented in (4.02).
{ }0
10
4321
1 ,4,3,21
1 ,4321
2351
123
123235
2351
22351
2
112351
1
123123123
123123123
1
≥
∈
≤
=∀≤
≥
≥
==∀≤≤
==∀≤≤
∑
∑
∑
=
=
=
n
n
nn
aj
nN
n
nj
anN
n
n
anN
n
n
naku
nnk
nakl
naju
nnj
najl
λ,θ
, λ θ
binary , λ
λ λ
,,, j IλI
YλY
YθλY
,...,Nn , k λIβλIλI β
,...,Nn ,,, j λXαλXλXs.t. α
θMax
(4.02)
The goal is to maximize non-aeronautical revenue (Y1) given international and domestic
passengers, cargo and ATM. Physical infrastructure (terminal and runway movements), costs
(labour and materials) and intermediate outputs define the reference set for each DMU as in
model (4.01). Aeronautical revenue (Y2) is included in the analysis as a non-discretionary
variable (Banker and Morey 1986). According to this model, benchmarks consist of airports
achieving higher non-aeronautical revenues, given similar levels of aeronautical revenue
whilst comparing airports of similar size and demand levels. In the following we will refer to
(4.01) as the independent model where the clusters were independently defined and the
efficiency of the aeronautical and non-aeronautical side estimated separately. Formulation
(4.02) presents the combined model since a common set of benchmarks are considered but
only non-aeronautical revenues are maximized.
The network PCA-DEA formulation for the radial, input-oriented, variable returns-to-
scale, mixed integer linear program proposed in this research is presented in formulation
(4.03). The cost minimization assumes variable returns-to-scale, since a doubling of output
should not necessarily result in a doubling of staff, materials and outsourcing costs (Gillen
and Lall 1997; Pels et al. 2003). As opposed to the output-oriented model, we have combined
domestic and international passengers into one intermediate variable Ipax in order to reduce the
number of variables. Furthermore, we have applied principal component analysis (PCA) to
reduce the over-estimation bias and improve the level of discrimination in the results. The
first principal component (PCcost) combines staff costs and other operating costs, explaining
Airport Benchmarking from a Managerial Perspective
117
89% of the variance in the original data. PCcap combines terminal and runways capacities,
explaining 85% of the original information. Including all PCs would provide precisely the
same solution as that achieved under the original DEA formulation.
{ }0,,,
10,,,,,
1,1
43
1 ,43 1
1 1
1 1
1231221
352524
3512325123241232412
1123
112
1231
1231
2
2
1
12123
1
cos22cos
2
1
12cos123
1cos
121
1
2
1
1112
1
cos11cos
2
1
11cos12
1cos
353535
352352352
252525
252252252
242424
241241241
21
≥∈
≤≤≤≤
==
=∀≥
≥
=+
=+
≥
=+
=+
==∀≤≤=∀≤≤
=∀≤≤=∀≤≤
=∀≤≤=∀≤≤
+
∑∑
∑
∑
∑∑
∑∑
∑
∑∑
∑∑
==
=
=
==
==
=
==
==
nn
nnn
nnnnnnnn
N
n
nN
n
n
aq
nN
n
nq
aPAX
nN
n
nPAX
acap
icap
i
icap
nN
n
ncap
at
it
i
it
nN
n
nt
aPAX
nN
n
nPAX
acap
icap
i
icap
nN
n
ncap
at
it
i
it
nN
n
nt
namu
nnm
naml
nau
nnnal
naPAXu
nnPAX
naPAXl
nau
nnnal
naPAXu
nnPAX
naPAXl
nau
nnnal
λ,θ
λ λ θ θ binary, λ λ λ
λ λλ λλ λλ λ
λ λ
, q IλI
IλI
δPCSlλPC
PCθSlλPC
IλI
δPCSlλPC
PCθSlλPC
,...,N n , m λIβλIλI β,...,N n λYαλYλY α
,...,N n λIβλIλI β,...,N n λYαλYλY α
,...,N n λIβλIλI β,...,N n λYαλYλYs.t. α
θ θMin
(4.03)
Model (4.03) clusters airports according to revenue and traffic mix, whereby the total
number of passengers is included in the commercial side and all intermediate activities are
included in the aeronautical side. Parameter values were set at αl=0.1, αu=3, βl=0.2, βu=2. nλ24 , nλ25 and nλ35 are binary variables and nλ12 and nλ123 are non-negative continuous variables.
Scost and Scap are slack variables and lcost and lcap are normalized eigenvectors based on costs
and capacities respectively. θ1 and θ2 represent relative efficiency scores on the terminal side
and airside respectively, where a score of 1 means that the airport is relatively cost efficient
and less that one indicates the level of input retraction required to achieve relative efficiency
in comparison to the benchmarks identified. θ1=1 indicates a cost minimization approach with
respect to the non-aeronautical activities of the airport and θ2=1 indicates cost minimization
with respect to all activities of the airport (passengers, cargo, ATM) whereby the source of
revenues draws from both the non-aeronautical and the aeronautical sides. To restrict the
variability of physical infrastructure, we assume that terminal and runway capacities may be
adjusted up to 30% in the medium term (δ=0.7).
Airport Benchmarking from a Managerial Perspective
118
4.4 Dataset
In this section we describe the variables collected, additional environmental variables that
may be required to adapt the model to ensure homogeneity of the production process and the
complete set of observations together with an initial exploratory data analysis. The dataset
consists of 43 European airports located in 13 different countries. We have pooled the data to
an unbalanced set of 294 observations covering the time period from 1998 to 2007 (Table 17
in Appendix 4.A lists the set of airports under study, the specific timeframe for which the data
was available and whether ground handling processes are undertaken in-house or outsourced).
All airports offer domestic and international routes, however airports located in smaller
countries such as the Netherlands generally have very few domestic destinations. The
passenger volume varies considerably from less than a million passengers at Tallinn and
Durham Tees Valley airports up to more than 50 million at London-Heathrow, the largest
European airport in terms of passenger throughput and number three in the world (ACI 2009).
Ten variables were collected in total for purposes of analysis based on publicly available
data. The variables are categorized into three groups; four inputs (X), four intermediate
products (I) and two outputs (Y). Table 13 presents summary information and specifies the
data sources. The operating inputs consist of staff costs and all other non-labour related
operating costs, which include materials and outsourcing. Although a smaller airport than
Heathrow in terms of air traffic movements, Frankfurt’s staff costs are highest due to the level
of vertical integration whereby the airport operates most of the services by itself or through
wholly-owned subsidiaries. As an example, the airport manages the ground handling
operations which represent one of the most labour intensive activities at an airport, a process
traditionally organized by airlines or independent third party providers at Heathrow.
Consequently, Heathrow spends the most on other operating costs, reflecting the high levels
of outsourcing undertaken.
Generally, as a proxy for capital, physical data such as the number of runways, gates,
check-in counters and overall terminal size is collected. However, such data is often
problematic because the number of runways does not include information on the
configuration or the impact of weather and on the number of runways open within a given
timeframe. Furthermore, the terminal area in square metres is somewhat subjective since
some airports report gross terminal area including sections of an airport that are not open to
the public. If the dataset covers more than one country, the monetary measurement of physical
capital also creates difficulties due to different national accounting standards and depreciation
Airport Benchmarking from a Managerial Perspective
119
methods or periods across countries. For example, the airports of the British Airports
Authority (BAA) depreciate their runways over 100 years whereas the airports operated by
the Aéroports de Paris (ADP) depreciate over a period of 10 to 20 years (Graham 2005). In
this research, terminal capacity is defined in terms of passengers handled within an hour, thus
combining the capacities of all terminal facilities including check-in counters, security
controls, baggage delivery and retail area into one common capacity figure. The airside is
defined by the declared runway capacity, specified as the number of departing and arrival
movements specified per hour. Airport stakeholders negotiate this parameter biannually which
is primarily used to avoid congestion at schedule facilitated airports and aid in the allocation
of slots at coordinated airports (IATA 2010). The advantage of using declared capacity is that
the parameters account for bottlenecks across the terminal and runway systems, providing two
individual capacity measures. Amsterdam possesses the highest agreed terminal and runway
capacities in our sample with 26,000 passengers and 110 movements per hour. Due to their
geographical location near the coast, they require a special runway configuration to operate as
a hub airport. The smallest airport with respect to runway capacity is Florence in the Tuscany
region, with a maximum hourly rate of twelve movements. Due to its short, single runway
system (1,688 m), the airport can handle aircrafts up to the size of a Boeing 737 or an Airbus
A319 (Aeroporto di Firenze 2010).
The annual traffic volume is represented by the number of passengers, commercial ATM
and tons of cargo (trucking is excluded). The passengers are divided according to domestic
and international destinations. Unfortunately, we could not collect enough data to separate the
passengers between intercontinental and European flights or account for transfer passengers,
which would be preferable since these groups probably generate different revenue streams.
Non-aviation revenues include revenues from retail activities and restaurants, concessions and
income from rents and utilities. Aviation revenues are generated from (often regulated)
landing and passenger charges, ground handling undertaken in-house and cargo activities. The
largest non-aeronautical revenues were generated at Heathrow, whereas Frankfurt earned the
highest aviation revenues. Commercial revenues equal 67% of total airport revenues on
average in the dataset, clearly supporting the argument that non-aeronautical activities should
not be ignored in a productivity analysis of airports from a managerial perspective,
particularly when considering the possibility of cross-subsidization.
Airport Benchmarking from a Managerial Perspective
120
Tab. 13: Variables in airport efficiency analysis
Variable Description Name Average Maximum Minimum Source
Staff costs Wages and salaries, other staff costs X1 81,704,359 1,080,756,267 5,962,213 Annual Reports
Other operating costs
Costs of materials, outsourcing and other X2 103,364,471 725,987,196 5,010,381 Annual Reports
Declared runway capacity Total movements per hour X3 49 110 12
IATA (2003), Airport and Coordinator Websites
Terminal capacity Total passenger throughput per hour X4 6,768 26,000 450 IATA (2003), Airport
Websites
International passengers Annual passenger volume I1 10,300,571 61,517,733 355,579 IATA (2003), Airport
Websites
Domestic passengers Annual passenger volume I2 2,433,287 9,932,208 48 IATA (2003), Airport
Websites
Cargo Metric tons (trucking excluded) I3 214,076 2,190,461 37 IATA (2003), Airport Websites
Air transport movements Total commercial movements I4 152,133 492,569 16,000 IATA (2003), Airport
Websites
Non-aeronautical revenues
Revenues from concessions own retail and restaurants, rents, utilities and other Y1 117,906,043 1,107,046,057 4,629,813 Annual Reports
Aeronautical revenues
Landing, passenger and aircraft parking charges; revenues from ground handling, cargo revenues and other
Y2 175,507,645 1,739,331,693 7,199,668 Annual Reports
All financial data is deflated to the year 2000 and adjusted by the purchasing power parity
according to the United States dollar in order to ensure comparability across countries. In
addition, the data has been normalized by the standard deviation to limit the influence of
outliers in the dataset.
4.5 Empirical results
In the following section we identify the impact of vertical integration and subsequently
include the information in the dynamic clusters. In section 4.5.2 we compare and contrast the
results of a basic DEA model with the network PCA-DEA formulation. Section 4.5.3
discusses the benchmarking results for a subset of airports, specifically Vienna as an example
of both an output- and input-oriented efficient airport, Hanover as an example of an input-
oriented inefficient case and Lyon as an example of the differences between the independent
output model (equations 4.01) which assesses the efficiency estimates of both revenue
generating activities separately and the combined output model (equations 4.02) in which only
commercial revenues are maximized.
Airport Benchmarking from a Managerial Perspective
121
4.5.1 Efficiency variation across groups
When estimating the relative efficiencies, it would appear that airports offering in-house
ground handling services operate on a different production frontier to airports that outsource
this activity. This is not immediately obvious since airports providing ground handling
services in-house have higher labour costs, outsourced have higher ‘other’ costs and both
have higher revenues than airports who permit third parties to provide the service hence no
costs appear on the books and only minor concessional fees from the suppliers on the output
side because the contracts themselves do not appear on the airport’s accounting books. To
evaluate the potential for different productivity levels, the non-parametric program evaluation
procedure was applied to basic DEA which contains the last stage of formulations (4.01) and
(4.03), combining both sources of revenues into one efficiency estimate. Based on Figure 13,
the output orientation assumes constant returns-to-scale and includes nodes {2345}, while the
input orientation includes the inputs and outputs from nodes {123} and assumes variable
returns-to-scale. In our sample, 21 airports offer ground handling and 22 outsourced or never
offered this service which translates into 156 DMUs in the ground handling group and 138
DMUs otherwise (see Table 17 in Appendix 4.A). The results are clear and significant that
airport operators providing ground handling appear to be revenue maximisers but were highly
inefficient in cost minimization relative to airports from the non-ground handling group.
Graph (a) in Figure 14 shows the DEA efficiency scores on the vertical axis for the two
groups from a revenue maximization perspective and (b) shows the DEA efficiency scores
from the cost minimization perspective across the two groups. Airports with ground handling
activities perform on average 10% better in maximizing their outputs as their aeronautical
revenues per passenger are naturally higher whereas in the input-oriented model, airports that
do not provide ground handling achieve on average 10% higher efficiencies since no costs are
associated with this service.
Airport Benchmarking from a Managerial Perspective
122
Fig. 14: Kruskal-Wallis ANOVA for outsourcing36
(a) Output-orientation
Source Sum of squares
Degrees of freedom
Mean Squares
Chi-sq
Prob > Chi-sq
Groups 243,606 1 243,606 33.7 6.4e-009
Error 1,873,769 292 6,417
Total 2,117,376 293
(b) Input-orientation
Source Sum of squares
Degrees of freedom
Mean Squares
Chi-sq
Prob > Chi-sq
Groups 357,302 1 357,302 49.4 2.0e-012 Error 1,760,162 292 6,028
Total 2,117,465 293
In order to test the robustness of our findings, the Banker F-test (1993) is also applied both in
the third stage of the program evaluation procedure and on basic DEA scores when two sub-
36 The vertical axis of the boxplot represents the efficiency scores computed in the third step of the program evaluation procedure (a score of one implies relative efficiency).
Airport Benchmarking from a Managerial Perspective
123
groups of DMUs face the same frontier as suggested in Banker (1993), assuming
exponential37 and half-normal38 inefficiency distributions (Table 14).
Tab. 14: Banker F-test for outsourcing
(a) Output-orientation
Inefficiency distribution Test applied on Test statistic Prob>F Entire dataset 2.5726 5.62797E-16 Exponential Program evaluation procedure 3.9766 4.69994E-31 Entire dataset 5.5646 9.70871E-24 Half-normal Program evaluation procedure 9.1030 3.84569E-36
(b) Input-orientation
Inefficiency distribution Test applied on Test statistic Prob>F Entire dataset 2.8054 1.33497E-18
Exponential Program evaluation procedure 9.1584 1.27351E-70 Entire dataset 7.8721 2.68437E-32 Half-normal Program evaluation procedure 22.7297 1.3301E-123
The results also proved to be consistent for a basic DEA model in which air traffic
movements, passengers, cargo and commercial income including ground-handling revenues
were selected as output in order to adjust for the effect of outsourcing. Staff and other
operating costs and runway and terminal capacities were defined as inputs. Having now
considered both costs and revenues in the efficiency estimation, the radial variable returns-to-
scale, input-oriented model still indicated significant efficiency differences across both groups
based on the results of a Kruskal-Wallis test. In summary, both the non-parametric Kruskal-
Wallis and parametric F-test reach the same significant result supporting a rejection of the
null hypothesis (Figure 14 and Table 14). After liberalization in 1996, airports that provided
ground-handling were required to permit competitors’ access. Munich and Frankfurt have
claimed substantial losses in this segment on a regular basis (Dietz 2009; Hutter 2009)39.
However, the strong labour unions in Germany have prevented airport management from
either cutting wages or outsourcing this service to third-party providers without guarantees
37 Test statistic= 1
2
1 11
2 21
/
/
Njj
Njj
u N
u N=
=
∑∑
%
%
and is distributed as F with (2N1, 2N2) degrees of freedom, where 1j ju θ= −% in the
output oriented and 1 1jj
uθ
= −%in the input oriented model and belong to the range [0, ∞).
38 Test Statistic= 1
2
21 11
22 21
/
/
Njj
Njj
u N
u N=
=
∑∑
%
%
and is distributed as F with (N1, N2) degrees of freedom.
39 Most German airports are fully or at least major publicly owned and if ground handling is operated in-house by the parent company, the airport pays salaries based on public tariffs, which are on average 20% higher compared to private ground handling companies. Some German airports, such as the minor-private airport Hamburg, outsourced the ground handling segment to a 100% subsidiary in order to set flexible tariffs however this was not deemed acceptable by the public shareholders of Munich for example.
Airport Benchmarking from a Managerial Perspective
124
that workers would continue under the same conditions. Thus, at least in the short-term, the
degree of outsourcing can be regarded as a political factor that is beyond managerial control
and ground handling is included in the network DEA formulation as an environmental
variable which will further limit the potential benchmark set via clustering.
4.5.2 Comparison of basic and network DEA
In contrast to our formulations (4.01 to 4.03), basic DEA does not restrict potential
benchmarks nor does it permit a limited deviation in one or more variables. In order to assist
the comparison between basic and network DEA results, we have exogenously divided the
dataset according to in-house or outsourced ground handling provision and applied DEA
individually to each category. For the output orientation, the technology of Figure 13 reduces
to nodes {235} and {24} with respect to aeronautical and commercial activities respectively,
while the input orientation case collapses to nodes {12} when assessing the non-aeronautical
side and {123} with respect to both activities.
The basic DEA results generate consistently efficient airports that belong either to the set
of smallest airports e.g. Bremen, Florence and Ljubljana, which provide ground handling in-
house and Malta, Durham Tees Valley and Leeds/Bradford which outsource, or the largest
airports in the sample such as Frankfurt. In neither output-oriented formulations do airports
achieve 100% efficiency over the entire review period, although Salzburg, Ljubljana and
Malta appear consistently close to the frontier. A notable exception is Cologne-Bonn, which
remains cost efficient with respect to both activities but operates very inefficiently (between
44% and 77% over time) with respect to the commercial side. Cologne-Bonn is the European
hub for the parcel service provider UPS, which rents office space and warehouses from the
airport, suggesting a behaviour different to others in the sample (Cologne-Bonn Airport
2010).
Under basic DEA, all airports are compared against a single Pareto frontier and Salzburg
represents an important benchmark for Vienna, Dusseldorf, Frankfurt, Hamburg and Munich.
However, it is doubtful that the management of a primary or secondary hub airport would
adopt the strategies of an airport that handles less than 2 million passengers per year with very
low cargo throughput too. Durham Tees Valley, a small airport in East England with less than
700,000 passengers per year was defined as a benchmark for Lyon, Geneva, Oslo and the
secondary hub airport in Zurich, which would not occur in the formulations we present due to
the dynamic clustering approach.
Airport Benchmarking from a Managerial Perspective
125
Airports in very small clusters are unique in character and in the extreme case tend to
form their own reference set. In the current dataset, these mostly included the smaller and less
congested airports such as Tallinn, Leeds-Bradford and Durham Tees Valley. These airports
can be identified as outliers according to the Andersen and Petersen (1993) super efficiency
procedure. Such observations frequently influence the basic DEA Pareto frontier, for example
Durham Tees Valley appears within the reference sets of Oslo and Zurich airports. In
comparison, the results of the cost minimization formulation (4.03) categorizes Copenhagen
and London Stansted as peer airports for Oslo and Zurich hence the modelling approach
indicates benchmarks that are more homogeneous in character. Another unique example
includes Dortmund, which acts a self benchmark in the cost minimization approach from
2003 to 2007, namely after their capacity expansion and severe reduction in cargo operations.
Dortmund is the only airport that exhibits operational losses over the entire timeframe. The
airport is partly owned by the local electricity distributor and losses are covered by their major
shareholder (Dortmund Airport 2007). Dortmund is located in the Ruhr area (Ruhrgebiet)
with a population of more than five million, representing the largest agglomeration in
Germany. Airport competition includes Dusseldorf, Cologne-Bonn and Paderborn which are
located in their catchment area (defined as 100 km around the airport) and intermodal
competition includes high speed rail and the motorway, especially on domestic routes and
traffic originating in Benelux. Hence despite their high capital investment, it may be
necessary for the airport to further decrease their aviation charges in order to attract airlines
and new destinations thereby generating additional commercial revenues.
The network DEA formulations provide the user with an exploratory data analysis that
does not exist in the basic DEA results. The results of formulations (4.01) and (4.02)
demonstrate that the average cluster size for each inefficient airport was reasonably small
because the capacity of airports varies considerably across the sample and some airports
suffer low utilization rates whereas other are highly congested. The operating costs at highly
congested airports were large mostly due to employee costs hence airports with similar
capacities did not necessarily belong to the same cluster. In general, large clusters indicate
that various airports in the sample possess similar characteristics which in our dataset
included Dusseldorf, Hamburg, Strasbourg, Venice and London-Gatwick.
Airport Benchmarking from a Managerial Perspective
126
4.5.3 Benchmarking airports
In this sub-section, we describe the type of results and analysis that are achievable by
collecting data and applying the formulations described in Section 4.3. We focus on relatively
efficient Vienna, relatively inefficient Hanover and finally Lyon, to describe the potential
balance between the two revenue streams. Vienna is an example of an airport that has
gradually improved in both input and output efficiency over time, achieving Pareto relative
efficiency by 2007. Vienna appears in the reference set of Cologne-Bonn and Dusseldorf in
the input-oriented case. Between 1998 and 2007 Vienna’s costs and revenues increased on
average by similar proportions (99% and 94% respectively)40, while traffic volume grew by
76% for total passengers, 54% for ATM and 37% for cargo. The input-oriented case in Figure
15 shows that from 1998 to 2003, Vienna lies close to the arrows that display the ratio of
intermediate outputs to costs. In 2004, both staff costs and other operating costs substantially
increased partly due to the introduction of a 100% hold baggage screening policy and the
founding of a subsidiary for infrastructure maintenance (Vienna International Airport 2004).
After 2004 greater emphasis has been allocated to the issue of runway utilization, viewed in
Figure 15 by the proximity of the later years to the capital asset related ratios. Vienna airport
moves in a positive direction towards an improved utilization of the runways which increased
from 48% to 66% between 2000 and 2007. Hence, despite substantial cost increases, the
airport still managed to increase its relative efficiency as their costs per ATM decreased over
time. On the aeronautical output side, the airport management changed their regulated tariff
structure by increasing passenger charges from an average price of 3.90€ in 1999 to 5.90€41 in
2007 and decreasing overall landing charges by 3%, whilst reducing them for larger aircraft
by up to 20%. In summation, Vienna airport increased passenger charges out of the total
aviation revenues collected from 33% in 2001 to 46% in 2007 and decreased the share of
landing fees from 44% to 28% over the same period (Vienna International Airport 2001,
2007). These policies appear to have aided Vienna to achieve revenue productivity.
40 Staff costs increased by 111% and other operating costs by 88%. Non-aeronautical revenues increased by 78% and aviation revenues by 109%.
41 These values are an average passenger price and were computed by dividing total passenger revenue charges by passenger throughput as obtained from the annual reports. They could therefore deviate from the passenger charges specified in the charges manual. Both prices were given in nominal values.
Airport Benchmarking from a Managerial Perspective
127
Fig. 15: Co-Plot graphic display of Vienna’s input-oriented strategy42
Hanover airport is the ninth largest airport in Germany handling 5.6 million passengers in
2008 (ACI 2009). It is partly owned by Fraport AG which owns and operates Germany’s
gateway hub in Frankfurt. Whilst Hanover’s non-aeronautical revenues from rents and
utilities increased over the decade analyzed due to the development of a large airport city, the
relative cost efficiency score θ1 consistently dropped from 72% in 1998 to 60% in 200643.
Over the same time period, passenger volume increased by 17%, ATM by 8% and cargo
dropped by 37% however staff costs and other operating costs increased by 80% and 64%
respectively, as shown in Table 15.
42 Coefficient of alienation is 0.06 and average of correlations is 0.89. 43 If δ=1 is assumed, terminal and runway capacities may decrease to a lower limit of zero. Hanover’s cluster of 100 DMUs is stable over time, a common set of benchmarks exists between 1998 and 2006 (Florence, Hamburg and Venice) and a comparison of θ1 over time is possible. Over the same period θ2 is close to 1 due to overestimation bias caused by the relatively limited cluster size of 30-40 DMUs.
Airport Benchmarking from a Managerial Perspective
128
Tab. 15: Benchmarking Hanover airport
Airport Declared runway capacity
Terminal capacity
Staff costs (US$)
Other operating
costs (US$)
Domestic passengers
International passengers
Air transport movements
Cargo (tons)
DMUs under review
HAJ1998 50 4,000 39,137,748 41,838,277 1,014,723 3,814,405 70,815 10,954
HAJ1999 50 4,000 44,100,404 40,270,806 1,080,384 4,017,528 76,914 7,724
HAJ2000 60 4,000 49,858,032 35,123,344 1,246,083 4,284,201 83,687 9,027
HAJ2001 60 4,000 49,344,453 39,876,433 1,067,834 4,089,724 75,368 6,712
HAJ2002 60 4,000 48,501,264 37,857,069 1,018,412 3,733,509 73,278 6,058
HAJ2003 60 4,000 51,602,584 46,057,783 1,010,975 4,033,895 74,960 6,338
HAJ2004 60 4,000 54,282,812 48,081,380 1,060,005 4,189,164 74,251 6,091
HAJ2005 60 4,000 61,893,620 58,038,723 1,137,940 4,499,445 76,585 6,551
HAJ2006 60 4,000 66,510,634 58,753,898 1,222,533 4,476,766 76,255 5,954
HAJ2007 60 4,000 70,453,727 68,607,888 1,215,036 4,429,546 76,263 6,912
Changing benchmarks over time according to formulation 4.03 and dual values (λ12) for δ=0.7 DMUs under review
Florence 2000
Hamburg 1998/9
Venice 2004/5
Genoa 2000
Nuremberg 2001/2/3/7
Vienna 1999
HAJ1998 0.46 0.29 0.25 HAJ1999 0.43 0.34 0.23 HAJ2000 0.28 0.28 0.43 HAJ2001 0.15 0.29 0.33 0.23 HAJ2002 0.28 0.17 0.37 0.18 HAJ2003 0.13 0.22 0.27 0.38 HAJ2004 0.35 0.21 0.35 0.09 HAJ2005 0.37 0.32 0.03 0.28 HAJ2006 0.19 0.14 0.55 0.13 HAJ2007 0.18 0.57 0.25
Figure 16 displays the gradual decline in productivity (see the red arrow) and the change
in benchmarks over time from Venice to Nuremberg (see white dots), the latter representing a
relatively more expensive airport to operate44. From the technical perspective, Hanover shows
potential to expand airport activities due to a declared runway capacity of 60 movements per
hour. Capacity utilization at Hanover remained at a stable 23%, whereas Florence and Venice
achieve 40% utilization and Hamburg slightly more than 50%. Nuremberg, Hanover’s
benchmark, achieves a capacity utilization of less than 40% which is still higher than that of
Hanover. Bremen and Dresden are the long term benchmarks according to basic DEA, which
appear as black dots on the left edge in Figure 1645. When comparing the results for Hanover
in 2007 with respect to network PCA-DEA (θ1=0.7) and basic DEA (θ1=0.49) it becomes
clear that the long term goal for Hanover, ceteris paribus, would be to close two of the three
runways. The medium-term network DEA results suggest that it would be sufficient to close
the equivalent of a single runway.
44 If δ=0.7 is assumed, terminal and runway capacities may be adjusted up to 30% in the medium term. As a result θ1=0.7 since 2001, although the dynamic clustering shows the change in productivity over time through changes in the set of benchmarks (see Table 15 and Figure 16). 45 According to basic DEA, θ1 decreased slightly from 54% in 1998 to 50% in 2006 and the benchmarks include Bremen, Dresden, Hamburg and Florence over the entire period.
Airport Benchmarking from a Managerial Perspective
129
Fig. 16: Co-plot of input minimization results with emphasis on Hanover46
Hanover’s management may find it rather difficult to improve capacity utilization due to
the airport’s highly competitive location. The airport faces direct competition from Hamburg
and Bremen that are in close proximity, as well as Dortmund, Paderborn and Münster-
Osnabrück, which primarily serve charter and low cost carriers and are less than two hours
drive by car, as shown in Fig. 8. Potential competition includes regional airports located in
Braunschweig-Wolfsburg, Kassel-Calden and Magdeburg-Cochstedt, none of which currently
operate commercially although plans exist to offer commercial flights from Kassel-Calden in
2012 (Flughafen Kassel-Calden). Additional intermodal competition includes the ICE high
speed rail alternative and a highly connected motorway network. In conclusion, Hanover
faces direct, potential and intermodal competition hence the airport needs to cut costs by as
much as 40%, further develop non-airport related activities and attempt to attract cargo
throughput. The latter strategy may increase runway utilization and seems reasonable given
the high level of connectivity of the city and the lack of night flights restrictions due to their
location.
46 Coefficient of alienation is 0.137 and average of correlations is 0.829.
Airport Benchmarking from a Managerial Perspective
130
Fig. 17: Catchment area of Hanover airport (2 hour drive)47
Our final analysis presents the benchmarking results for Lyon Saint-Exupéry which is the
fourth largest airport in France, with a passenger throughput of 7.9 million in 2008 (ACI
2009). Like the majority of French regional airports, Lyon is fully publicly owned and
operated by the regional Chamber of Commerce. The airport became a major regional hub
airport for the national carrier Air France at the end of the nineties and today Easyjet is their
second largest customer (Lyon Aéroport 2008). In the independent revenue maximizing
formulation (4.01), Lyon improved in aeronautical efficiency (θ2) from a score of 2.09 to 1.02
between 1998 and 200548, benefiting from a change in the tariff structure similar to that of
Vienna airport. The important peer airports include privatized BAA Glasgow and Basel-
Mulhouse, both of which focus on low cost carrier traffic. Glasgow serves Easyjet and
Scottish Loganair in competition with Ryanair at Prestwick and Basel-Mulhouse serves
Easyjet, which achieved a market share of almost 50% in 2007 (Flughafen Basel-Mulhouse).
With respect to non-aeronautical activities, Lyon increased its efficiency (θ1) from 1.68 in
1998 to 1.41 in 2005, in part due to the large increase in car parking revenues, rents and
utilities which contributed to a 67% increase in overall commercial revenues (see Table 16).
Benchmarks on the non-aeronautical side include Basel-Mulhouse and Marseille, with the
former generating more than 50% of their revenue from commercial sources, the majority of
which are derived from retail sales, rents and utilities.
47 Source: adapted from ADV (2010). 48 Lyon’s benchmark clusters are stable over time according to formulation 4.01.
Airport Benchmarking from a Managerial Perspective
131
Tab. 16: Output benchmarks for Lyon airport
Airport Domestic passengers
International passengers Movements Cargo Aeronautical
revenues (US$) Non-aeronautical revenues (US$)
DMUs under review
LYS1998 2.478.508 2.742.712 106.170 39.749 25.552.145 33.325.320
LYS1999 2.565.033 2.935.515 116.894 39.050 29.032.107 37.403.371
LYS2000 2.715.196 3.311.666 129.373 40.126 35.476.207 40.519.260
LYS2001 2.714.678 3.393.929 132.903 38.902 41.348.698 44.483.257
LYS2002 2.523.982 3.254.242 120.529 35.349 48.933.813 44.982.158
LYS2003 2.571.177 3.368.718 118.489 35.494 58.338.026 47.915.432
LYS2004 2.633.962 3.594.650 123.958 34.874 61.171.391 50.077.362
LYS2005 2.682.123 3.879.242 128.868 38.725 68.845.652 55.678.876
Output benchmarks for Lyon airport in 2005
Benchmark Intensity (λ) Domestic passengers
International passengers Movements Cargo Aeronautical
revenues (US$)
Non-aeronautical
revenues (US$)
Short-term benchmark for non-aeronautical activities (model 4.02, θ1=1.15)
MLH2004 0.98 651.102 1.893.772 57.915 34.227 30.882.599 38.867.805
GLA2000 0.29 3.568.259 3.453.741 92.000 10.000 69.790.936 38.013.125
GLA2005 0.16 4.604.022 4.237.878 97.610 9.461 76.125.667 63.356.734
NCE2006 0.06 4.332.382 5.615.653 164.617 13.940 100.059.952 83.755.849
Medium-term benchmark for non-aeronautical activities (model 4.01, θ1=1.4)
MLH2004 1.00 651.102 1.893.772 57.915 34.227 30.882.599 38.867.805
MLH2002 0.61 792.765 2.264.199 88.000 31.285 34.865.786 45.079.574
MRS1998 0.39 3.943.382 1.568.411 87.030 55.993 19.795.312 31.681.597
Medium-term benchmark for aeronautical activities (model 4.01, θ2=1.03)
MLH2004 0.96 651.102 1.893.772 57.915 34.227 30.882.599 38.867.805
GLA2000 0.52 3.568.259 3.453.741 92.000 10.000 69.790.936 38.013.125
NCE2006 0.05 4.332.382 5.615.653 164.617 13.940 100.059.952 83.755.849
Long-term benchmark for both activities (basic DEA, θ=1.4)
LCY2002 0.4 417.551 1.187.449 53.000 1.000 38.509.245 12.353.716
MLH2002 0.35 792.765 2.264.199 88.000 31.285 34.865.786 45.079.574
OSL2007 0.19 9.477.511 9.566.489 223.000 97.000 177.975.321 245.588.135
ATH2007 0.08 5.953.814 10.571.571 205.295 119.000 453.224.152 137.319.560
Given that aeronautical revenue maximization is not necessarily an optimal policy,
irrespective of ownership form, the combined formulation (4.02) defines aeronautical revenue
as a non-discretionary output and maximizes commercial revenue alone. The results of this
model suggest that Lyon’s short-term, commercial revenue target should be $63 million, an
increase of 15%, given current aeronautical revenues. The medium-term target (formulation
4.01) suggests an increase of 40% in non-aeronautical revenues to become efficient and the
longer term, standard DEA target requires the same increase of 40% both on the commercial
and aeronautical side respectively (Figure 18). In the combined model, Basel-Mulhouse
appears as an important benchmark and Glasgow acts as a benchmark of increasing intensity
over the years. As also shown in Figure 19, Lyon airport is moving in the direction of Basel-
Mulhouse and Glasgow hence is increasing in efficiency over time. Marseille no longer
Airport Benchmarking from a Managerial Perspective
132
appears as a benchmark in the results of the combined model since the airport generates
substantially lower aeronautical revenues in comparison.
Fig. 18: Current and target output values for Lyon
Non-aeronautical revenues
20.000.000
30.000.000
40.000.000
50.000.000
60.000.000
70.000.000
80.000.000
90.000.000
1998 1999 2000 2001 2002 2003 2004 2005
current value independent model (4.01) combined model (4.02)
Aeronautical revenues
20.000.000
30.000.000
40.000.000
50.000.000
60.000.000
70.000.000
80.000.000
90.000.000
1998 1999 2000 2001 2002 2003 2004 2005
current value independent model (4.01) combined model (4.02)
In summary, were Lyon to adopt the strategies of the short-term benchmarks, aeronautical
revenues were sufficiently high in 2003 and with respect to medium-term benchmarks, the
aeronautical charges were sufficiently high in 2005 (Figure 18). However, Lyon could still
optimize commercial revenues in order to increase managerial productivity. The targets
obtained from the basic DEA results would be very challenging in the short- or medium-term
and should therefore be considered only as a long-term target if at all. As shown in the graphs
of Figure 18, several paths to the Pareto frontier can be defined for the airport in which both
revenues can be expanded equi-proportionally or with greater emphasis on the non-
aeronautical revenues such that the airport remains profit maximizing.
Airport Benchmarking from a Managerial Perspective
133
Fig. 19: Co-plot of output maximization results with emphasis on Lyon49
4.6 Conclusion
DEA has been ubiquitous in the study of airport productivity however the basic DEA
models treat the airport technology as a black box which reduces the usefulness of the model
for purposes of benchmarking. The focus of this paper is to model the airport production
process from a managerial perspective in order to provide a set of models that would aid
benchmarking by applying a network DEA model. Usually network-DEA is applied to
determine the efficiency of sub processes and overall efficiency whereas in our research,
network DEA helps decision-makers to describe the production process, demonstrating the
sequential effects separating final and intermediate outputs including those under partial
managerial control and those that are known to be non-discretionary. Consequently, the
approach connects aeronautical and commercial activities via intermediate products.
To improve the set of peer airports chosen, a dynamic clustering mechanism limits DEA’s
dual variables (benchmark intensities), ensuring appropriate comparability within the dataset.
The dynamic clustering approach proposed by Golany and Thore (1997) restricts the selection
of best practice DMUs according to predefined boundaries within the basic DEA framework.
We extended this method by using integer linear programming which forms reference sets
based on similar mixes of inputs or outputs and intermediate products. As a result each DMU
49 Coefficient of alienation is 0.107 and average of correlations is 0.815
Airport Benchmarking from a Managerial Perspective
134
optimizes only the last stage of the network, taken into account the information from previous
stages. In addition, PCA-DEA is applied to reduce the number of variables when clusters are
too small to avoid the curse of dimensionality. By identifying individual reference sets using
dynamic clustering we provide individual benchmarks for inefficient DMUs, permitting
identification of strategy changes over time and uniqueness with respect to economic
regulation and airport infrastructure. The formulation was further adapted to ensure partial
flexibility with respect to an expensive and complicated infrastructure system. Finally, the
provision of ground handling was shown to severely affect efficiency estimates leading to a
separation in the comparison of those airports that undertake the process in-house compared
to those that outsource.
Data proved to be the most difficult issue for this application. After defining salient
variables (as in Figure 11), we were then forced to reduce the model drastically in the light of
data availability issues (as in Figure 13). It would be extremely helpful were various
government organizations to publicly publish the data that they already collect. However, the
results have shown that compared with the basic DEA approach, network DEA formulations
provide more appropriate benchmarks which may enable airport managers to improve
performance in the short and medium-term. In the case of Hanover, we show that in the short
or medium term it is sufficient to close one of the three existing runways or expand their
cargo operations to increase utilization, whereas basic DEA benchmarks require the airport to
close the equivalent of two runways. In the case of Lyon we demonstrate that in the short-
term the airport earns a sufficient level of aeronautical revenues and simply needs to focus on
improvements on the commercial side. In comparison, the results of basic DEA require Lyon
to increase aeronautical revenues by 40% in order to operate efficiently.
To be in a position to undertake benchmarking exercises requires the collection and
publication of airport related data openly at the federal level since such information would be
of public interest. Furthermore, an airport also produces undesirable outputs such as delays.
Besides the capacity utilization which has been considered in our research, delay substantially
affects airport and airline efficiency and should clearly be included in a benchmarking study.
For improved managerial benchmarking, disaggregated data with regard to non-aeronautical
activities would help to identify successful strategies on the commercial side. Other factors
that are beyond managerial control include the competitive environment, ownership structure
and economic regulation. These aspects influence managerial behaviour and accounting for
them may further improve comparability and permit the relevant authorities to analyze the
impact of cost or incentive based regulation on managerial efficiency.
Airport Benchmarking from a Managerial Perspective
135
4.A Appendix
Tab. 17: Airport dataset50
Code Airport Country Time Period Ground Handling
ABZ Aberdeen UK 1999 not provided AMS Amsterdam Netherlands 1998-2007 not provided ATH Athens Greece 2005-2007 not provided BHX Birmingham UK 2005 not provided BLQ Bologna Italy 2000-2005 provided BRE Bremen Germany 1998-2007 provided CGN Cologne-Bonn Germany 1998-2007 provided CPH Copenhagen Denmark 1998-2007 not provided DRS Dresden Germany 1998-2004 provided DTM Dortmund Germany 2001-2007 provided DUS Dusseldorf Germany 1998-2007 provided FLR Florence Italy 2000-2005 provided FRA Frankfurt Germany 2002-2007 provided GLA Glasgow UK 1999-2006 not provided GOA Genoa Italy 2000-2005 provided GVA Geneva Switzerland 1998-2007 not provided HAJ Hanover Germany 1998-2007 provided HAM Hamburg Germany 1998-2007 provided LBA Leeds-Bradford UK 1998-2006 not provided LCY London-City UK 2002 not provided LEJ Leipzig Germany 1998-2002 provided LGW London-Gatwick UK 1998-2002 not provided LHR London-Heathrow UK 1998-2006 not provided LJU Ljubljana Slovenia 2004-2007 provided LTN London-Luton UK 1998-2005 not provided LYS Lyon France 1998-2005 not provided MAN Manchester UK 1998-2006 not provided MLA Malta Malta 2005-2006 not provided MLH Basel-Mulhouse France 1998-2007 not provided MME Durham Tees Valley UK 2002 not provided MRS Marseille France 1998-2006 not provided MUC Munich Germany 1998-2007 provided NCE Nice France 1998-2006 not provided NUE Nuremberg Germany 1998-2007 provided OSL Oslo Norway 1999-2007 not provided RIX Riga Latvia 2004-2006 provided STN London-Stansted UK 1998-2006 not provided STR Stuttgart Germany 1998-2007 provided SZG Salzburg Austria 2004-2007 provided TLL Tallinn Estonia 2002-2007 provided VCE Venice Italy 2000-2005 provided VIE Vienna Austria 1998-2007 provided ZRH Zurich Switzerland 1998-2007 not provided
50 Source: adapted from SH&E (2002) and Airport Websites
References
136
5 REFERENCES
Abbott, M., and S. Wu (2002): ‘Total factor productivity and efficiency of Australian
airports’, Australian Economic Review, 35, 244–260.
ACCC (2003): ‚Quality of service: price-monitored airports 2002-03’, ACCC, Melbourne.
ACCC (2010): ‘Airport monitoring report 2008-09: Price, financial performance and quality
of service monitoring’, ACCC, Canberra.
ACI (2009): World Airport Traffic Report 2008, ACI World, Geneva.
Adler, N., and J. Berechman (2001): ‘Measuring airport quality from the airlines’ viewpoint:
an application of data envelopment analysis’, Transport Policy, 8(3), 171–181.
Adler, N., and B. Golany (2001): ‘Evaluation of deregulated airline networks using data
envelopment analysis combined with principal component analysis with an application to
Western Europe’, European Journal of Operational Research, 132(2), 260–273.
Adler, N., and B. Golany (2002): ‘Including principal component weights to improve
discrimination in data envelopment analysis’, The Journal of the Operational Research
Society, 53(9), 985–991.
Adler, N., Liebert, V., and E. Yazhemsky (2010): ‘Benchmarking airports from a managerial
perspective’, manuscript submitted for publication.
Adler, N., Oum, T.H., and C. Yu (2009): ‘A response to ‘Understanding the complexities and
challenges of airport performance benchmarking’’, Journal of Airport Management, 3(2),
159–163.
Adler, N., and A. Raveh (2008): ‘Presenting DEA graphically’, Omega, 36(5), 715–729.
References
137
Adler, N., and E. Yazhemsky (2010): ‘Improving discrimination in data envelopment
analysis: PCA-DEA or variable reduction’, European Journal of Operational Research,
202(1), 273–284.
ADV (2010): ‘Mitglieder/ Flughafenkarte’, http://www.adv.aero/mitglieder.html, 19.03.2010.
Aeroporto di Firenze (2010): ‘Technical Data’, http://www.aeroporto.firenze.it, 17.03.2010.
Aigner, D.J., Lovell, C.A.K., and P. Schmidt (1977): ‘Formulation and estimation of
stochastic frontier production function models’, Journal of Econometrics, 6, 21-37.
Andersen, P., and N.C. Petersen (1993): ‘A procedure for ranking efficient units in data
envelopment analysis’, Management Science, 39(10), 1261-1264.
Assaf, A. (2008): ‘Accounting for size in efficiency comparisons of airports’, Journal of Air
Transport Management, 15(5), 256-258.
Assaf, A. (2010a): ‘Bootstrapped scale efficiency measures of UK airports’, Journal of Air
Transport Management, 16(1), 42-44.
Assaf, A. (2010b): ‘The cost efficiency of Australian airports post privatisation: A Bayesian
methodology’, Tourism Management, 31(2), 267-273.
Averch, H., and L.L. Johnson (1962): ‘Behavior of the firm under regulatory constraint’, The
American Economic Review, 52(5), 1052-1069.
Banker, R.D. (1993): ‘Maximum likelihood, consistency and data envelopment analysis: a
statistical foundation’, Management Science, 39(10), 1265–1273.
Banker, R.D., Charnes, A., and W.W. Cooper (1984): ‘Some models for estimating technical
and scale inefficiencies in data envelopment analysis’, Management Science, 30(9), 1078–
1092.
Banker, R.D., and R.C. Morey (1986): ‘Efficiency analysis for exogenously fixed inputs and
outputs’, Operations Research, 34(4), 513-521.
Banker, R.D., and R. Natarajan (2008): ‘Evaluating contextual variables affecting
productivity using data envelopment analysis’, Operations Research, 56(1), 48-58.
Barbot, C. (2010): ‘Competition in Complementary Goods: Airport Handling Markets and
Council Directive 96/67/EC’, CEF.UP Working Paper Series.
Barros, C.P. (2008a): ‘Airports in Argentina: Technical efficiency in the context of an
economic crisis’, Journal of Air Transport Management, 14(6), 315–319.
References
138
Barros, C.P. (2008b): ‘Technical change and productivity growth in airports: A case study’,
Transportation Research Part A, 42(5), 818–832.
Barros, C.P. (2008c): ‘Technical efficiency of UK airports’, Journal of Air Transport
Management, 14(4), 175–178.
Barros, C.P. (2009): ‘The measurement of efficiency of UK airports, using a stochastic latent
class frontier model’, Transport Reviews, 29(4), 479-498.
Barros, C.P., and A. Assaf (2009): ‘Productivity change in USA airports: The Gillen and Lall
approach revisited’, Working Paper 22/2009/DE/UECE, School of Economics and
Management, Technical University of Lisbon.
Barros, C.P., and P.U. Dieke (2007): ‘Performance evaluation of Italian airports: a data
envelopment analysis’, Journal of Air Transport Management, 13(4), 184–191.
Barros, C.P., and P.U. Dieke (2008): ‘Measuring the economic efficiency of airports: A
Simar–Wilson methodology analysis’, Transportation Research Part E, 44(6), 1039–
1051.
Barros, C.P., and R.C. Marques (2008): ‘Performance of European airports: regulation,
ownership and managerial efficiency’, Working Paper 25/2008/DE/UECE, School of
Economics and Management, Technical University of Lisbon.
Barros, C.P., and A. Sampaio (2004): ‘Technical and allocative efficiency in airports’,
International Journal of Transport Economics, 31(3), 355–377.
Barros, C.P., and W.L. Weber (2009): ‘Productivity growth and biased technological change
in UK airports’, Transportation Research Part E, 45(4), 642–653.
Battese, G.E., and T.J. Coelli (1992): ‘Frontier production functions, technical efficiency and
panel data: with application to paddy farmers in India’, Journal of Productivity Analysis,
3(1), 153–169.
Battese, G.E., and T.J. Coelli (1995): ‘A model for technical inefficiency effects in a
stochastic frontier production function for panel data’, Empirical Economics, 20(2), 325-
332.
Bauer, P.W., Berger, A.N., Ferrier, G.D., and D.B. Humphrey (1998): ‘Consistency
conditions for regulatory analysis of financial institutions: a comparison of frontier
efficiency methods’, Journal of Economics and Business, 50(2), 85–114.
References
139
Bazargan, M., and B. Vasigh (2003): ‘Size versus efficiency: a case study of US commercial
airports’, Journal of Air Transport Management, 9(3), 187–193.
Beesley, M. E. (1999): ‘Airport Regulation’, in: Beesley, M. E. (ed.) Regulating Utilities: A
New Era?, Institute of Economic Affairs, London, ch. 4.
Bel, G. and X. Fageda (2009): ‘Privatization, regulation and airport pricing: an empirical
analysis for Europe’, Journal of Regulatory Economics, 37(2), 142-161.
Boardman, A.E., and A.R. Vining (1989): ‘Ownership and performance in competitive
environments: A comparison of the performance of private, mixed, and state-owned
enterprises’, Journal of Law and Economics, 32(1), 1-33.
Brockett, P.L., and B. Golany (1996): ‘Using rank statistics for determining programmatic
efficiency differences in data envelopment analysis’, Management Science, 42(3), 466–
472.
Button, K.J., and T.G. Weyman-Jones (1992): ‘Ownership structure, institutional organization
and measured X-efficiency’, The American Economic Review, 82(2), 439–445.
CAA (2000): ‘The Use of Benchmarking in the Airport Reviews’, Consultation paper, CAA,
London.
Caves, D.W., Christensen, L.R., and W.E. Diewert (1982a): ‘Multilateral comparisons of
output, input, and productivity using superlative index numbers’, The Economic Journal,
92, 73–86.
Caves, D.W., Christensen, L.R. and W.E. Diewert (1982b): ‘The economic theory of index
numbers and the measurement of input, output, and productivity’, Econometrica: Journal
of the Econometric Society, 50(6), 1393–1414.
Charkham, J.P. (1995): Keeping good company: a study of corporate governance in five
countries, Oxford University Press.
Charnes, A., Cooper, W.W., Golany, B., and J. Seiford (1985): ‘Foundations of data
envelopment analysis for Pareto-Koopmans efficient empirical production functions’,
Journal of Econometrics, 30(1-2), 91–107.
Charnes, A., Cooper, W.W., and E. Rhodes (1978): ‘Measuring the efficiency of decision
making units’, European Journal of Operational Research, 2(6), 429-444.
References
140
Charnes, A., Cooper, W.W. and E. Rhodes (1981): ‘Evaluating Program and Managerial
Efficiency: An Application of Data Envelopment Analysis to Program Follow Through’,
Management Science, 27(6), 668-697.
Chen, C.M. (2009): ‘A network-DEA model with new efficiency measures to incorporate the
dynamic effect in production networks’, European Journal of Operational Research,
194(3), 687-699.
Chi-Lok, A.Y., and A. Zhang, A (2009): ‘Effects of competition and policy changes on
Chinese airport productivity: An empirical investigation’, Journal of Air Transport
Management, 15(4), 166–174.
Chow, W., Kong, C., and M.K. Fung (2009): ‘Efficiencies and scope economies of Chinese
airports in moving passengers and cargo’, Journal of Air Transport Management, 15(6),
324-329.
Christensen, L.R. and D.W. Jorgenson (1969): ‘The measurement of US real capital input,
1929–1967’, Review of Income and Wealth, 15(4), 293–320.
Coelli, T., Estache, A., Perelman, S., and L. Trujillo (2003): A primer on efficiency
measurement for utilities and transport regulators, World Bank, Washington.
Coelli, T., Perelman, S. and E. Romano (1999): ‘Accounting for environmental influences in
stochastic frontier models: with application to international airlines’, Journal of
Productivity Analysis, 11(3), 251-273.
Coelli, T., Rao, D.S.P., and G.E. Battese (1998): An introduction to efficiency and
productivity analysis, Kluwer Academic Publishers, Boston.
Coelli, T., Rao, D.S.P., O’Donnell, C.J., and G.E. Battese (2005): An introduction to
efficiency and productivity analysis, second edition, Springer, Berlin.
Cologne Bonn Airport (2010); ‚Press Office’, http://www.airport-cgn.de/main.php?lang=
2&id=140, 18.03.2010.
Cook, A. J., Tanner, G. and S. Anderson (2004): Evaluating the true cost to airlines of one
minute of airborne or ground delay: final report, Project Report, Eurocontrol, Brussels.
Cooper, W.W., Seiford, L.M., and K. Tone (2007): Data envelopment analysis: a
comprehensive text with models, applications, references and DEA-solver software,
Second Edition, Springer Verlag, Berlin.
References
141
Cornwell, C., Schmidt, P., and R. Sickles (1990): ‘Production frontiers with cross-sectional
and time-series variation in efficiency levels’, Journal of Econometrics, 46, 185–200.
Czerny, A.I. (2006): ‘Price-cap Regulation of Airports: Single till Versus Dual till’, Journal
of Regulatory Economics, 30(1), 85-97.
de Borger, B., Kerstens, K., and A. Costa (2002): ‘Public transit performance: What does one
learn from frontier studies?’, Transport Reviews, 22(1), 1–38.
de la Cruz, S.F. (1999): ‘A DEA approach to the airport production function’, International
Journal of Transport Economics, 26, 255–270.
Dence, R. (1995): ‘Best-Practices benchmarking’, in Holloway, J., Lewis, J. and G. Mallory
(ed.) Performance Measurement and Evaluation, Sage Publications, ch. 4.
Dietz, P. (2009): ‚Verluste auf dem Rollfeld’, Frankfurter Rundschau, 11.11.2009.
Doganis, R. (2002): ‘Flying Off Course: The Economics of International Airlines’, 3rd ed.,
Taylor & Francis Ltd.
Doganis, R., Graham, A. and A. Lobbenberg (1995): ‘The economic performance of
European airports’, Research Report 3, Dept of Air Transport, Colleague of
Aeronautics, Cranfield University.
Doganis, R.S. and G.F. Thompson (1973): The economic of British airports, Transport
Studies Group.
Dortmund Airport (2007): Geschäftsbericht 2007, Flughafen Dortmund GmbH.
Efron, B. (1979): ‘Bootstrap methods: another look at the jackknife’, The Annals of Statistics,
7(1), 1-26.
Emrouznejad, A., and E. Thanassoulis (2005): ‘A mathematical model for dynamic efficiency
using data envelopment analysis’, Applied Mathematics and Computation, 160(2), 363-
378.
Eurocontrol (1999-2005): ‘Delays to Air Transport in Europe: Annual Report, Eurocontrol
Central Office for Delay Analysis, https://extranet.eurocontrol.int, 05.11.2010.
Eurocontrol (2006): ‘Delays to Air Transport in Europe: Digest-Annual 2005’, Eurocontrol
Central Office for Delay Analysis, https://extranet.eurocontrol.int, 05.11.2010.
Eurocontrol (2007-2008): ‘Digest-Annual: Delays to Air Transport in Europe’, Eurocontrol
Central Office for Delay Analysis, https://extranet.eurocontrol.int, 05.11.2010.
References
142
European Commission (2007): ‘Flying together’, Office for official publications of the
European Communities, Luxemburg.
Färe, R. (1991): ‘Measuring Farrell efficiency for a firm with intermediate inputs’, Academia
Economic Papers, 19(2), 329–340.
Färe, R., and S. Grosskopf (1996): ‘Productivity and intermediate products: a frontier
approach’, Economics Letters, 50(1), 65-70.
Färe, R., and S. Grosskopf (2000): ‚Network DEA’, Socio-Economic Planning Sciences,
34(1), 35-49.
Farrell, M.J. (1957): ‘The measurement of productive efficiency’, Journal of the Royal
Statistical Society. Series A, 120(3), 253-290.
Fernandes, E., and R.R. Pacheco (2002): ‘Efficient use of airport capacity’, Transportation
Research Part A, 36(3), 225–238.
Ferrier, G.D., and C.A. Lovell (1990): ‘Measuring cost efficiency in banking: econometric
and linear programming evidence’, Journal of Econometrics, 46(1-2), 229–245.
Flughafen Basel-Mulhouse (2007): ‘2007-The Year-Key Figures’, Flughafen Basel
Mulhouse.
Forsyth, P. (2000): ‘Models of airport performance’, in: Hensher, D.A. and Button, K.J. (ed.)
Handbook of Transport Modelling, Elsevier, Oxford, ch. 37.
Forsyth, P. (2004): ‘Replacing Regulation: Airport Price Monitoring in Australia’, in Forsyth,
P, Gillen, D., Knorr, A., Mayer, O. and H.-M. Niemeier (ed.) The economic regulation of
airports: recent developments in Australasia, North America and Europe, Ashgate, ch. 1.
Forsyth, P.J., Hill, R.D., and C.D. Trengove (1986): ‘Measuring airline efficiency’, Fiscal
Studies, 7(1), 61–81.
Fraport (2009) Groundbreaking Ceremony form Frankfurt Airport’s Runway Northwest,
http://www.ausbau.fraport.com/cms/default/dok/352/352869.groundbreaking_ceremony_f
or_frankfurt_ai.htm, 05.11.2010
Fried, H.O., Lovell, C.A.K., and S.S. Schmidt (2008): The measurement of productive
efficiency and productivity change, Oxford University Press, Oxford.
Fung, M.K., Wan, K.K.H., Hui, Y.V., and J.S. Law (2008): ‘Productivity changes in Chinese
airports 1995–2004’, Transportation Research Part E, 44(3), 521–542.
References
143
Gillen, D. (2010): ‘The evolution of airport ownership and governance’, Journal of Air
Transport Management, doi:10.1016/j.jairtraman.2010.10.003.
Gillen, D., and A. Lall, A. (1997): ‘Developing measures of airport productivity and
performance: an application of data envelope analysis’, Transportation Research Part E,
33(4), 261–273.
Gillen, D., and Lall (2001): ‘Non-parametric Measures of efficiency of US airports’,
International Journal of Transport Economics, 28(3), 283–306.
Gillen, D., and H.-M. Niemeier (2008): The European Union: Evolution of privatization,
regulation, and slot reform, in: Winston, C., and G. de Rus, (ed.) Aviation Infrastructure
Performance: A Study in Comparative Political Economy, Brookings Institution,
Washington, ch. 3.
Golany, B., and S. Thore (1997): ‘Restricted best practice selection in DEA: an overview with
a case study evaluating the socio-economic performance of nations’, Annals of Operations
Research, 73, 117–140.
Golaszewski, R. (2003): ‘Network industries in collision: aviation infrastructure capacity,
financing and the exposure to traffic declines’, Journal of Air Transport Management,
9(1), 57-65.
González, M.M., and L. Trujillo (2009): ‘Efficiency measurement in the port industry: A
survey of the empirical evidence’, Journal of Transport Economics and Policy, 43(2),
157–192.
Graham, A. (2004): Managing airports: an international perspective, second edition,
Elsevier, Amsterdam.
Graham, A. (2005): ‘Airport benchmarking: a review of the current situation’, Benchmarking:
An International Journal, 12(2), 99–111.
Graham, A., and T. Holvad (2000): ‘Efficiency Measurement for Airports’ paper submitted to
the Trafik Dags PAA Aalborg Universitet 2000 Conference, Aalborg University.
Greene, W. (2005): ‘Reconsidering heterogeneity in panel data estimators of the stochastic
frontier model’, Journal of Econometrics, 126(2), 269-303.
Guttman, L. (1968): ‘A general nonmetric technique for finding the smallest coordinate space
for a configuration of points’, Psychometrika, 33(4), 469–506.
References
144
Hensher, D.A., and W.G. Waters (1993): ‘Using total factor productivity and data
envelopment analysis for performance comparisons among government enterprises:
concepts and issues’, Institute of Transport Studies Working Paper No. ITS-WP-93-10,
The University of Sydney.
Hooper, P., Cain, R., and S. White (2000): ‘The privatisation of Australia's airports’,
Transportation Research Part E, 36(3), 181-204.
Hooper, P.G., and D.A. Hensher (1997): ‘Measuring total factor productivity of airports—an
index number approach’, Transportation Research Part E, 33(4), 249–259.
Hutter, D. (2009): ‚2000 Mitarbeitern droht die Kündigung’, Süddeutsche Zeitung,
29.05.2009.
IATA (2003): Airport capacity/demand profiles - 2003 edition, Air Transport Consultancy
Services, Monetral-Geneva.
IATA (2010): ‘World Scheduling Guidelines’, 19th Edition, http://www.iata.org/NR/Content
Connector/CS2000/SiteInterface/sites/whatwedo/scheduling/file/fdc/WSG-12thEd.pdf,
17.03.2010.
Jeong, J. (2005): ‘An investigation of operating costs of airports: focus on the effects of output
scale’, MSc-Thesis, Sauder School of Business, University of British Columbia.
Kamp, V., Niemeier, H.M., and J. Müller (2007): ‘What can be learned from benchmarking
studies? Examining the apparent poor performance of German airports’, Journal of
Airport Management, 1(3), 294–308.
Kao, C. (2009): ‘Efficiency decomposition in network data envelopment analysis: A relational
model’, European Journal of Operational Research, 192(3), 949-962.
Keeler, T.E. (1970): ‘Airport Costs and Congestion’, The American Economist, 14, 47-53.
Kincaid, I., and M. Tretheway (2006): ‘Guidelines for benchmarking airports’, paper
submitted to the German Aviation Research Society Conference, Hamburg, 22 February
2006.
Kumbhakar, S.C., and C.A.K. Lovell (2000): Stochastic frontier analysis, Cambridge
University Press, Cambridge.
Lam, S.W., Low, J.M., and L.C. Tang (2009): ‘Operational efficiencies across Asia Pacific
airports’, Transportation Research Part E, 45(4), 654–665.
References
145
Lewis, H.F., and T.R. Sexton (2004): ‘Network DEA: efficiency analysis of organizations
with complex internal structure’, Computers and Operations Research, 31(9), 1365-1410.
Liebert, V. (2010): ‘A review of empirical studies on the productivity and efficiency of
airports’, Mimeo. Jacobs University Bremen.
Lin, L.C., and C.H. Hong (2006): ‘Operational performance evaluation of international major
airports: an application of data envelopment analysis’, Journal of Air Transport
Management, 12(6), 342–351.
Liebert, V. and N. Adler (2010): ‘Joint Impact of Competition, Ownership Form and
Economic Regulation on Airport Performance’, Mimeo, Jacobs University Bremen.
Lipczynski, J., Wilson, J.O.S. and J. Goddard (2009): Industrial Organization: Competition,
Strategy & Policy, third edition, Financial Times Prentice Hall, Harlow.
Littlechild, S.C. (1983): Regulation of British Telecommunications' profitability: report to the
Secretary of State, February 1983, London, Department of Industry.
Lovell, C.A.K., and J.T. Pastor (1995): ‘Units invariant and translation invariant DEA
models’, Operational Research Letters, 18, 147-151.
Lozano, S., Gutiérrez, E., and J.L. Salmerón (2009): ‘Network DEA models in transportation-
Application to airports’, paper submitted to the German Aviation Research Society
Seminar on Airport Benchmarking, Berlin, 20 and 21 November 2009.
Lyon Aéroport (2008): ‘Les indicateurs trafic de l’aéroport de Lyon-Saint Exupéry’,
Aéroport Lyon - Saint Exupéry.
Main, B.G., Lever, B. and J. Crook (2003): Central Scotland airport study, Report, The David
Hume Institute.
Martín, J.C., and C. Román (2001): ‘An application of DEA to measure the efficiency of
Spanish airports prior to privatization’, Journal of Air Transport Management, 7(3), 149-
157.
Martín, J.C., and C. Román, C. (2006): ‘A benchmarking analysis of Spanish commercial
airports. A comparison between SMOP and DEA ranking methods’, Networks and Spatial
Economics, 6(2), 111–134.
Martín, J., and C. Román (2008): ‘The relationship between size and efficiency: A
benchmarking analysis of Spanish commercial airports’, Journal of Airport Management,
2(2), 183-197.
References
146
Martín, J.C., Román, C., and A. Voltes-Dorta (2009): ‘A stochastic frontier analysis to
estimate the relative efficiency of Spanish airports’, Journal of Productivity Analysis,
31(3), 163–176.
Martín, J., and A. Voltes-Dorta (2007): ‘Stochastic frontier efficiency estimation in airports: a
Bayesian approach’, Paper submitted to the 11th Air Transport Research Society
Conference 2007, University of California Berkeley, 21 to 23 June 2007.
Martın-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), 233–238.
Maurer, P. (2003): ‚Luftverkehrsmanagement’, third edition., Oldenbourg Wissenschafts-
Verlag, München.
Meeusen, W., and J. van den Broeck (1977): ‘Efficiency estimation from Cobb-Douglas
production functions with composed error’, International Economic Review, 18(2), 435-
444.
Megginson, W.L. and J.M. Netter (2001): ‘From state to market: A survey of empirical
studies on privatization’, Journal of Economic Literature, 39(2), 321-389.
Morrison, W.G. (2009): ‘Understanding the complexities and challenges of airport
performance benchmarking’, Journal of Airport Management, 3(2), 145–158.
Murillo-Melchor, C. (1999): ‘An analysis of technical efficiency and productivity changes in
Spanish airports by using the Malmquist index’, International Journal of Transport
Economics, 26, 271–292.
Nyshadham, E.A., and V.K. Rao (2000): ‘Assessing efficiency of European airports: a total
factor productivity approach’, Public Works Management & Policy, 5(2), 106-114.
O’Donnell, C.J., Rao, D.S.P. and G.E. Battese (2007): ‘Metafrontier frameworks for the study
of firm-level efficiencies and technology ratios’, Empirical Economics, 34(2), 231-255.
Orea, L., and S.C. Kumbhakar (2004): ‘Efficiency measurement using a latent class stochastic
frontier model’, Empirical Economics, 29(1), 169-183.
Oum, T.H., Adler, N., and C. Yu (2006): ‘Privatization, corporatization, ownership forms and
their effects on the performance of the world's major airports’, Journal of Air Transport
Management, 12(3), 109–121.
References
147
Oum, T.H., Tretheway, M.W., and W.G. Waters (1992): ‘Concepts, methods and purposes of
productivity measurement in transportation’, Transportation Research Part A, 26(6), 493–
505.
Oum, T.H., Waters, W.G., and C. Yu (1999): ‘A survey of productivity and efficiency
measurement in rail transport’, Journal of Transport Economics and Policy, 33(1), 9-42.
Oum, T.H., Yan, J., and C. Yu (2008): ‘Ownership forms matter for airport efficiency: A
stochastic frontier investigation of worldwide airports’, Journal of Urban Economics,
64(2), 422-435.
Oum, T.H., and C. Yu (2004): ‘Measuring airports’ operating efficiency: a summary of the
2003 ATRS global airport benchmarking report’, Transportation Research Part E, 40(6),
515–532.
Oum, T.H., Yu, C., and X. Fu (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), 285–297.
Oum, T.H., Zhang, A., and Y. Zhang (2004): ‘Alternative forms of economic regulation and
their efficiency implications for airports’, Journal of Transport Economics and Policy,
38(2), 217–246.
Pacheco, R.R., and E. Fernandes (2003): ‚Managerial efficiency of Brazilian airports’,
Transportation Research Part A, 37(8), 667–680.
Pacheco, R.R., Fernandes, E., and M.P. de Sequeira Santos (2006): ‘Management style and
airport performance in Brazil’, Journal of Air Transport Management, 12(6), 324–330.
Parker, D. (1999): ‘The performance of BAA before and after privatisation: A DEA study’,
Journal of Transport Economics and Policy, 33(2), 133-145.
Pathomsiri, S., Haghani, A., Dresner, M., and R.J. Windle (2008): ‘Impact of undesirable
outputs on the productivity of US airports’, Transportation Research Part E, 44(2), 235–
259.
Pels, E., Nijkamp, P., and P. Rietveld (2001): ‘Relative efficiency of European airports’,
Transport Policy, 8(3), 183–192.
Pels, E., Nijkamp, P., and P. Rietveld (2003): ‘Inefficiencies and scale economies of
European airport operations’, Transportation Research Part E, 39(5), 341–361.
References
148
Pitt, M.M., and L.F. Lee (1981): ‘The measurement and sources of technical inefficiency in
the Indonesian weaving industry’, Journal of Development Economics, 9(1), 43-64.
Ramanathan, R. (2003): ‘An introduction to data envelopment analysis: a tool for
performance measurement’, Sage Publications Pvt. Ltd., New Delhi.
Reinhold, A., Niemeier, H.-M., Kamp, V., and J. Müller (2008): ‘The pros and cons of
benchmark regulation for airports’, Paper submitted to the 12th Air Transport Research
Society Conference 2008, Athens, 6 to 10 July 2008.
Reinhold, A., Niemeier, H.-M., Kamp, V., and J. Müller (2010): ‘An evaluation of yardstick
regulation for European airports’, Journal of Air Transport Management, 16(2), 74-80.
Sappington, D.E.M., and Stiglitz, J.E. (1987): ‘Privatization, information and incentives’,
Journal of Policy Analysis and Management, 6(4), 567-582.
Sarkis, J. (2000): ‘An analysis of the operational efficiency of major airports in the United
States’, Journal of Operations Management, 18(3), 335–351.
Sarkis, J., and S. Talluri (2004): ‘Performance based clustering for benchmarking of US
airports’, Transportation Research Part A, 38(5), 329–346.
Schmidt, P., and R.C. Sickles (1984): ‘Production frontiers and panel data’, Journal of
Business & Economic Statistics, 2(4), 367-374.
Shapiro, C., and R.D. Willig, (1990): ‘Economic rationales for the scope of privatization’, in:
Suleiman, E.N., and J. Waterbury (ed.) The political economy of public sector reform and
privatization, Westview Press, Boulder.
Shleifer, A. (1985): ‘A theory of yardstick competition’, The RAND Journal of Economics,
16(3), 319-327.
Shleifer, A., and R.W. Vishny (1994): ‘Politicians and firms’, The Quarterly Journal of
Economics, 109(4), 995-1025.
SH&E (2002): ‘Appendices: Study on the quality and efficiency of Ground handling services
at EU airports as a result of the implementation of Council Directive 96/67/EC’, Report to
European Commission, SH&E International Air Transport Consultancy, London.
Simar, L., and P.W. Wilson (1998): ‘Sensitivity analysis of efficiency scores: How to
bootstrap in nonparametric frontier models’, Management Science, 44(1), 49-61.
Simar, L., and P.W. Wilson (2000): ‘A general methodology for bootstrapping in non-
parametric frontier models’, Journal of Applied Statistics, 27(6), 779–802.
References
149
Simar, L., and P.W. Wilson (2007): ‘Estimation and inference in two-stage, semi-parametric
models of production processes’, Journal of Econometrics, 136(1), 31-64.
Starkie, D. (2002): ‘Airport regulation and competition’, Journal of Air Transport
Management, 8(1), 63-72.
Stone, M. (2002): ‘How not to measure the efficiency of public services (and how one
might)’, Journal of the Royal Statistical Society: Series A, 165(3), 405–434.
Sueyoshi, T., and S. Aoki (2001): ‘A use of a nonparametric statistic for DEA frontier shift:
the Kruskal and Wallis rank test’, Omega, 29(1), 1–18.
Templin, C. (2007): ‚Bodenabfertigungsdienste an Flughäfen in Europa: Deregulierung und
ihre Konsequenzen’, Kölner Wissenschaftsverlag, Cologne.
The Guardian (2010): ‘Third runway plan for Heathrow scrapped by BAA’,
http://www.guardian.co.uk/environment/2010/may/24/third-runway-heathrow-scrapped-
baa, 24.05.2010.
Tolofari, S., Ashford, N.J. and R.E. Caves (1990): The cost of air service fragmentation,
Department of Transport Technology, University of Loughborough.
Tone, K., and M. Tsutsui (2009): ‘Network DEA: A slacks-based measure approach’,
European Journal of Operational Research, 197(1), 243-252.
Tovar, B., and R.R. Martín-Cejas (2009): ‘Are outsourcing and non-aeronautical revenues
important drivers in the efficiency of Spanish airports?’, Journal of Air Transport
Management, 15(5), 217-220.
Tovar, B., and R.R. Martín-Cejas (2010): ‘Technical efficiency and productivity changes in
Spanish airports: A parametric distance functions approach’, Transportation Research
Part E, 46(2), 249-260.
Tretheway, M.W., and Kincaid, I. (2010): ‘Competition between airports: occurrence and
strategy’, in Forsyth, P., Gillen, D., Müller, J., and H.-M. Niemeier (ed.) Airport
competition: The European experience, Ashgate, ch. 9.
van Dender, K. (2007): ‘Determinants of fares and operating revenues at US airports’,
Journal of Urban Economics, 62(2), 317-336.
Vasigh, B., and J. Gorjidooz (2006): ‘Productivity analysis of public and private airports: A
causal investigation’, Journal of Air Transportation, 11(3), 144-63.
References
150
Vasigh, B. and M. Haririan (2003): ‘An empirical investigation of financial and operational
efficiency of private versus public airports’, Journal of Air Transportation, 8(1), 91-109.
Vickers, J., and G. Yarrow (1991): ‘Economic perspectives on privatization’, The Journal of
Economic Perspectives, 5(2), 111-132.
Vienna International Airport (2001, 2004, 2007): Geschäftsbericht, Flughafen Wien AG.
Vogel, H.A. (2006): ‘Impact of privatisation on the financial and economic performance of
European airports’, Aeronautical Journal, 110(1106), 197-213.
von Hirschhausen, C., and A. Cullmann (2005): ‚Questions to airport benchmarkers -
some theoretical and practical aspects learned from benchmarking other sectors’, Paper
submitted to the German Aviation Research Society Conference on Benchmarking and
Airport Competition, Vienna, 24 and 25 November 2005.
White, H. (1980): ‘A heteroskedasticity-consistent covariance matrix estimator and a direct
test for heteroskedasticity’, Econometrica: Journal of the Econometric Society, 48(4),
817–838.
Yokomi, M. (2005): ‘Measurement of Malmquist index of privatized BAA plc’, Paper
submitted to the 9th Air Transport Research Society Conference 2005, Rio de Janeiro, 3 to
7 July 2005.
Yoshida, Y. (2004): ‘Endogenous-weight TFP measurement: methodology and its application
to Japanese-airport benchmarking’, Transportation Research Part E, 40(2), 151–182.
Yoshida, Y., and H. Fujimoto (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), 533–546.
Yu, M.M. (2004): ‘Measuring physical efficiency of domestic airports in Taiwan with
undesirable outputs and environmental factors’, Journal of Air Transport Management,
10(5), 295–303.
Yu, M.M., and E.T. Lin (2008): ‘Efficiency and effectiveness in railway performance using a
multi-activity network DEA model’, Omega, 36(6), 1005–1017.
Zhang, A., and Y. Zhang (2003): ‘Airport charges and capacity expansion: effects of
concessions and privatization’, Journal of Urban Economics, 53(1), 54–75.
Zhang, A. and Y. Zhang (2010): ‘Airport capacity and congestion pricing with both
aeronautical and commercial operations’, Transportation Research Part B, 44(3), 404-413.
References
151
So which airport is really efficient?
Source: www.leipzig-halle-airport.de
Source: http://www.homato.com/photo/flughafen.jpg