Description
An airport is a location where aircraft such as fixed-wing aircraft, helicopters, and blimps take off and land. Aircraft may be stored or maintained at an airport. An airport consists of at least one surface such as a runway for a plane to take off and land, a helipad, or water for takeoffs and landings, and often includes buildings such as control towers, hangars and terminal buildings
ABSTRACT
Title: ASSESSMENT OF PRODUCTIVE EFFICIENCY OF AIRPORTS Somchai Pathomsiri Doctoral of Philosophy, 2006 Directed By: Professor Ali Haghani Department of Civil and Environmental Engineering
The move towards commercialization and privatization has pressured airports to become more productive and competitive. The need to devise an overall (total) productivity measure is increasingly important in airport business. The dissertation made three major research contributions. First, it assessed the productivity of airports operating in multiple airport systems (MASs). Second, it developed a more complete total factor productivity measure by considering joint production of desirable and undesirable outputs. Third, it developed models for explaining variations in productive efficiency. These are accomplished in two case studies. In case study 1, the Data Envelopment Analysis (DEA) is used to assess relative total productivity of 72 airports operating in 25 MASs during 2000 – 2002. The results indicate that highly utilized airports such as O’Hare International, Los Angeles International, Heathrow/London and LaGuardia are classified as efficient. The Censored Tobit regression model suggests that runway utilization market dominance, proportion of international passengers and ownership can be used to explain variations in productive efficiency.
In case study 2, the directional output distance function is applied to assess the productivity of 56 U.S. commercial airports during 2000 – 2003. Delays are considered as undesirable outputs. There are several important findings and insightful implications. First, about half of U.S. airports are actually operated efficiently. These airports include busy airports such as Hartsfield-Jackson Atlanta, LaGuardia, and Memphis together with less busy airports with relatively low delays such as Baltimore/Washington International and Oakland International. Second, the overall system has potential to accommodate about 1,550 million passengers, 26 million movements and 34 million tons of cargo. Third, during 2000 - 2003, annual growth of productivity is modest in the range of -1.3% to +1.8%. Fourth, by ignoring delays the assessment provides drastically different results in terms of number of efficient airports, level of inefficiency, ranking, and estimated maximum possible outputs. Fifth, the consideration of undesirable output is as important as the consideration of additional inputs and desirable outputs. The Censored Tobit regression model suggests that runway utilization, proportion of international passengers and average delay per passenger can be used to explain variations in productive efficiency.
ASSESSMENT OF PRODUCTIVE EFFICIENCY OF AIRPORTS
By Somchai Pathomsiri
Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park in partial fulfillment of the requirements for the degree of Doctoral of Philosophy 2006
Advisory Committee: Professor Ali Haghani, Chair Assistant Professor Kelly J. Clifton Professor Martin E. Dresner Professor Hani S. Mahmassani Professor Paul M. Schonfeld Professor Robert J. Windle
© Copyright by Somchai Pathomsiri 2006
Assessment of Productive Efficiency of Airports
Preface
As a transportation professional, I am rather fortunate to be able to work in all modes of transportation, i.e., land, water and air. This dissertation adds another chapter in air transportation experience to my career. I was interested in aviation since I was an undergraduate student studying pavement design for airfield. But it was not until 1990s that I realized the fascination of aviation systems planning when I worked as a transport engineer in two projects, i.e., 1) Airport Systems Master Plan Study in Thailand and 2) Feasibility Study and Master Plan Development for Joint Military-Civilian Used UTaphao International Airport. It was so memorable time to work with several aviation professionals including an old-hand project manager and a good friend of my family, Mr. Clifford R. King (then with Louis Berger International, Inc., currently a senior project manager at Bechtel Corporation). Since then, I have had aviation in my heart. This dissertation was accidentally started while I was completing a term paper on airport choice modeling in Baltimore-Washington multiple airport system for a class ENCE688Y (Advances in Transportation Demand Analysis) in the Fall 2002 taught by Professor Hani Mahmassani. The literature review led me to learn further that aviation community is interested in measuring performance of airport for a variety of reasons, including benchmarking and investment appraisal. I was so surprised to know that research to assess overall airport performance had just started in late 1990s. I then started working on the topic seriously and published our first paper “Benchmarking efficiency of airports in the competitive multiple-airport systems: the international perspective” at the 19th Annual Transport Chicago Conference in June 2004 (with Professor Ali Haghani). Subsequently, Professors Paul Schonfeld, Martin Dresner and Robert Windle kindly ii
Assessment of Productive Efficiency of Airports
accepted the invitation to jointly work and improve the quality of papers. We co-authored another two papers for the 84th and 85th Transportation Research Board Annual Meeting, Washington DC, 2005 - 2006. I am particularly grateful to many valuable comments from my co-authors and several anonymous reviewers. Their comments enabled us to extend and expand the scope of our research, essentially to answer practical aviation issues. One of the frequent comments is about my overemphasis on quantity of outputs and ignoring their quality, although measures such as delays are a major concern of the airport management. Such comments bring the dissertation to the last phase, i.e., the assessment of airport productivity with joint consideration of desirable and undesirable outputs. I stumbled upon a relatively new theory in production economics, i.e., the directional output distance function, while I was searching for a method to deal with undesirable outputs in the productivity assessment. I received useful guidance from Professors Rolf Färe and Shawna Grosskopf of Oregon State University who devised the theory and eminently populated applications in recent years. We could start a new research on airport productivity by jointly considering delays as major output measures, though undesirable, along with other traditional desirable outputs (e.g., number of passengers, aircraft movement, and freight throughput). We have published the new findings in three papers so far, i.e., National Urban Freight Conference, Long Beach, CA (February 2006), the 47th Annual Transportation Research Forum, New York (March 2006) and the 10th Annual Air Transport Research Society (ATRS) World Conference, Nagoya, Japan (May 2006). This dissertation is partly based
iii
Assessment of Productive Efficiency of Airports
on the above-mentioned six publications and another paper which is under review by Transportation Research Part E: Logistics and Transportation Review.
iv
Assessment of Productive Efficiency of Airports
Dedication
To my mother, U-SA Kow whose vision is always clear and correct. You enabled so many things in my life and enlighten me in many ways. I can be successful today because I have you, mom. I really can’t thank you enough.
v
Assessment of Productive Efficiency of Airports
Acknowledgements
Always for this level of accomplishment, there are many people directly or indirectly involved. I apologize if I unintentionally omit one of you. First of all, I wish to express my gratitude to my academic advisor, Professor Ali Haghani for continuous support throughout my doctoral study. Working under his supervision has been particularly rewarding and most gratifying, especially since I have been able to carry out research in various fields (e.g., logistics, travel behavior and demand modeling, GIS, operations research, and aviation system planning) with his full support and confidence. I also want to thank Professors Kelly Clifton, Martin Dresner, Hani Mahmassani, Paul Schonfeld, and Robert Windle for serving in my dissertation committee. Their comments and suggestions were of particularly great value to the quality of the dissertation. In particular, I highly appreciate friendly advice from Professors Paul Schonfeld, Robert Windle and Martin Dresner while we were co-authoring several papers. I learned endless lessons from them and really enjoyed working with them. Hopefully, we could continue the productive collaboration in the future endeavors. Special thanks also go to Professors Rolf Färe and Shawna Grosskopf of Oregon State University for their guidance on the theory and rich applications of the directional output distance function. I greatly benefited from a string of communications with them. In addition, comments and discussion with Professors Steven Burks of the University of Minnesota and B. Starr McMullen of the Oregon State University greatly improve the quality of the dissertation. There are also many airport staff members and managers in the U.S. and around the world who willingly shared their airport information and had fruitful discussions about airport business. I highly appreciate their contribution.
vi
Assessment of Productive Efficiency of Airports
Finally, I want to share my success and pleasure with my wonderful family members who have stood beside me with love and encouragement for years. I am forever indebted to their immeasurable, constant and continuous and tireless support during my study, especially, my wife, Laddawan whose sacrifice has made nothing impossible. I am so proud to have you and our lovely three children, Nawanont, Nontthida and Phuwanont in my life. I also want to express the gratitude to my wife’s family members who always strongly encourage and support us. It is the highest honor of my life to have such a great family support.
vii
Assessment of Productive Efficiency of Airports
TABLE OF CONTENTS
Page Preface ……………………………………………………………………………………ii Dedication ………………………………………………………………………………. v Acknowledgements ……………………………………………………………………...vi TABLE OF CONTENTS ……………………………………………………………...viii LIST OF TABLES ……………………………………………………………………...xii LIST OF FIGURES …………………………………………………………………... xv CHAPTER 1 1.1 1.2 1.3 1.4 1.5 1.6 INTRODUCTION ……………………………………………………. 1
Airport business ………………………………………………………………….. 1 Importance of airport productivity study ………………………………………… 2 Motivation of the dissertation research …………………………………………... 5 Research objectives and scope …………………………………………………... 5 Research contributions …………………………………………………………… 6 Organization of the dissertation ………………………………………………….. 6 LITERATURE REVIEW ……………………………………………. 9
CHAPTER 2 2.1
Productivity Measures …………………………………………………………… 9 2.1.1 2.1.2 Partial Factor Productivity (PFP) measure ………………………………. 9 Total Factor Productivity (TFP) measure ………………………………. 12
2.2
Methodology for computing TFP measure ……………………………………... 13 2.2.1 2.2.2 Parametric approach ……………………………………………………..13 Non-parametric approach ………………………………………………..17
viii
Assessment of Productive Efficiency of Airports
TABLE OF CONTENTS (Continue)
Page CHAPTER 2 2.3 LITERATURE REVIEW
Data Envelopment Analysis (DEA) …………………………………………….. 35 2.3.1 2.3.2 Background ……………………………………………………………... 35 Model development …………………………………………………….. 37
2.4
Discussion ………………………………………………………………………. 43 RESEARCH METHODOLOGY …………………………………... 48
CHAPTER 3 3.1 3.2 3.3 3.4
Characterization of production possibility set ……………………………………. 48 Output distance function ………………………………………………………….. 49 Directional output distance function ……………………………………………… 51 Malmquist productivity index with the presence of undesirable outputs ………… 58 CASE STUDY 1
CHAPTER 4
PRODUCTIVITY OF AIRPORTS IN MULTIPLE AIRPORT SYSTEMS ……… 68 4.1 4.2 4.3 4.4 Definition of multiple airport system ……………………………………………...69 Modeling airport operations ……………………………………………………….74 Input and output measures of airport operations …………………………………..74 Data collection ……………………………………………………………………. 76 CASE STUDY 1: RESULTS AND DISCUSSION………………… 81
CHAPTER 5 5.1 5.2 5.3 5.4
Selection of a DEA model ………………………………………………………... 81 Efficiency scores ………………………………………………………………….. 82 Determination of airport productivity …………………………………………….. 86 Factors affecting productive efficiency of airports in MASs ……………………...88
ix
Assessment of Productive Efficiency of Airports
TABLE OF CONTENTS (Continue)
Page CHAPTER 6 CASE STUDY 2
PRODUCTIVITY OF U.S. AIRPORTS ……………………………………………... 98 6.1 6.2 Modeling airport operations ……………………………………………………….98 Input and output measures of airport operations …………………………………..99
6.3 Sample characteristics …………………………………………………………… 102 6.3.1 Size of sample ……………………………………………………………... 102 6.3.2 Analysis period ……………………………………………………………. 104 6.3.3 Data source and definition ………………………………………………… 105 CHAPTER 7 7.1 CASE STUDY 2: RESULTS AND DISCUSSION ………………. 115
Impact of the inclusion of undesirable outputs …………………………………...115 7.1.1 Classification of efficient airports ……………………………………….. 117 7.1.2 The number of efficient airports ………………………………………… 120
7.1.3 Difference in efficiency scores …………………………………………...121 7.1.4 7.1.5 Ranking …………………………………………………………………..121 Maximum possible production outputs ………………………………….. 126
7.2 Lumpiness of airport investment ………………………………………………... 136 7.3 Changes in productivity over time ………………………………………………. 137
7.4 Scenario analysis ………………………………………………………………… 148 7.5 Determination of airport productivity …………………………………………… 167
7.6 Factors affecting productive efficiency of U.S. airports ………………………… 169
x
Assessment of Productive Efficiency of Airports
TABLE OF CONTENTS (Continue)
Page CHAPTER 8 8.1 CONCLUSIONS AND FUTURE RESEARCH ………………….. 175
Conclusions from assessing productivity of airports in MASs …………………..175 8.1.1 Productivity of airports in MASs ………………………………………... 175 8.1.2 Underutilized airports …………………………………………………….176
8.1.3 Factors affecting productive efficiency of airports in MASs …………….177 8.2 Conclusions from assessing productivity of U.S. airports ………………………. 177 8.2.1 Productivity of U.S. commercial airports ……………………………….. 179 8.2.2 Productivity growth of U.S. commercial airports ……………………….. 180 8.2.3 8.2.4 Impact of delays on airport productivity ………………………………… 181 Selection of input and output measures …………………………………. 183
8.2.5 Factors affecting productive efficiency of U.S. airports ………………… 183 8.3 Suggested future research ……………………………………………………….. 184
REFERENCES ………………………………………………………………………..188 Bibliography ………………………………………………………………………….. 206
xi
Assessment of Productive Efficiency of Airports
LIST OF TABLES Page Table 2.1 Examples of partial factor productivity measures in aviation sector ……….. 11 Table 2.2 List of publications on airport productivity studies by parametric Approach ……………………………………………………………………. 16 Table 2.3 List of publications on airport productivity studies using index number method ……………………………………………………………... 18 Table 2.4 List of publications on airport productivity studies using DEA ……………. 23 Table 2.5 Summary of DEA models …………………………………………………... 44 Table 4.1 List of multiple airport systems and corresponding Herfindahl Concentration Indices, 2002 ………………………………………………... 72 Table 4.2 Descriptive statistics of 72 airports in MASs, 2000 – 2002 ………………... 80 Table 5.1 Efficiency scores, 2000 – 2002 …………………………………………….. 84 Table 5.2 Comparisons of statistics between efficient and inefficient airports ……….. 92 Table 5.3 Censored Tobit regression: preliminary model estimation results …………. 93 Table 5.4 Censored Tobit regression: proposed model estimation results ……………. 94 Table 6.1 List of 56 US airports under consideration and their outputs in 2003 ……...109 Table 6.2 Descriptive statistics of samples 2000 – 2003 …………………………….. 114 Table 7.1 Efficiency scores for Case 1 and Case 2 …………………………………... 118 Table 7.2 Comparisons of efficiency scores between Cases 1 and 2 by paired sample t-test ………………………………………………………... 122 Table 7.3 Comparisons of efficiency scores between Cases 1 and 2 by Nonparametric Paired tests ………………………………………………... 123 Table 7.4 Ranking of airport productivity …………………………………………… 124
xii
Assessment of Productive Efficiency of Airports
LIST OF TABLES (Continue) Page Table 7.5 Maximum possible passengers, aircraft movements and cargo throughput in 2000 ………………………………………………………… 128 Table 7.6 Maximum possible passengers, aircraft movements and cargo throughput in 2001 ………………………………………………………… 130 Table 7.7 Maximum possible passengers, aircraft movements and cargo throughput in 2002 ………………………………………………………… 132 Table 7.8 Maximum possible passengers, aircraft movements and cargo throughput in 2003 ………………………………………………………… 134 Table 7.9 Luenberger productivity indexes, Case 1 …………………………………..138 Table 7.10 Luenberger productivity indexes, Case 2 ………………………………...143 Table 7.11 Comparisons of Luenberger productivity indexes between Cases 1 and 2 ……………………………………………………………. 145 Table 7.12 Comparisons of Luenberger productivity indexes by paired sample t-test ……………………………………………………………... 147 Table 7.13 Comparisons of Luenberger productivity indexes by nonparametric tests ……………………………………………………… 147 Table 7.14 Efficiency scores for Cases 1 and 3 ……………………………………... 151 Table 7.15 Comparisons of efficiency scores between Cases 1 and 3 by paired sample t-test ……………………………………………………… 153 Table 7.16 Comparisons of efficiency scores between Cases 1 and 3 by nonparametric paired tests ………………………………………………. 154
xiii
Assessment of Productive Efficiency of Airports
LIST OF TABLES (Continue) Page Table 7.17 Maximum possible passengers, aircraft movements and cargo throughput in 2000, Cases 1 and 3 ………………………………………. 156 Table 7.18 Maximum possible passengers, aircraft movements and cargo throughput in 2001, Cases 1 and 3 ………………………………………. 158 Table 7.19 Maximum possible passengers, aircraft movements and cargo throughput in 2002, Cases 1 and 3 ………………………………………. 160 Table 7.20 Maximum possible passengers, aircraft movements and cargo throughput in 2003, Cases 1 and 3 ………………………………………. 162 Table 7.21 Luenberger productivity indexes, Cases 1 and 3 …………………………164 Table 7.22 Comparisons of Luenberger productivity indexes by paired sample t-test Cases 1and 3 ………………………………………………. 166 Table 7.23 Comparisons of Luenberger productivity indexes by nonparametric paired tests ………………………………………………. 167 Table 7.24 Censored Tobit regression model estimation results ……………………. 172
xiv
Assessment of Productive Efficiency of Airports
LIST OF FIGURES Page
Figure 2.1 Difference between DEA and regression concept ………………………... 37 Figure 2.2 Determination of efficient production frontier …………………………… 39 Figure 3.1 Output possibility set and distance functions …………………………….. 50 Figure 3.2 Graphical illustration of directional output distance function concept …… 55 Figure 3.3 The Luenberger productivity indicator …………………………………… 64 Figure 6.1 Locations of 56 airports …………………………………………………. 104 Figure 6.2 Scatter plot between number of delayed flights and number of passengers, 2003 ………………………………………………………… 112 Figure 6.3 Scatter plot between delay/passenger and density of movements, 2000 – 2003 ………………………………………………………………112
xv
Assessment of Productive Efficiency of Airports
CHAPTER 1 INTRODUCTION
1.1 Airport business The aviation industry has experienced swift ups and downs after the deregulation of air transportation in the late 1970s. The competition in the air transportation industry is now fierce. In recent years, it has become more common for airports to advertise and promote their services to lure customers just like other businesses do. Airport business models have changed dramatically from being perceived as a fundamental public service in the same way as roads and public transport, to a commercial activity. Over the past twenty years, it has become obvious that airports can actually be run as highly successful and profitable businesses (Doganis, 1992). On one hand, airports are an essential part of the air transportation system. They provide all the infrastructure needed to enable passengers and freight to transfer from surface to air mode of transport and to allow airlines to take off and land. On the other hand, airports also offer a wide variety of commercial facilities ranging from shops, restaurants, hotels, and conference services (Graham, 2003; Jarach, 2005). There is no doubt that an airport can be a big business. The public floatation of the British Airports Authority in the summer of 1987 is valued at £1.3 billion (Doganis, 1992). It is expected that rapid growth of air traffic would require enormous amount of funding to support airport improvement programs. This places increasing pressures on public finance. Such pressures have led governments all over the world to consider privatization and commercialization to relieve them from the financial burden of airport ownership -1-
Assessment of Productive Efficiency of Airports
(Ashford, 1999; Francis and Humphreys, 2001). Airport managers have to adapt themselves in response to such pressures as well. They are now acting more like corporate business managers. They think strategically, identify markets, set objectives and goals, cast competitive strategies, implement, monitor, evaluate outcomes, and respond to the dynamics of market competition. Their job is much more complicated than before. 1.2 Importance of airport productivity study To study airport productivity is to study the relationship between inputs and outputs of airport operation. With such a relation, airport managers can easily determine the probable traffic level that airports should accommodate, given any level of inputs. This is very useful in monitoring, managing and planning airports. In addition, it allows managers to benchmark their operational performance with peers and set appropriate output targets for improving their business. Until the 1980s, the systematic monitoring and comparing of airport performance was not a widely practiced activity within the airport industry. This can largely be attributed to insufficient commercial and business pressures for airports and the general lack of experience of benchmarking techniques within the public sector. With airport privatization and commercialization has come a marked interest in performance comparisons and benchmarking. As airports become more commercially-oriented, they have been keen to identify the strong performers in the industry and adopt what are seen as best practices (Graham, 2003). Hooper and Hensher (1997) commend that the growing importance of airport performance measurement is accompanied by the trend toward corporatized or even privatized airports. In recent survey from the world’s top 200 busiest
-2-
Assessment of Productive Efficiency of Airports
passenger airports (Francis, Humphreys and Fry, 2002; Humphreys and Francis, 2002), the results reveal that airport managers are now using several performance measures to monitor their businesses. Several authors comment on the importance of studying airport productivity. For instance Sarkis (2002) argues that evaluating airport operational efficiency is important for a number of reasons including communities’ reliance on airports for economic wellbeing; air carriers’ ability to choose among competing airports due to deregulation, and the fact that federal funding for airport improvements is based on performance measures. Performance evaluation and improvement studies of airport operations have important implications for a number of airport stakeholders. They assist air carriers in identifying and selecting more efficient airports on which to base their operations. Likewise municipalities would benefit from efficient airports in terms of attracting business and passengers. They also assist federal government in making effective decisions on optimal allocation of resources to airport improvement programs, and in evaluating the efficacy of such programs on the bottom line efficiency of airports. Finally, benchmarking their own airports against comparable airports is one way for operations managers to ensure competitiveness (Sarkis and Talluri, 2004). The need to develop appropriate service and productivity indicators for airport operation has been recognized and there is a small, but growing literature on the subject. Though there have been appeals to measure “overall productivity”, there is little evidence that the tools of productivity measurement that have been applied in other parts of the transport sector have had serious application in the case of airports (Tretheway, 1995). As the literature review in Chapter 2 will reveal, it only began in the late 1990s. Until
-3-
Assessment of Productive Efficiency of Airports
recently, a good number of studies have been conducted comparing productivity and operational efficiency of airports around the world, including Australia (Abbott and Wu, 2001; Hooper and Hensher, 1997), U.K. (Parker, 1999), U.S. (Gillen and Lall, 1997, 1998; Bazargan and Vasigh, 2003; Sarkis, 2000; Sarkis and Talluri, 2004), Spain (Martin and Roman, 2001), Brazil (Fernandes and Pacheco, 2001, 2002, 2005; Pacheco and Fernandes, 2003), Japan (Yoshida, 2004; Yoshida and Fujimoto, 2004). Occasionally the scope was expanded beyond a country to one continent such as Europe (Pels, Nijkamp and Rietveld., 2001, 2003) and international level (Adler and Berechman, 2001; Oum and Yu, 2004; Oum, Yu and Fu, 2003). Surprisingly, none has ever studied the productivity of airports operating in a specific market such as multiple airport systems (MAS), although they involve much more capital investment. At best, MAS airports are treated in the mixed samples with airports from single airport systems. The exception are only Pathomsiri and Haghani (2004); Pathomsiri, Haghani and Schonfeld (2005); Pathomsiri, Haghani, Dresner and Windle, (2006a) which will be summarized within this dissertation. Moreover, airport productivity research is rather restricted in the sense that productive efficiency is solely based on consideration of marketed outputs. Nonmarketed outputs or so-called “undesirable outputs” such as delays have been largely ignored, though they are also a major concern to airport stakeholders. This may be due to the lack of analysis technique. In principle where there is joint production of desirable and undesirable outputs, accounting for both of them intuitively should provide a more complete measure of airport productivity.
-4-
Assessment of Productive Efficiency of Airports
1.3
Motivation of the dissertation research This dissertation is motivated by several factors. First, it is clear that the aviation
society still has insufficient understanding regarding the productivity of airports operating in specific markets such as multiple airport systems. Second, the aviation society lacks understanding of the relationship between airport’s inputs and outputs, especially when undesirable byproducts are taken into consideration. Development of an applicable model is eminently necessary. The results should give a more complete measurement of airport productivity. Third, it is believed that the results have substantial implications which are very useful for managing airports. Last but not least, as the literature review in chapter 2 will reveal, the dissertation is a pioneering work. It likely creates an impact and entices researchers to re-think the way they assess productivity of airports. It is expected that further development of the applicable models for fairer assessment will follow. 1.4 Research objectives and scope This dissertation attempts to address the shortcomings of the previous airport productivity studies. In particular, it aims to accomplish the following four main objectives. 1) Assess the productivity of airports operating specifically in multiple airport systems as well as develop a model for predicting their relative efficiency. 2) Assess the productivity of U.S. commercial airports by accounting for joint production of desirable and undesirable outputs as well as develop a model for predicting their relative efficiency.
-5-
Assessment of Productive Efficiency of Airports
3) Estimate changes of airport productivity and sources of productivity growth during the study period, i.e., 2000 - 2003. 4) Analyze the impact of the inclusion of undesirable outputs on the productivity measurement and productivity growth. Since the research aims to provide timely information useful for managing airport in the modern era, the study period will span over recent years, i.e., 2000 – 2002 for research objective 1) and 2000 – 2003 for research objectives 2) to 4). Most data are expected to be from consolidated databases such as Federal Aviation Administration (FAA), Airports Council International (ACI), and Air Transport Research Society (ATRS). Supplement data may be collected directly from primary sources. 1.5 Research contributions The dissertation makes three major research contributions. First, rather than using mixed sample of airports, it assesses the productivity of airports operating in a similar market structure, i.e., multiple airport systems (MASs). Second, unlike previous airport productivity studies, this dissertation makes the first attempt to develop a more complete total factor productivity measure by also taking into account undesirable byproducts from airport operations, i.e., delays. Third, this dissertation also develops causal models for explaining variations in productive efficiency. The three contributions are accomplished by using recent panel data in two case studies. 1.6 Organization of the dissertation The dissertation is organized into eight chapters. The first chapter discusses the revolution of airport business, importance of airport productivity study, motivation of the
-6-
Assessment of Productive Efficiency of Airports
dissertation research as well as objectives and scope of the research. Chapter 2 reviews literature related to productivity study with emphasis in airport sector. Classification of productivity measures and a methodology to compute them are described. Since Data Envelopment Analysis is the most widely-used method for measuring airport productivity, its concept is briefly explained in this chapter to provide basic understanding of model development and its weaknesses. The DEA model will be used in one of the two case studies in this dissertation. Chapter 3 explains in details the proposed research methodology for assessing productivity of airports where joint production of desirable and undesirable outputs is taken into account. This chapter starts with the characterization of production possibility set, and illustration of output distance function and its modification, i.e., the directional output distance function which is the adopted model for analysis in the case study. The chapter also illustrates the computation of Malmquist and Luenberger productivity indexes and their components that are useful for explaining changes of productivity over time. Chapter 4 describes the first case study. The study is to assess productive efficiency of airports operating in MASs by using DEA as well as develop causal models for explaining variations in efficiency level. Chapter 5 presents and discusses the results from case study 1. Chapter 6 describes the second case study of 56 U.S. commercial airports. The contents cover modeling of airport operation, selection of inputs and output measures and characteristics of samples. The directional output distance function is applied to access the productivity of these airports. Chapter 7 presents and discusses the results. It provides contrast comparisons between with and without consideration of
-7-
Assessment of Productive Efficiency of Airports
undesirable outputs (delays). Other substantial results include productivity growth during 2000 – 2003, statistical analysis and scenario analysis. Important findings and insightful information are pointed out. Lastly, chapter 8 concludes the dissertation and suggests some potential areas for future research.
-8-
Assessment of Productive Efficiency of Airports
CHAPTER 2 LITERATURE REVIEW
This chapter rigorously reviews previous work directly related to airport productivity. The focus is on the classification of productivity measurement, the applied methodologies for measuring productivity, discussion of their advantageous and disadvantageous as well as the consideration for use. Summary of major findings and implications are also discussed. 2.1 Productivity measures In economics, productivity is defined as the amount of output per unit of input. In other words, the productivity measure is the ratio between output(s) and input(s). The definition, though very concise, is quite problematic to be applied in assessing productivity of airports. This is essentially due to the nature of airport operation which takes multiple inputs (such as labor and capital) for producing multiple outputs (such as movement of aircrafts, number of passengers and cargo throughput). Given various possible inputs and outputs, there are really many different ways of computing the productivity measure. Nevertheless, productivity measures can be categorized broadly into two groups of either partial factor or total (overall) factor productivity measures. 2.1.1 Partial Factor Productivity (PFP) measure Partial factor productivity (PFP) measures generally relate an airport’s output to a single input (factor). Labor productivity measures such as passengers per employee, aircraft movements per employee and ton landed per employee, are good examples. Table
-9-
Assessment of Productive Efficiency of Airports
2.1 summarizes more examples of PFP measures that have been used in airport business. A recent survey (Humphreys and Francis, 2002) revealed that the move towards privatization and commercialization has led to new performance measures being introduced to reflect the changing management goals. New measures fall into three categories, i.e., 1) financial measures to monitor commercial performance, 2) measures to meet the requirements of government regulators and 3) environmental measures. PFP measures have the advantage of being easy to compute, requiring only limited data and are easy to understand. As a result, many airport managers around the world usually adopt PFP measures to benchmark their performance (Francis, Humphreys and Fry, 2002; Humphreys and Francis, 2002). It is common to see such measures appear routinely in aviation trade publications (ACI 2002-2004; ATRS 2002 – 2003). Nevertheless, the measures can often be misleading when looking at the overall picture of the airport operation. For instance, it is possible to raise productivity in terms of one input, at the expense of reducing the productivity of other inputs. In the case of airports, which are fairly capital intensive, a partial productivity measure of labor productivity does not give a very clear picture of whether the performance of the institution is being improved (Abbott and Wu, 2002). Moreover, there are many possible PFP ratios, given multiple inputs and outputs of airport operation. There is usually a tradeoff among those measures. Airport may look better on one measure but can be worse on the others. As far as the overall assessment is concerned, it is preferable to use some form of overall (total) productivity measures that better shows the relation between all outputs and inputs.
- 10 -
Assessment of Productive Efficiency of Airports
Table 2.1 Examples of partial factor productivity measures in aviation sector
Scope of measure
Global performance of airport
Category
Profitability
Examples of performance measures
income per passenger rate of return on capital revenue to expenditure ratio profit per workload unit (WLU) cost per WLU (excluding depreciation and interest) operating cost per WLU capital cost per WLU labor cost per WLU aeronautical cost per WLU total revenue per WLU aeronautical revenue as a share of total aeronautical revenue per WLU non-aeronautical revenue per WLU concession revenue per area value added per unit of capital costs WLU per unit of net asset value total revenue per unit of net asset value WLU per employee revenue per employee value added per employee passengers/employee aircraft movements per runway aircraft movements per length of runway aircraft movements per hourly capacity passenger per aircraft movement service time for check-in time to reclaim baggage gate utilization rates passengers per terminal area baggage handled per unit of time baggage service reliability over time distances to reach departure gates crowding (passenger density) variability in service times passenger service ratings average time required to deliver freight at cargo terminal prior to aircraft departure theft and breakage rates index of aeronautical charges index of non-aeronautical charges aircraft turn-around times
Cost-efficiency
Cost-effectiveness (revenue earning)
Partial productivity measures
Capital productivity Labor productivity
Performance of particular processes
Runways
Passenger processing
Baggage handling Customer service Passengers
Cargo Airlines
Note: A “workload unit (WLU)” is equal to one passenger or 100 kilogram of cargo. Source: Hooper and Hensher (1997); Francis, Humphreys and Fry (2002); Humphreys and Francis (2002); Oum, Yu and Fu (2004).
- 11 -
Assessment of Productive Efficiency of Airports
2.1.2
Total Factor Productivity (TFP) measure In early 1990s, the literature on performance measurement for airports was
focused on the use of partial measures that yield an incomplete representation of the important relationship between multiple inputs and outputs. The lack of published research on overall measures of performance places a limit on our understanding of productive processes in the airport sector (Hooper and Hensher, 1997). As partial factor productivity measure indicates, performance has many dimensions. The growing literature on measuring the performance of airports is addressing the limitations of PFP measures in capturing all of those dimensions. A common way to deal with the problem of too many PFP measures is to derive an aggregate measure that takes into account all significant inputs and outputs simultaneously. Such measure is often called “Total Factor Productivity (TFP)” measure. Such overall TFP measure is useful for managers who are assessing the global productivity of an airport. It considers that different airports face different economic conditions and therefore may use input factors in varying proportions. For example, an airport that exhibits low labor productivity may not necessarily be inefficient from an overall perspective; it may merely be substituting capital with labor to take advantage of a wage rate (Nyshadham and Rao, 2000). TFP based measures have recently received increased attention in air transportation research and become a preferred measure. See for example Gillen and Lall (1997, 1998); Hooper and Hensher (1997); Oum and Yu (2004); Pathomsiri, Haghani, Dresner and Windle (2006a); Pels, Nijkamp and Rietveld (2001, 2003); Windle and Dresner (1992); and Yoshida and Fujimoto (2004). Since the TFP
- 12 -
Assessment of Productive Efficiency of Airports
measure is more suitable than PFP measures for assessing the productivity of airports, the subsequent review will focus on the methodology to derive the TFP measure. 2.2 Methodology for computing TFP measure There are several methods for deriving the TFP measure. The methods generally fall into two broad categories i.e., parametric and non-parametric approaches. Each approach has advantages and disadvantages. Their applicability usually depends on the availability of data. In some cases, both approaches are used to obtain complementary results (Pels, Nijkamp and Rietveld, 2001, 2003) or confirm the conclusions (Yoshida and Fujimoto, 2004). 2.2.1 Parametric approach Conceptually speaking, the parametric approach works in three major steps, i.e., 1) Transforming inputs into a common unit by assigning appropriate weights to individual inputs 2) Transforming outputs into a common unit by assigning appropriate weights to individual outputs so that an aggregate output can be computed and 3) Given a priori production function which represents logical relationship between the composite output in 2) and various transformed inputs in 1), estimate a set of parameters associated with individual transformed inputs. The results will give an estimated production function of airport operation explaining the transformation of inputs into outputs. With this function, it is possible to estimate the probable output level for a given set of inputs. Whenever the actual output is below the probable level, an airport is not being operated efficiently. In addition, by - 13 -
Assessment of Productive Efficiency of Airports
assuming that airports in the sample are similar; the productivity of airports can be benchmarked by comparing the difference between the actual output and the probable level. The further from the probable level means the less efficient operation. There are two major issues involved when one decides to use the parametric approach. First, what are the appropriate weights for transforming inputs and outputs? Second, what is the suitable functional form? Regarding the first question, Hooper and Hensher (1997) argue that the appropriate input weights should be the cost shares which represent the contributions of each input to costs. They also suggested that the output weights be the cost elasticities as long as they are readily available from prior research. However, in the most of empirical studies the absence of such elasticities has led to the use of revenue shares as proxies. Nyshadham and Rao (2000) have also adopted cost and revenue share respectively as input and output weights for their productivity assessment of 25 European airports. Hooper and Hensher (1997) commented that the use of prices as output proxies implicitly presumes that the airport is pricing efficiently but, since monopoly pricing is an issue of concern, it is problematic to derive an output measure from income. Indeed better measures for output quantity would have been landings for aeronautical output and passenger plus meeter-greeter throughput and the volume of cargo handled for nonaeronautical output. As for the second issue, the choice of a priori production function is rather subjective; and its suitability is usually based on the goodness-of-fit. Martin-Cejas (2002) estimates a deterministic cost frontier using translog function to assess the productive efficiency of 31 Spanish airports during 1996 – 1997. Pels, Nijkamp and Rietveld (2001)
- 14 -
Assessment of Productive Efficiency of Airports
estimate two stochastic production frontiers in their productivity study of 34 European airports during 1995 – 1997. The first function has number of passengers as the dependent (output) variable. The second function aims to explain the number of aircraft movements. Both of them are translog function. Based on the same dataset, their subsequent publication (Pels, Nijkamp and Rietveld, 2003) also estimate two stochastic production frontiers with the same two dependent (output) variables, but with different set of explanatory variables. The literature review indicates that translog is the most widely -used function in airport productivity studies. Although there are issues on weights and selection of production form, the parametric does have some advantages over the non-parametric approach. First of all, it can both measure and explain inefficiency simultaneously. Second, the parametric method allows for statistical testing of the presence of a deviation from the efficient frontier and returns to scale. Table 2.2 summarizes previous studies that used a parametric approach. It can be seen that there are very few studies. Availability of cost and revenue sharing data seem to be a big hurdle that limits the applicability of this approach. Many researchers therefore have resorted to an alternative approach, i.e., nonparametric.
- 15 -
Assessment of Productive Efficiency of Airports
Table 2.2 List of publications on airport productivity studies by parametric approach Author(s) Pels, Nijkamp and Rietveld (2001) Productivity model Year: 1995-1997 Stochastic (pooled cross-section production time series) frontier (SPF) Size: 34 European airports Sample Functional form
translog Air transport movements (ATM) = f{constant, number of runways, number of aircraft parking positions at the terminal, number of remote aircraft parking positions} Number of passengers (PAX) = f{constant, number of baggage claim units, number of aircraft parking positions at the terminal, number of remote aircraft parking positions} translog Total cost (TC) = f{unit of traffic transported, labor price, capital price}
Remark - Compute the most productive scale size (mpss) which represents the maximum productivity for any given input-output combination (Banker, 1984).
Martin-Cejas (2002)
Pels, Nijkamp and Rietveld (2003)
Year: 1996 – 1997 (pooled cross-section time series) Size: 31 Spanish airports Year: 1995-1997 (pooled cross-section time series) Size: 34 European airports
Deterministic cost frontier
- Unit of traffic transported (UT) = number of passengers + (kilograms of freight/100) - All variables (except dummies) are standardized around mean. - Treat number of runways as a fixed factor - Estimate also the DEA model (see Table 2.4 for the same authors)
Stochastic production frontier (SPF) and inefficiency model
translog Air transport movements (ATM) = f{constant, year dummy, airport area, number of runways, number of aircraft parking positions at the terminal, number of remote aircraft parking positions} ATM Inefficiency = f{slot coordination dummy, time restriction dummy} Air passenger movements (APM) = f{constant, , year dummy, predicted ATM, number of check-in desks, number of baggage claim units } APM Inefficiency = f{ constant, time restriction dummy, average airlines’ load factor}
- 16 -
Assessment of Productive Efficiency of Airports
2.2.2
Non-parametric approach The key characteristic of non-parametric approach is that it does not need to
specify a priori production function. No parameter needs to be estimated. Among other methods, index number and Data Envelopment Analysis (DEA) are the most popular in previous airport productivity studies. The index number method works similarly to the first two steps of parametric approach. Each input (output) needs to be assigned an appropriate weight so that individual inputs (outputs) are transformed into the same unit of measurement. Thus a weighted aggregate input (output) can be computed. The resulting aggregate input/output are called input/output indexes. By definition the total productivity index is the ratio of the weighted aggregate output index to a weighted aggregate input index. The higher value of TFP indicates higher efficiency. Thus the TFP measure can be used to rank performance of airports. Since the method involves weights, the discussion of weight issue in parametric approach is applicable here. In their productivity study of four Australian airports during 1989 – 1992, Hooper and Hensher (1997) use cost and revenue shares respectively as associated weights to inputs and outputs and obtain aggregate input and output indexes. Similarly, Nyshadham and Rao (2000) also use cost and revenue shares in their productivity study of 25 European airports. Other studies that adopted index number approach to compute TFP measure include Oum, Yu and Fu (2003), Oum and Yu 2004Yoshida (2004), Yoshida (2004), as well as Yoshida and Fujimoto (2004). Table 2.3 summarizes publications on airport productivity studies using index number method.
- 17 -
Assessment of Productive Efficiency of Airports
Table 2.3 List of publications on airport productivity studies using index number method Author(s) Hooper and Hensher (1997) Sample Year Year: 1988- 9, 1991-2 Size: 6 Australian airports Model Multilateral translog index number of TFP Input 1. capital stock 2. labor expenditures 3. other costs Output 1. deflated aeronautical revenues 2. deflated nonaeronautical revenues Remark - Use share of revenues as weights to compute output index - Estimate two regression models for estimating output-adjusted TFP. One model regressed TFP with output index (composite of cargo tonnages, movements, passengers, employers and labor costs). The other adds airport dummy variable. - A work load unit is defined as either one passenger or 100 kilograms of cargo. - Use percentage share of the revenue as weights to compute output index - Use percentage share of cost as weights to compute input index - Compute Spearman rank correlation between TFP and several PFP measures; then estimate a regression model for explaining TFP by PFP measures.
Nyshadham and Rao (2000)
Year: 1995 Size: 25 European airports
Multilateral translog index number of TFP
1. operating cost per work load unit 2. capital cost per work load unit 3. other costs per work load unit
1. aeronautical revenue per work load unit (WLU) 2. non-aeronautical revenue per work load unit (WLU)
- 18 -
Assessment of Productive Efficiency of Airports
Table 2.3 List of publications on airport productivity studies using index number method (Continued) Author(s) Oum, Yu and Fu (2003) Sample Year Year: 1999 Size: 50 major airports in AsiaPacific, Europe and North America Model Endogenous Weighted (EW) TFP input and output index numbers Input 1. number of fulltime equivalent employees who work directly for an airport operator 2. number of runways 3. number of gates 4. other costs l 1. runway length 2. terminal area Output 1. aircraft movements 2. number of passengers 3. cargo throughput 4. non-aeronautical revenues 1. aircraft movements 2. number of passengers 3. cargo throughput 1. aircraft movements 2. number of passengers 3. cargo throughputs - Use DEA efficiency score as truncated dependent variable and estimate a Tobit regression model to check inefficiency of regional airports and airports operated in 1990s. - Use EW-TFP index number as a dependent variable and estimate a regression model to check inefficiency of regional airports and airports operated in 1990s. Remark - Estimate a regression model for explaining the variation in TFP: TFP = f{constant, Asia-pacific dummy, airport size, % international passengers, % of aeronautical revenues}
Yoshida (2004)
Year: 2000 Size: 30 Japanese airports
Yoshida and Year: 2000 Fujimoto Size: 67 Japanese (2004) airports
Endogenous weighted (EW) TFP input and output index numbers DEA-InputCRS DEA-InputVRS Endogenous weighted (EW) TFP index number
1. runway length 2. terminal area 3. number of employees in the terminal 4. average access cost
- 19 -
Assessment of Productive Efficiency of Airports
Table 2.3 List of publications on airport productivity studies using index number method (Continued) Author(s) Oum and Yu (2004) Sample Year Year: 2000 – 2001 Size: 76 airports in Asia Pacific, Europe and North America Model Variable Factor Productivity (VFP – input index number) Input 1. number of fulltime equivalent employees who works directly for an airport operator 2. other costs Output 1. aircraft movements 2. number of passengers 3. cargo throughput 4. non-aeronautical revenues Remark - Estimate a regression model for explaining the variation in VFP: VFP = f{constant, airport size, % international passengers, % cargo traffic, capacity constraints, % of non-aeronautical revenues, outsourcing dummy}
- 20 -
Assessment of Productive Efficiency of Airports
One of the difficulties in using this method is that it requires a complete set of prices and quantity data. In many cases, these data are not available. Due to data limitations Hooper and Hensher (1997) were restricted to a few years during the early 1990s for four airports, so the study essentially presented a very limited indication of the performance of the Australian airport systems. Another weakness in this approach is the use of total revenue of the airports as an indicator of output. It is justifiable as long as prices, and therefore revenue, are not a reflection of the degree of market power of the institution considered. In the case of airports this might be the case and so it is preferable to use a total factor productivity valuation approach that does not depend upon prices that might be distorted by market imperfections (Abbott and Wu, 2002). Martin and Roman (2001) argued that some financial measures can be misleading indicators, as a consequence of the relative market power that might exist. Monopolistic airports might be able to make substantial profits even if they were inefficient. More importantly, prices are applicable for marketed outputs only, but it is difficult to calculate for non-marketed outputs, such as delays, noise and other externalities. During the past decade, aviation researchers have resorted to use an alternative non-parametric method which gets away from weight issue i.e., Data Envelopment Analysis (DEA). DEA is perhaps the most widely used method for assessing productivity of airport, regardless of approaches. DEA may be a true non-parametric method. It does not require any weights. It does not need to assume a production function. Instead, it builds an empirical piecewise linear production function from sample data. The only required data are the quantity of inputs and outputs. This is perfectly applicable in airport context where the breakdown between revenue and average prices for freight cargo and passenger
- 21 -
Assessment of Productive Efficiency of Airports
traffic are not made available, but output and input volume figures are. Therefore, DEA is the ideal method for estimating TFP measure. During the past decade, DEA seems to be the prevailing method used in assessing airport productivity. Table 2.4 lists publications that adopted DEA as an analytical method. Since DEA is the prevailing method in airport productivity study, the next section will be devoted to the review of DEA. It should be noted that the review is by no means exhaustive, but is focused on model development and some important features. The publications in Table 2.4 will also be referred to more. For more theoretical insights and applications about DEA, a good number of textbooks can be consulted. See for examples in Charnes, Cooper, Lewin and Seiford (1994); Cooper, Seiford and Tone (2000); Zhu (2003); Cooper, Seiford and Zhu (2004); Ray (2004); Cook and Zhu (2005).
- 22 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA Author(s) Gillen and Lall (1997) Sample Year Year: 1989 – 1993 Size: 23 of the top U.S. airports Model DEAOutput-CRS Input I. Terminal services 1. number of runways 2. number of gates 3. terminal area 4. number of employees 5. number of baggage collection belts 6. number of public parking spaces II. Movements 1. airport area 2. number of runways 3. runway area 4. number of employees Output I. Terminal services 1. number of passengers 2. pounds of cargo Remark - Estimate two Tobit regression models for explaining terminal and movements efficiency
DEAOutput-VRS
II. Movements 1. air carrier movements 2. commuter movements
- 23 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Gillen and Lall (1998) Sample Year Year: 1989 – 1993 Size: 22 of the top U.S. airports Model DEAOutput-CRS Input I. Terminal services 1. number of runways 2. number of gates 3. terminal area 4. number of employees 5. number of baggage collection belts 6. number of public parking spaces II. Movements 1. airport area 2. number of runways 3. runway area 4. number of employees Output I. Terminal services 1. number of passengers 2. pounds of cargo Remark - Compute Malmquist TFP by component
DEAOutput-VRS
II. Movements 1. air carrier movements 2. commuter movements
- 24 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) MurilloMelchor (1999) Sample Year Year: 1992 – 1994 Size: 33 Spanish civil airports under management of AENA (Spanish Airports and Air Transport) Model DEAInput-CRS DEAInput-VRS Input 1. number of workers 2. accumulated capital stock approximated by the amortization estimated in constant value 3. intermediate expenses 1. employment 2. capital stock 3. non labor cost 4. capital cost 5. changes in gross domestic product (GDP) Output 1. number of passengers Remark - Compute Malmquist index for individual pair of years
Parker (1999)
Salazar De la Cruz (1999)
Year: financial years (as of March 31) from 1988/89 – 1996/97 Size: 22 UK airports, including all of British Airports Authority (BAA)’s major airports Year: 1993 – 1995 Size: 16 main Spanish airports serviced mixed domestic and international passenger traffic; range 1 – 20 million passengers
DEAInput-VRS
1. number of passenger 2. cargo and mail
- Compute mean efficiency rating over 88/89 – 96/97 and use it to rank 22 airports before and after privatization.
DEAOutput-CRS
1. total economic cost e.g., cost for annual operations, the current costs and the internal interest on the net assets
1. annual passengers 2. total returns 3. returns on infrastructure services 4. operative returns 5. final returns
- Empirically, observe the extent to which input and output contribute to the change in efficiency by visualizing from graph
- 25 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Sarkis (2000) Sample Year Year: 1990 – 1994 Size: 44 major U.S airports Model DEAInput-CRS DEAInput-VRS Input 1. operating costs 2. number of airport employees 3. number of gates 4. number of runways Output 1. operational revenue 2. number of passengers 3. aircraft movements 4. general aviation movements 5. amount of cargo shipped Remark - Include the following variants 1. Simple cross-efficiency (SXEF) (Doyle and Green, 1994) 2. Aggressive cross-efficiency (AXEF) (Doyle and Green, 1994) 3. Ranked efficiency (RCCR) (Anderson and Peterson, 1993) 4. Radii of classification ranking (GTR) (Rousseau and Semple, 1995) - Perform nonparametric MannWhitney U-test to test the differences of efficiency scores between hub/non-hub, MAS/SAS, and snowbelt/non-snowbelt
- 26 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Adler and Berechman (2001) Sample Year Year: 1996 Size: 26 airports in Western Europe, North America and the Far East Model DEAInput-VRS (dual formulation) Input 1. peak short and medium haul charges 2. inversed number of passenger terminal 3. inversed number of runways 4. distance to the city center 5. minimum connecting time international – international 6. average delay per aircraft movement in minutes Output Three principal components derived from the following five measures of service quality from airlines’ perspective 1. suitability 2. operational reliability and convenience 3. cost of using airport 4. overall satisfaction and airport quality 5. factual questions with respect to the wave system and demand Remark - Survey airport quality of service from airlines rating 14 questions on Likert scale; and due to excessive number of total variables (inputs + outputs), the authors apply Principal Component Analysis (PCA) statistical method to reduce the total number inputs/outputs - Apply super-efficient DEA model (Anderson and Peterson, 1993) to fully rank the airports and report unbound results (infeasibility in primal) for some airports.
- 27 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Sample Year Fernandes Year: 1998 and Pacheco Size: 35 Brazilian (2001) domestic airports Model Input DEA-Non1. mean number of oriented-CRS employees 2. payroll expenditure, including direct and indirect benefits 3. operating expenditures 4. apron area 5. departure lounge area 6. number of check-in counters 7. length of curb frontage 8. number of vehicle parking spaces 9. baggage claim area DEA1. labor expense Output-VRS 2. capital expense, DEAincluding Output-CRS amortization of fixed assets 3. material expense Output 1. number of passengers, 2. cargo plus mail, 3. operating revenues 4. commercial revenues 5. other revenues Remark
Martin and Roman (2001)
Year: 1997 Size: 37 Spanish airports
1. air traffic movements 2. number of passengers 3. tonnage of cargo
- Compute technical efficiency by using reciprocal of efficiency score obtained from solving DEA - Compute scale efficiency - Interpret target output and input slack
- 28 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Pels, Nijkamp and Rietveld (2001) Sample Year Year: 1995 – 1997 (pooled crosssection time series) Size: 34 European airports Model Air transport movements (ATM) model DEAInput-CRS DEAInput-VRS Input 1. airport area 2. runway length 3. number of aircraft parking positions at the terminal 4. number of remote aircraft parking positions Output 1. Air transport movements (ATM) Remark - Estimate also the stochastic production frontier (see Table 2.2 for the same authors)
1. terminal area Air passenger 2. number of movements aircraft parking (APM) model positions at the terminal 3. number of remote aircraft parking positions 4. number of check-in desks 5. number of baggage claim units
1. Air passenger movements (APM)
- 29 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Abbott and Wu (2002) Sample Year Year: 1989/1990 to 1999/2000 Size: 12 main Australian airports Year: 1998/99 Size: 12 main Australian and 13 other international airports Fernandes Year: 1998 and Pacheco Size: 33 Brazilian (2002) major domestic airports Model DEAInput-CRS Input 1. number of staffs 2. capital stock 3. runway length 1. number of staffs 2. runway length 3. land area 4. number of aircraft standing areas 1. area of apron 2. area of departure lounge 3. number of check-in counters 4. length of frontage curb 5. number of parking spaces 6. baggage claim area Output 1. number of passengers 2. freight cargo in tons Remark - Compute Malmquist total factor productivity (TFP) index, - Estimate Tobit regression for explaining variation in Malmquist TFP
DEAInput-CRS
DEAOutput-VRS
1. number of passengers 2. freight cargo in tons 1. domestic passengers
- Analyze inefficiency level, slacks, potential number of domestic passengers in comparison to demand forecast
- 30 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Bazargan and Vasigh (2003) Sample Year Year: 1996 – 2000 Size: Top 45 U.S. airports, 15 each from large, medium and small hubs (by FAA’s definition) during the study period Model DEAInput-CRS Input 1. operating expenses 2. non-operating expenses 3. number of runways 4. number of gates including gates with jet ways and other non jet- way gates 1. average number of employees 2. payroll, including direct and indirect benefits 3. operating expenses 1. land area 2. number of runways 3. area of runways Output 1. number of passengers 2. air carrier operations 3. number of commuters, GA and military 4. aeronautical revenues 5. non-aeronautical revenues 6. percentage of ontime operations 1. domestic passengers 2. cargo plus mail 3. operating revenues 4. commercial revenues 5. other revenues, 1. aircraft movements 2. number of passengers Remark - Achieve a full ranking of all airports by introducing a virtual super efficient airport with existing airports so that there will be only one efficient airport. Its inputs and outputs are as follows: - Test the difference among three hub types by non-parametric Kruskal-Wallis test.
Pacheco and Year: 1998 Fernandes Size: 35 Brazilian (2003) domestic airports
DEAInput-VRS
- Use efficient scores from Fernandes and Pacheco (2002) as physical efficiency score and management efficiency score from this study to create Boston Consultancy Group (BCG) matrix - Perform paired-sample t-test to see if there is significant difference in efficiency scores before and after September-11. - Compute target inputs and outputs for inefficient airports
Pathomsiri and Haghani (2004)
Year: 2000, 2002 Size: 63 airports in multiple airport system worldwide
DEAOutput-VRS
- 31 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Pels, Nijkamp and Rietveld (2003) Sample Year Year: 1995-1997 (pooled crosssection time series) Size: 34 European airports Model Air transport movements (ATM) model DEAInput-CRS DEAInput-VRS Input 1. airport area 2. number of runways 3. number of aircraft parking positions at the terminal 4. number of remote aircraft parking positions Output 1. Air transport movements (ATM) Remark - Estimate also the stochastic production frontier (see Table 2.2 for the same authors) - Number of runways is treated as a fixed factor and adopted Banker and Morey (1986) formulation.
Air passenger 1. ATM movements 2. number of (APM) model check-in desks 3. number of baggage claim units
1. Air passenger movements (APM)
- 32 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Sarkis and Talluri (2004) Sample Year Year: 1990 – 1994 Size: 44 major U.S. airports Model DEAInput-CRS Input 1. operational costs 2. number of airport employees 3. number of gates 4. number of runways Output 1. operational revenue 2. passengers 3. aircraft movements 4. number of general aviation movements 5. total cargo I. Financial performance 1. operating revenues 2. commercial revenues 3. other revenues II. Operating performance 1. passengers embarked plus disembarked 2. tonnage of cargo embarked plus disembarked Remark - Rank airports by mean crossefficiency scores (AXEF) (Doyle and Green, 1994) - Identify benchmarks by using the hierarchical clustering technique based on correlation coefficients of the columns in the crossefficiency matrix. The average linkage method is utilized to derive the clusters. Airports in each cluster have a benchmark.
Fernandes Year: 1998 and and Pacheco 2001 (2005) Size: 58 airports administered by the Brazilian Airport Infrastructure Enterprise, Infraero
DEAInput-VRS
1. payroll, including direct and indirect benefits 2. operating and other expenses 3. average number of employees
- 33 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Pathomsiri, Haghani and Schonfeld (2005) Pathomsiri, Haghani, Dresner and Windle (2006a) Sample Year Year: 2000 , 2002 Size: 72 airports in multiple airport system worldwide Year: 2000 - 2002 Size: 72 airports in multiple airport systems worldwide Model DEAOutput-VRS Input 1. land area 2. number of runways 3. area of runways 1. land area 2. number of runways 3. area of runways Output 1. aircraft movements 2. number of passengers 1. aircraft movements 2. number of passengers Remark - Use parametric and nonparametric statistical methods to test the difference of efficiency scores before and after September 11 - Estimate Tobit regression model to explain variation in airport productivity
DEAOutput-VRS
- 34 -
Assessment of Productive Efficiency of Airports
2.3 2.3.1
Data Envelopment Analysis (DEA) Background DEA is a relatively new “data oriented” approach for evaluating performance of a
set of peer entities called Decision Making Units (DMUs) which convert multiple inputs into multiple outputs. The story of DEA begins with Edwardo Rhodes’s Ph.D. dissertation research at Carnegie Mellon University. The research was to evaluate Program Follow Through – the educational program for disadvantaged students (mainly black and Hispanic) undertaken in U.S. public schools with support from the Federal Government. It was the challenge of estimating relative technical efficiency of the schools involving multiple outputs and inputs, without using the information on prices that resulted in the formulation of the CCR (Charnes, Cooper and Rhodes) ratio form of DEA and the first publication (Charnes, Cooper and Rhodes, 1978). The DEA models use the optimization method of mathematical programming to generalize the Farrell (1957) single-output/input technical efficiency measure to the multiple-output/multiple-input case. Thus DEA began as a new Management Science tool for technical-efficiency analyses of public sector DMUs (Charnes, Cooper, Lewin and Seiford, 1994). The definition of a DMU is generic and flexible (Cook and Zhu, 2005; Cooper, Seiford, Zhu, 2004). Since the introduction in 1978, researchers in a number of fields have quickly recognized its usefulness and applicability. In recent years, there have been a great variety of applications of DEA in evaluating the performances of many kinds of entities engaged in many different activities in many different contexts in many different countries (Cooper Seiford and Tone, 2000; Cook and Zhu, 2005; Cooper, Seiford and Zhu, 2004). Seiford (1996) provides a bibliography since its first publication in 1978 to - 35 -
Assessment of Productive Efficiency of Airports
1995. Some textbooks exclusively cover DEA (Cooper, Seiford and Tone, 2000; Zhu, 2003; Ray, 2004; Cooper, Seiford and Zhu, 2004; Cook and Zhu, 2005). From time to time journals publish special issues on DEA theory and applications (Haynes, Stough, and Shroff, 1990; Cooper, Seiford, Thanassoulis and Zanakis, 2004). Emrouznejad (2006) has maintained a website that describes a rich family of DEA models. DEA opened up possibilities for use in cases which have been resistant to other approaches because of the complex (often unknown) nature of the relationship between the multiple inputs and multiple outputs involving DMUs. DEA requires very few assumptions. It does not need a priori assumption on functional form. This has made DEA applications quickly pervasive. In transportation, DEA has been applied to assess productivity of several activities such as public transit (Kerstens, 1996; Pina and Torres, 2001; Boame, 2004; Boame and Obeng, 2005), railway (Coelli and Perelman, 1999), large-scale distribution systems (Ross and Droge, 2004), ports (Tongzon, 1995; Budria, Diaz-Armas, Navarro-Ibanez and Ravelo-Mesa, 1999; Tongzon, 2001; Itoh, 2002; Turner, Windle and Dresner, 2004), and airlines (Schefczyk, 1993; Scheraga, 2004; Pires Capobianco and Fernandes, 2004). DEA has become a useful analytical tool for productivity study and performance analysis during the past two decades. In airport sector, researchers started using DEA in the late 1990s. The early works include Gillen and Lall (1997, 1998); Murillo-Melchor (1999); and de la Cruz (1999). Recently, there are a good number of publications using DEA to assess productivity of airports in different regions. Table 2.4 summarizes DEA publications in airport sector. For each study, the Table describes author(s), sample characteristics, analysis period, type of applicable DEA model, as well as set of inputs and outputs. The remark in the last
- 36 -
Assessment of Productive Efficiency of Airports
column notes major extra work beyond application of DEA in those studies. For example, after solving DEA models, Gillen and Lall (1997); Pathomsiri, Haghani, Dresner and Windle (2006a) estimated Tobit regression models for explaining variation in output efficiency scores. Meanwhile, Gillen and Lall (1998) and Abbott and Wu (2001) compute Malmquist index to explain changes of total factor productivity over time. 2.3.2 Model development
Output Y
B
CB = efficient production frontier
A4 A8 B’
C’B’ = average production function
A2
Positive residual
A5
Negative residual
A3
A6 A7
C C’ O A1
Input X
Figure 2.1 Difference between DEA and regression concept DEA is a methodology directed to frontiers rather than central tendency. Suppose
that there is a set of hypothetical airports whose airside operation take only single input X (e.g., runway) and produces single output Y (e.g., aircraft movements). Their input and output measures are scatter plotted in Figure 2.1. If one were to use a regression model to
- 37 -
Assessment of Productive Efficiency of Airports
estimate the production function of this operation, the fitted line would be C’B’ which passes through the “cloud” of data points. This regression line basically explains the average production. For airports A1, A6, and A7 this production function gives negative residuals or overestimated production. Meanwhile the fitted line will give positive residuals or underestimated production for airports A2, A4 and A8. The residuals are the portion of production that results from other factors beyond this single output. Instead of fitting a line to the data, DEA tries to learn from airports that lie above the line (airports with positive residuals). These airports are outliers that provide a good benchmark meaning that for a given input X, there is no other airport producing more Y. An alternative line CB is therefore drawn to represent the maximum possible production function or efficient production frontier that encompasses all airports. Any airports on this frontier are regarded as efficient whereas other airports within the frontier are inefficient. The further an airport is from the frontier, the more inefficient it is. DEA determines the efficient production frontier by estimating the distance for individual airports. Figure 2.2 explains the mechanism. DEA checks each airport to find out whether it lies on the frontier. Consider airport A3 which is below the frontier CB. Denote a scalar multiplier ? to current output y2 for boosting the production to the maximum level at A9 on the frontier. A9 may be viewed as a virtual airport whose input and output levels are the linear combination of airports A2 and A8, i.e., (?1x1+ ?2x2, ?1y1, ?2y2). Intuitively, for efficient airports, i.e., A2 and A8, their multipliers equal to one because they do not need to boost the production any further. All other inefficient airports will have some value depending on how inefficient they are. In real application, the production consists of multiple inputs and
- 38 -
Assessment of Productive Efficiency of Airports
outputs, rather than single input/output as shown in Figure 2.2; therefore it would be impossible to visualize. In this case the multiplier ? can be estimated by solving the following LP problem:
Output Y
B
A8 (x2, y2)
A9 (?1x1+ ?2x2, ?1y1, ?2y2)
A4
A11 A2 (x1, y1) yo A10
?
A5 A3 (xo, yo) A6 A7
C O
A1 xo Figure 2.2 Determination of efficient production frontier
Input X
max ? k s.t.
k?K
?? ??
k
+ y km ? s m = ?y km , m = 1,..........., M , ? x kn + s n = x kn , n = 1,.............., N ,
(2.1)
k?K
k
? k ? 0, k = 1,.........., K
Where k , m and n represent index of airports (k = 1,2......, K ) , index of outputs
(m = 1,2......, M ) and index of inputs (n = 1,2......, N ) respectively. ?k is an intensity vector associated with each airport and has k elements. x kn , y km are quantity of input n
- 39 -
Assessment of Productive Efficiency of Airports
+ ? and output m of airport k respectively. s m and s n are output and input slacks
respectively. ? or the output efficiency score is a scalar by which the current output level has to be multiplied in order to reach the frontier. If an airport is on the frontier, solving this LP will result in an optimal objective function ? * k = 1 . In other words, it is
sufficiently productive and does not need to increase output. ? is bound by [1, ?). The efficiency score can be used as a TFP measure. The LP needs to be solved k times, each time for an individual airport. The LP formulation in (2.1) is known as “Output-Oriented with Constant Returnto-Scale Characterization DEA model”, or in short DEA-Output-CRS hereinafter. As the name implies, the formulation seeks to determine if an airport is on the frontier in the output direction, for a given level of inputs. The analysis provides an assessment of how efficiently the inputs are being utilized. The DEA-Output-CRS has been used by several researchers including Gillen and Lall (1997, 1998); Fernandes and Pacheco (2002); de la Cruz (1999); Martin and Roman (2001); Pathomsiri and Haghani (2004); Pathomsiri, Haghani and Schonfeld (2005); Pathomsiri, Haghani, Dresner and Windle (2006a). In fact, the inefficiency can be determined in other directions as well. Figure 2.2 shows another two possible directions. The first is in the direction of input, i.e., projecting A3 to the frontier at A10. If this is the case, the corresponding LP formulation is given in (2.2).
- 40 -
Assessment of Productive Efficiency of Airports
min ? k s.t.
k?K
?? ??
k
+ y km ? s m = y km , m = 1,..........., M , ? x kn + s n = ?x kn , n = 1,.............., N ,
(2.2)
k?K
k
? k ? 0, k = 1,.........., K
All notations are the same as defined above. ? k or the input efficiency score is a scalar by which the current input level has to be multiplied in order to reach the frontier. The LP formulation in (2.2) is known as “Input-Oriented with Constant Return-to-Scale Characterization DEA model”, or in short DEA-Input-CRS hereinafter. The model determines whether there is inefficiency in input, for a given level of output. If an airport is on the frontier, solving this LP will result in an optimal objective function ? k* = 1 . In other words, the current level of input is probable and does not need to be reduced. ? is bound by (0, 1]. The input efficiency score can be used as a TFP measure. The LP needs to be solved k times, each time for an individual airport. The DEA-Iutput-CRS has been used by several researchers including Abbott and Wu (2002); Adler and Berechman (2001); Bazargan and Vasigh (2003); Fernandes and Pacheco (2005); Murillo-Melchor (1999); Pacheco and Fernandes (2003); Parker (1999); Pels, Nijkamp and Rietveld (2001, 2003); Sarkis (2000); Sarkis and Talluri (2004); and Yoshida and Fujimoto (2004). In Figure 2.2 there is another direction which is the shortest possible distance, i.e., projecting A3 to the frontier at A11. In this case DEA does not care about direction. The LP formulation simultaneously expands the output and contracts inputs. The efficiency score indicates inefficiency level in both input and output. The model is called “Nonoriented with Constant Return-to-Scale Characterization DEA model”, or in short DEA-
- 41 -
Assessment of Productive Efficiency of Airports
Non-oriented-CRS. The model is rarely used in airport productivity studies since it is not practical to freely adjust inputs/outputs mix. An airport manager is unlikely to choose the combination of capital inputs (e.g., runway, taxiway, and terminal building) and passenger throughput. Either input or output may be not controllable. It is found that only Fernandes and Pacheco (2001) adopted the DEA-Non-Oriented-CRS to assess the productivity of 35 Brazilian airports. The formulation is not given here, but can be found in several textbooks including Zhu (2003), Ray (2004) and Cooper, Seiford and Zhu (2004). Regardless of the chosen orientation, there is no effect on the classification of efficient airports because the resulting efficient frontiers are identical. However, it does affect results regarding inefficient airports. Researchers have to justify the choice of orientation. Regarding the use of input orientation, Abbott and Wu (2002) justify by reasoning that “airports have fewer controls over outputs than they do over inputs. The volume of airline traffic is somewhat exogenous to the control of airports’ managers depending as it does mainly on the general level of economic activity, both in the host city and the Australian and international economies more generally.” Meanwhile, Pacheco and Fernandes (2003) justify that they were dealing with Brazilian airports of various sizes. Martin and Roman (2001) justify the use of output orientation in their assessment of Spanish airports by reasoning that “We think that once an airport has invested in the building of new runways or new terminals, it is difficult for managers to disinvest to save costs, therefore invalidating the input-orientation.” Meanwhile, Fernandes and Pacheco
- 42 -
Assessment of Productive Efficiency of Airports
(2002) argue that the main issue of their analysis is the potential output from organizations with various sizes. In addition to the CRS frontier type, DEA can be carried out under the assumptions of variable returns to scale by introducing a scale constraint into the model. In VRS frontier type DMUs are not penalized for operating at a non-optimal scale (Banker, 1984; Banker and Thrall, 1992). Ganley and Cubbin (1992) consider the CRS frontier type as the long-term view as opposed to short-term view for VRS frontier. Martin and Roman (2001) argue that due to the existence of different scale airports in Spain, a VRS frontier should be used. Nonetheless they estimate also the CRS model. Parker (1999) argues that given the variation in the size of the airports in his dataset, VRS is the more realistic assumption than CRS. Murillo-Melchor (1999) however, argues that scale efficiency requires that the production size corresponds to the long-run. For this reason, this efficiency is assessed with respect to the technology of a long-run model i.e., constant returns to scale. Table 2.5 summarizes some important DEA models that have been used in previous airport productivity studies. The efficient targets in the last row compute the probable levels of input and outputs for those inefficient airports.
2.4 Discussion
The data availability on prices tends to limit the applicability of parametric approach. Literature review clearly indicates that non-parametric approach such as index number and DEA are more widely used by researchers. During the past decade, many researchers have adopted DEA to assess productivity of airports in different regions around the world.
- 43 -
Assessment of Productive Efficiency of Airports
Table 2.5 Summary of DEA models
Frontier type CRS
min ? s.t.
k?K
Input-oriented
Output-oriented
max ? s.t.
+ y km ? s m = y km , m = 1,..........., M , k?K
?? ??
k
?? ??
k
+ = ?y km , m = 1,..........., M , y km ? s m ? = x kn , n = 1,.............., N , x kn + s n
k?K
k
x kn + s = ?x kn , n = 1,.............., N ,
? n
k?K
k
? k ? 0, k = 1,.........., K
VRS Add
? k ? 0, k = 1,.........., K
k?K
?? ?? ??
k
=1
NIRS
Add
k?K
k
?1
NDRS
Add
k?K
k
?1
Efficient target
? ˆ kn = ?x kn ? s n x , n = 1,........., N + ˆ km = y km + s m y , m = 1,........., M
? ˆ kn = x kn ? s n x , n = 1,........., N + ˆ km = ?y km + s m y , m = 1,.........., M
Based on the review, it can be observed that previous studies assess productivity by only looking at desirable outputs such as passengers, aircraft movements, cargo and revenues. Inherently in the nature of airport operations, there are always undesirable byproducts being produced such as delays, mishandled baggage and accidents. In addition, airport operations also create externalities, notably noise and pollution. These byproducts may also be considered to be airport outputs, although undesirable, and they
- 44 -
Assessment of Productive Efficiency of Airports
are major concerns of the aviation industry. All airport stakeholders wish to minimize these undesirable outputs, or at least keep them at acceptable levels. Accounting for undesirable outputs in decision making is therefore, a goal of managers in the aviation industry. However, none of them considers joint production of desirable and undesirable outputs in the assessment, except Yu (2004). In that study, the author considered aircraft noise (in 1000 New Taiwan dollars) as the lone undesirable output. There are several limitations in this work. It is not clear how noise is measured and transformed into monetary unit. The sample size of 14 Taiwanese airports is too small when compare to the number of inputs (5) and outputs (3) measures. In DEA framework, the sample size should be much greater than number of inputs times outputs (Cooper, Seiford and Tone, 2000: page 252). Otherwise the discriminatory power will be deteriorated. That is the reason why Yu (2004) reports many efficient airports. Furthermore, other major undesirable outputs are excluded. In the US, delays are a major concern of air services. BTS (2006) routinely records on-time performance of flights and delays. In Europe, the situation about air traffic control is getting worse. In 2000, around 30% of flights experienced delays more than 15 minutes and air traffic control was the most important causes of delays (Martin and Roman, 2001). In fact some researchers have discussed about undesirable outputs but did not address them. In their ad-hoc Tobit regression models, Gillen and Lall (1997) noted that greater noise restriction tend to lower movement performance. To clean up noise, airports need to trade their movements low. Some researchers have pointed out the association between efficient airports and delays. Salazar de la Cruz (1999) observed that those airports that define the frontier show very high level of utilization, confirmed by further
- 45 -
Assessment of Productive Efficiency of Airports
congestion problems and expansion works. Furthermore, he suggests that it will be more prudent to consider the inefficiencies associated with level of usage, climate conditions, economy of design, construction or quality level including delays, etc. Especially, the necessity to consider the impact of capacity and delays jointly requires the introduction of specific behavioral models for each airport, information for which is not easily available. Based on previous work (Adler and Berechman, 2001; Bazargan and Vasigh, 2003; Fernandes and Pacheco, 2002; Gillen and Lall, 1997, 1998; Martin and Roman, 2001; Pacheco and Fernandes 2003; Pathomsiri and Haghani, 2004; Pathomsiri, Haghani and Schonfeld, 2005; Pathomsiri, Haghani, Dresner and Windle, 2006a; Pel, Nijkamp and Rietveld, 2001, 2003; Sarkis, 2000; Sarkis and Tulluri, 2004), DEA results tend to identify busy airports as efficient. Frequently, these efficient airports are also congested. It may be that one airport creates greater numbers of delayed flights than another, but produces the same level of desirable outputs per unit of input. Unless delays are taken into account, both airports would show the same productivity level. Consideration of undesirable outputs is not as straightforward as desirable outputs, but quite problematic. In DEA literature, there is a general guideline for distinguishing between input and output variables. If the lower level of measure is better, it should be classified as an input; but if the higher quantity is desirable, that variable is classified as an output. This is not true in airport operation where the higher quantity of undesirable outputs such as noise, pollution, delays, and accident are not desirable. Moreover, these outputs are not inputs in airport operation either. Even so, Adler and Berechman (2001) consider delay as an input.
- 46 -
Assessment of Productive Efficiency of Airports
Estimation is also an issue. The limitation lies in the mathematical mechanism for determining if an airport is on the efficient frontier. Using DEA-Output orientation would seek to maximize the expansion of all outputs, rather than maximize only the desirable outputs and minimize the undesirable. In reality, an airport manager never wishes to expand both number of passengers and delays simultaneously. The ad-hoc DEA is not applicable either (Färe and Grosskopf, 2004a, 2004b; Seiford and Zhu, 2002, 2005). The issue will need a special mathematical formulation. As a result, it is a very challenging task to analyze airport productivity where there is joint production of desirable and undesirable outputs. Accounting for both types of outputs should provide a more complete measure of airport productivity. Furthermore, consideration of undesirable outputs such as reduction in delays may lead to different evaluation which in turn results in different management policy. For example, it could affect the time when expansions and new facilities must be operated. This pioneer research will address this problem and attempt to point out the effects of joint consideration of desirable and undesirable outputs. It should be noted that all studies that are categorized as TFP are termed as such because they consider more than one input and output. It is virtually impossible to consider all of the factors in a productivity study. Since this dissertation also considers multiple important inputs and outputs, it can also be considered as a TFP study.
- 47 -
Assessment of Productive Efficiency of Airports
CHAPTER 3 RESEARCH METHODOLOGY
This chapter presents the methodology that will be used to assess the productive efficiency of airport operation with joint consideration of both desirable and undesirable outputs. The methodology is based on the production theory from economics discipline. The chapter starts off with the characterization of production technology in order to represent the relationship between input and output measures. The traditional axiom of production theory i.e., the distance function is then introduced. Taking this as a building block an optimization model, called the directional output distance function, is developed. It is of a non-parametric type applicable for modeling production system with multiple inputs and outputs and provides measures of performance without appealing to prices. Finally, the productivity index number is devised for use in analyzing productivity changes over time.
3.1 Characterization of production possibility set
In environmental economics one often wishes to distinguish between desirable
+ + ) outputs. In the production context the former output ( y ? RM ) and undesirable (b ? R J
is typically a marketable goods and the latter is often not marketed, but rather a byproduct which may have deleterious effects on the environment or human health, and therefore its disposal is often subject to regulation. As a result, it should be useful to explicitly model the effects of producing both types of outputs, taking into account their characteristics and their interactions (Färe and Grosskopf, 2004b).
- 48 -
Assessment of Productive Efficiency of Airports
Let’s consider a production process that desirable and undesirable outputs may be jointly produced, i.e., b is a byproduct of the production of y. Here, the application is an airport operation that processes throughputs of passengers, aircraft movements and cargo by using its infrastructure such as land, runway, and terminal. In this case, the desirable or marketable outputs are number of passengers, movements and amount of cargo transported. There is also undesirable byproduct, i.e., delays (others may include noise and pollution). The basic problem is that given technology, producing these throughputs means simultaneously producing delays even though their production is undesirable. The production technology T describes the possible transformations of inputs
+ + ( x ? RN ) into ( y ? RM ) and undesirable (b ? R J+ ) outputs. The production possibility
set is defined as a set of desirable and undesirable outputs that can be produced from a given level of inputs. This set is represented by:
P ( x) = {( y, b) : x can produce ( y, b)}
3.2 Output distance function
(3.1)
For the sake of illustration, assume that airport operation produces only two desirable outputs y1 and y 2 (which may be passengers and aircraft movements) from a given input vector. Figure 3.1 shows a hypothetical output possibility set. Note that the true shape of the set is unknown. The frontier of the set is defined as the output vector that cannot be increased by a scalar multiple without leaving the set. In the Figure, the frontier represents efficient combinations of outputs y1 and y 2 . However, not all airport operations are efficient; therefore there must be an inefficient airport that lies below the efficient frontier. The basic idea in distinguishing efficient airports from inefficient ones
- 49 -
Assessment of Productive Efficiency of Airports
is to determine how far the current operations are from the frontier. In Figure 3.1, airports A and B are right on the frontier; hence obviously are efficient airports. Meanwhile, airport C is away from the frontier by the distance BC or AC depending on the direction of measurement. As a result, airport C is not efficient.
y2 B A
E g
#
C
#
D
P(x)
0
O
Figure 3.1 Output possibility set and distance functions
y1
Shephard’s output distance function (Shephard, 1970) can be used to determine how far an airport is from the frontier. It is defined as the ratio of actual output to maximum potential output and equals to the reciprocal of Farrell’s output technical efficiency measure (Farrell, 1957). For any airport, the Shephard’s output distance function is: v Do ( x , y )
= inf{? : ( x, y ) ? P( x)} ? (3.2)
- 50 -
Assessment of Productive Efficiency of Airports
?1
y ? ? = ?sup{? : ( x, ) ? P( x)}? ? ? ?
(3.3)
r OC . This is a measure of how far the For airport C in Figure 3.1, Dc ( x, y ) = OA operation of airport C is from the frontier. The function is equal to one for all efficient airports and less than one for inefficient ones. The higher value of distance function indicates higher operational efficiency. The reciprocal of the output distance function ( OA ) or the Farrell measure gives the maximum proportional expansion in all outputs OC
that is feasible given inputs. The distance function completely characterizes the v production technology T , because as long as y ? P( x) ? Do ( x, y ) ? 1 (Färe, Grosskopf, Norris and Zhang (1994b); Färe and Primont, 1995). However, the generalization of the output distance function in (3.2) to include
v y b undesirable output by simply redefining it as Do ( x, y, b) = inf{? : ( x, , ) ? P( x)} ? ?
would not be meaningful since it would mean proportionate expansion of undesirable and desirable outputs as much as possible, without crediting the reduction of undesirables. In assessing productive efficiency of airports where there is joint production of desirable and undesirable outputs, this is not well-applicable. A rational airport manager should aim at maximizing only desirable, but minimizing undesirable outputs.
3.3 Directional output distance function
Due to the existence of both desirable and undesirable outputs in the output possibility set (3.1), the Shephard’s output distance function needs to be modified so that the efficiency measure will be able to credit for expansion of desirable and reduction of
- 51 -
Assessment of Productive Efficiency of Airports
+ undesirable outputs. First necessary notations are formally defined. Let y ? RM denote a + vector of desirable outputs, b ? R J+ denote a vector of undesirable outputs, and x ? R N
denote a vector of inputs. In airport context, K airports with ( x k , y k , bk ) are examined. The output possibility set P ( x) = {( y, b) : x can produce ( y, b)} in (3.1) satisfies certain axioms laid out by Shephard (1970), namely:
Property 1 Property 2 Property 3 Property 4
P (0) = {0,0}
+ P( x) is convex and compact for each x ? R N
( y, b) ? P( x) and ( y ' , b) ? ( y, b) imply ( y ' , b) ? P( x)
P( x) ? P( x' ) implies x ? x'
Property 1 states that zero inputs essentially yield zero outputs and any nonnegative input yields at least zero output. Sometimes this property is called a condition of no free lunch. Property 2 requires that only finite output should be produced given finite inputs. Property 3 imposes strong or free disposability of desirable outputs which means that it allows any desirable outputs to be disposed costlessly and still remain in P ( x). In other words, the disposal of any output can be achieved without incurring any costs in term of reducing the production of other outputs. Property 4 imposes strong or free disposability of inputs. The inputs are also allowed to be disposed costlessly. It also implies that an increase in any one input does not reduce the size of P ( x). Although in production theory it is common to assume that outputs are strongly disposable, it may not be appropriate for production technologies such as present airport operation in which undesirable outputs such as delays and noise cannot be costlessly - 52 -
Assessment of Productive Efficiency of Airports
disposed. Under regulated environment, an airport is forced to clean up its undesirable outputs or to reduce its levels. Desirable and undesirable outputs should be treated asymmetrically in terms of their disposability characteristics (Zaim, 2005). Even in the absence of regulations, increased environmental consciousness from stakeholders still require careful treatment of undesirable outputs as weakly disposable. To model the idea that there is a cost to reducing undesirable outputs, the next property is assumed.
Property 5
Weak disposability between desirable and undesirable outputs: If
( y, b) ? P ( x) and 0 ? ? ? 1, then (?y,?b) ? P( x). Property 5 implies that if undesirables are to be decreased, then the desirable outputs must also be decreased, holding inputs x constant. In other words, both desirable and undesirable outputs may be proportionally contracted, but undesirable outputs cannot, in general, be freely disposed. It models the idea that there is a cost to ‘cleaning up’ undesirable outputs. In the airport operation context, it implies that fewer delays can be achieved by letting an airport to service fewer aircraft movements. Finally, to recognize the nature of joint production of desirable and undesirable outputs, the following property is assumed:
Property 6
Null-jointness, if ( y, b) ? P( x), and b = 0, then y = 0.
This property states that if an output vector ( y, b) is feasible and there are no undesirable outputs produced, then under the null-jointness only zero desirable output can be produced. Equivalently, if some positive amount of the desirable output is produced then undesirable output must also be produced. In our airport operation context, nulljointness implies that where there are aircraft movements, there must be some delayed
- 53 -
Assessment of Productive Efficiency of Airports
flights which could be occurred by any cause (e.g., air carrier, extreme weather, nonextreme weather conditions, airport operations, late-arrival aircraft, security, human error, and accident). There are several ways of integrating the above six properties into the representation of output possibility set, including parametric and nonparametric approaches. The focus here is on the nonparametric model using DEA. The representation of output set is in the form of piecewise linear. Based on the six properties, the production technology for an individual airport k or P ( x k ) may be represented by the following output set: P ( x k ) = {( y, b) : (3.4)
k?K
?? ??
k
y km ? y km , m = 1,..........., M ,
kj
k?K
?? b
k k
= bkj , j = 1,............, J ,
k?K
x kn ? x kn , n = 1,.............., N ,
?k ? 0, k = 1,.........., K }
The constraints for the undesirable outputs b j , j = 1,....., J are equality
constraints, which under the constant returns to scale models the idea that these outputs are not freely disposable. Meanwhile free disposability of desirable outputs y m , m = 1,..., M and inputs x n , n = 1,...., N are allowed by using the inequalities in their respective constraints. ? k is an intensity vector. Figure 3.2 represents the construct of P( x) from four hypothetical airports i.e., A, B, C, and D. These airports are assumed to use the same amount of inputs, x , but produce
- 54 -
Assessment of Productive Efficiency of Airports
different amounts of desirable output, y and undesirable output, b. Since the linear programming of DEA approach is being used to estimate the output distance function, P( x) is drawn as piecewise linear rather than smooth curve as in Figure 3.1. The output possibility set, P( x), is bounded by 0ABCD0. Airports A, B, and C form an efficient frontier. y = desirable output G B I J H C
F (y, b)
A g = (gy, -gb)
P(x)
D 0 E b = undesirable output
Figure 3.2 Graphical illustration of directional output distance function concept
This figure illustrates how the assumptions are used in the construct. The origin (0,0) is included in P( x) because of the null-jointness assumption. The assumption of weak disposability implies that for any point on or inside P( x) , a proportional contraction in both ( y, b) is feasible. The vertical line segment CD occurs because of strong disposability between desirable outputs. The negative slope portion BC is possible because sometimes traffic may be blocked due to a long queue of delayed flights; hence
- 55 -
Assessment of Productive Efficiency of Airports
reducing throughput. Note that if undesirable outputs are ignored, P( x) will be the area bounded by 0GBCD0. Next, the interest is to assess the level of inefficiency for all airports which will tell how far each airport is from the efficient frontier. For airport F, the distance should be measured along the diagonal line FJ or in the direction of vector g = ( g y ,? g b ) . This measurement is justified on the premise that it seeks to maximize the expansion of desirable outputs and contraction of undesirable outputs simultaneously. The directional output distance function is then formulated as follows:
D 0 ( x, y, b; g y ,? g b ) = sup{? : ( y + ? g y , b ? ? g b ) ? P( x)}
?
(3.5)
To assess the level of inefficiency for an individual airport, the following linear programming problem is solved:
max ? s.t.
k?K
?? ??
k
y km ? y km + ? g y , m = 1,..........., M ,
kj
k?K
?? b
k k
= bkj ? ? g b , j = 1,............, J ,
(3.6)
k?K
x kn ? x kn , n = 1,.............., N ,
? k ? 0, k = 1,.........., K
The selection of a directional vector g = ( g y ,? g b ) is rather flexible. For example, using g = (0, b) implies that the level of inefficiency is measured along the horizontal line FH or projecting airport F to the frontier at H. Meanwhile, using g = ( y,0) yields the projection on the frontier at I. Using g = (1,?1) gives the same weight to both desirable and undesirable outputs. In this study, the vector g = ( y,?b) will be used, which means
- 56 -
Assessment of Productive Efficiency of Airports
that the projected direction depends on individual airport’s outputs. The linear programming in (3.6) is then rewritten as (3.7). max ? s.t.
k?K
?? ??
k
y km ? (1 + ? ) y km , m = 1,..........., M ,
kj
k?K
?? b
k k
= (1 ? ? )bkj , j = 1,............, J ,
(3.7)
k?K
x kn ? x kn , n = 1,.............., N ,
? k ? 0, k = 1,.........., K
The directional output distance function D 0 ( x, y, b; g y ,? g b ) or an optimal ? takes the minimum value of zero when it is not possible to expand the desirable outputs and contract undesirable outputs. This means that the airport is efficiently producing at the maximum possible outputs. To assess the productivity of K airports, the linear programming in (3.7) is solved K times, once for each individual airport. Thereafter, the optimal ? k will be called a efficiency score. A higher value of ? k indicates a lower level of efficiency. As a result, it can also be used to rank the performance of airports. The terms (1 + ? ) y km plus the corresponding output slacks and (1 ? ? )bkj in (3.7) give the projection of desirable and undesirable outputs onto the frontier. For an efficient airport with ? = 0 , the terms are simply ( y km , bkj ) or the current level of outputs. For inefficient airports, these terms represent the maximum possible production outputs or highest potential outputs that an airport could have produced. The results may provide benchmarks for airports to improve operational efficiency. However, as is shown in
?
- 57 -
Assessment of Productive Efficiency of Airports
Chapter 5, the selection of an appropriate set of outputs is crucial to the reasonableness of the benchmark. In order to relate the Shephard’s output distance function and the directional output distance function, let g = ( y, b) , then through (3.5), it becomes:
?
D 0 ( x, y, b; y, b)
= sup{? : ( Do ( x, ( y, b) + ? ( y, b)) ? 1} = sup{? : (1 + ? ) Do ( x, y, b) ? 1}
= sup{? : ? ?
1 ? 1} Do ( x , y , b ) (3.8)
=
1 ?1 Do ( x, y, b)
The expression in (3.8) shows that Shephard’s output distance function is a special case of the directional output distance function. The relation between the two can be written as: D 0 ( x, y, b; y,?b) =
?
1 ?1 Do ( x, y, b)
(3.9)
or equivalently,
Do ( x, y,?b)
=
1 v 1 + Do ( x, y, b; y,?b)
(3.10)
3.4
Malmquist productivity index with the presence of undesirable outputs
The concept of Malmquist productivity index was first introduced by Malmquist (1953) to compare the input of a production unit at two different points in time in terms of the maximum factor by which the input in one period could be decreased such that the
- 58 -
Assessment of Productive Efficiency of Airports
production unit could still produce the same output level of the other time period. The idea leads to the Malmquist input index. It has further been studied and developed in the non-parametric framework by several authors. See for example, among others, Caves, Christensen and Diewert (1982), Färe, Grosskopf, Lindgren and Roos (1994), Färe R., Grosskopf S., Norris M., Zhang Z. (1994) and Färe, Grosskopf and Roos (1998). It is an index representing the Total Factor Productivity (TFP) growth of a decision making unit (DMU), in that it reflects progress or regress in efficiency along with progress of the frontier technology over time under multiple inputs and multiple outputs framework. Given panel data, the Malmquist index evaluates the productivity change of an airport between two time periods. It is defined as the product of “Catch-up” and “Frontier-shift” terms. The catch-up (or recovery) term relates to the degree that an airport attains for improving its efficiency, while the frontier-shift (or innovation) term reflects the change in the efficient frontier surrounding the airport between the two time periods. Suppose that undesirable outputs are ignored. To analyze change of productivity over time, ( x, y ) is superscripted with corresponding time period. Then ( x t , y t ) and ( x t +1 , y t +1 ) are measures of inputs and outputs in period t and t + 1 respectively. From time t to t + 1 operational efficiency of airport k may change or (and) the frontier may shift. The output-oriented Malmquist productivity index using period t as the base period is the following (Caves, Christensen and Diewert, 1982).
t ? Do ( x t +1 , y t +1 ) ? Mo = ? ? t t t ? Do ( x , y ) ?
(3.11)
- 59 -
Assessment of Productive Efficiency of Airports
Alternatively, by using period t + 1 as the base period, the output-oriented Malmquist productivity index can be written as follows
? D t +1 ( x t +1 , y t +1 ) ? M o = ? o t +1 t t ? ? Do ( x , y ) ?
(3.12)
In order to avoid choosing an arbitrary frontier as reference, Färe, Grosskopf, Lindgren, and Roos (1994) suggest using the geometric average of the two indexes above. The resulting index is
t t +1 ? Do ( x t +1 , y t +1 ) Do ( x t +1 , y t +1 ) ? =? ? t t t t +1 t t ? Do ( x , y ) Do ( x , y ) ? 1/ 2
M
t +1 t
(3.13)
An alternative way of writing the Malmquist total factor productivity index is:
t t D t +1 ( x t +1 , y t +1 ) ? Do ( x t +1 , y t +1 ) Do (xt , yt ) ? = o t t t ? ? t +1 t +1 t +1 t +1 Do ( x , y ) ? Do ( x , y ) Do (xt , yt ) ? 1/ 2
M
t +1 t
(3.14)
Where the ratio outside the bracket measures the change in relative efficiency (i.e., the change in how far observed production is from maximum potential production) between period t and t + 1 . The geometric mean of the two ratios inside the bracket captures the shift in technology between the two periods evaluated at x t and x t +1 , that is
t +1 Do ( x t +1 , y t +1 ) t Do (xt , yt )
Efficiency change (EFFCH) =
(3.15)
t ? D t ( x t +1 , y t +1 ) Do (xt , yt ) ? Technical change (TECHCH) = ? t o +1 t +1 t +1 t +1 t t ? ? Do ( x , y ) D o ( x , y ) ?
1/ 2
(3.16)
Although the Malmquist index can in principal deal with undesirable outputs since it does not require knowledge on prices, the distance functions on which it is based
- 60 -
Assessment of Productive Efficiency of Airports
do not credit an airport for reducing level of undesirable outputs. Chung, Färe, and Grosskopf (1997) defined an output-oriented Malmquist-Luenberger (ML) productivity index that is comparable to the Malmquist index. Take undesirable outputs into consideration, ( x t , y t , b t ) and ( x t +1 , y t +1 , b t +1 ) are measures of inputs, desirable and undesirable outputs in t and t + 1 respectively. If the directional vector g t = ( y t ,?b t ) and g t +1 = ( y t +1 ,?b t +1 ) are chosen in corresponding periods, the output-oriented Malmquist-Luenberger productivity index is:
vt t t t t v t +1 t t t t 1/ 2 ? ? (1 + Do ( x , y , b ; y ,?b t )) (1 + Do ( x , y , b ; y ,?b t )) v t t +1 t +1 t +1 t +1 t +1 v t +1 t +1 t +1 t +1 t +1 t +1 ? =? ? (1 + Do ( x , y , b ; y ,?b )) (1 + Do ( x , y , b ; y ,?b )) ?
ML
t +1 t
(3.17)
v t +1 Here Do means that the reference technology is constructed based on data from
period t + 1 and the data being evaluated is included in the parentheses with its associated time period; for example ( x t , y t , b t ) would mean that the data to be evaluated are from period t. The directional vector g is time dependent. The definition is such that when the directional vector g is ( y, b) rather than ( y,?b) , the Malmquist-Luenberger index coincides with the Malmquist index. The Malmquist-Luenberger measure indicates productivity improvements if its value is greater than one. The value of less than one indicates decreases in productivity. In other words, it means that with the same amount of inputs as in period t + 1 , the greater quantity of outputs is produced as in period t (Murillo-Melchor, 1999). The productivity remains unchanged if M o is unity. Similar to the case of Malmquist, the Malmquist-Luenberger index can also be decomposed into two components, namely. - 61 -
Assessment of Productive Efficiency of Airports
MLEFFCH
t +1 t
vt t t t t 1 + Do ( x , y , b ; y , ?b t ) v t +1 t +1 t +1 t +1 t +1 t +1 = 1 + Do ( x , y , b ; y ,?b )
(3.18)
MLTECHCH tt +1 =
v t +1 t t t t v t +1 t +1 t +1 t +1 t +1 t +1 1 / 2 ? (1 + Do ( x , y , b ; y ,?b t )) (1 + Do ( x , y , b ; y ,?b )) ? vt t t t t v t t +1 t +1 t +1 t +1 t +1 ? ? t ? (1 + Do ( x , y , b ; y ,?b )) (1 + Do ( x , y , b ; y ,?b )) ?
(3.19) The product of (3.17) and (3.18) equals to MLtt+1 . The decomposition makes it possible to measure the change of technical efficiency and the movement of the frontier for a specific airport. Equation (3.18) measures the magnitude of technical efficiency change between periods t and t + 1 . The value of less than 1 indicates regress in technical efficiency. In other words, given a level of inputs, the same average output of all samples would have lead to produce more efficiently in period t + 1 than in period t (MurilloMelchor, 1999). Meanwhile the value greater than 1 indicates improvements. The technical efficiency remains unchanged if the value is unity. The second term measures the shift of frontier between periods t and t + 1 . Alternatively, Färe, and Grosskopf (2004b) construct another productivity index that has an additive structure, i.e., in terms of differences rather than ratios of MalmquistLuenberger ratio indexes in (3.17) – (3.19). The index is an output-oriented version of The Luenberger Productivity index introduced by Chambers (1996). Specifically, the index is: Ltt+1 =
v t +1 t +1 t +1 t +1 t +1 t +1 1 v t +1 t t t t Do ( x , y , b ; y ,?b t ) ? Do ( x , y , b ; y ,?b ) 2
[
- 62 -
Assessment of Productive Efficiency of Airports
vt t t t t v t t +1 t +1 t +1 t +1 t +1 + Do ( x , y , b ; y ,?b t ) ? Do ( x , y , b ; y , ?b )
]
(3.20)
Following the idea of Chambers, Färe, and Grosskopf (1996) the Luenberger productivity index can be additively decomposed into an efficiency change and a technical change component,
vt t t t t v t +1 t +1 t +1 t +1 t +1 t +1 LEFFCH tt +1 = Do ( x , y , b ; y ,?b t ) ? Do ( x , y , b ; y , ?b )
(3.21)
and LTECHCH tt +1 =
v t t +1 t +1 t +1 t +1 t +1 1 v t +1 t +1 t +1 t +1 t +1 t +1 D o ( x , y , b ; y , ?b ) ? D o ( x , y , b ; y , ?b ) 2
[
v t +1 t t t t vt t t t t + Do ( x , y , b ; y ,?b t ) ? Do ( x , y , b ; y ,?b t )
]
(3.22)
respectively. The sum of these two components equals the Luenberger productivity index. The index and its components signal improvements with values greater than zero, and declines in productivity with values less than zero. As usual, selection of the directional vector is flexible. If the vector g = ( y,?b) is chosen, i.e., choosing the observed desirable and (negative) undesirable output vector to determine the direction, then each airport may be evaluated in a different direction, i.e., in its own direction. This is just typically the case for Shephard type distance functions.
- 63 -
Assessment of Productive Efficiency of Airports
y = desirable output
M
#
Pt+1(x)
D’ N
L H A
#
D
g = (1, -1)
Pt(x)
0
b = undesirable output
Figure 3.3 The Luenberger productivity indicator
Figure 3.3 illustrates how the Luenberger productivity indicator is constructed. For simplicity, it is assumed that inputs are the same in period t and t + 1 and are represented by x = x t = x t +1 . Without loss of generality, a directional vector g = (1,1) is assumed and illustration is for the case of technological progress (the frontier shifts to the left at period t + 1 ). Given g = (1,1), the directional output distance function is an estimate of the simultaneous unit expansion in the desirable output and unit contraction in the undesirable output. An airport is observed to produce at point D in period t and at D' in period t + 1 . If the airport was to eliminate technical inefficiency it could operate at H in period t and at M in period t + 1 . The Luenberger efficiency change indicator is LEFFCH tt +1 = DH D' M ? og og and the Luenberger technical change indicator is
- 64 -
Assessment of Productive Efficiency of Airports
LTECHCH tt +1 =
1 ? DL DH D' M D' N ? 1 ? HL NM ? ? + ? = ? + . 2? og og og ? og ? ? og ? 2 ? og ? and
v t t +1 t +1 t +1 Do ( x , y , b ;1,1) < 0,
By
construction, indicates
v t +1 t t t Do ( x , y , b ;1,1) > 0
so
LTECHCH > 0
technical progress. To compute the Malmquist-Luenberger productivity index in (3.17) and Luenberger productivity index in (3.20), including their decomposed components in (3.18), (3.19), (3.21) and (3.22), four distance functions must be estimated,
t t +1 t i.e., Do ( x t , y t , b t ; y t ,?b t ), Do ( x t +1 , y t +1 , b t +1 ; y t +1 ,?b t +1 ), Do ( x t +1 , y t +1 , b t +1 ; y t +1 , b t +1 ) and
t +1 ( x t , y t , b t ; y t ,?b t ) . The latter two are mixed-period distance functions which are Do
obtained by evaluating performance of an airport from one period in another period. Färe, Grosskopf, Norris, and Zhang (1994) make use of the fact that the output distance function is reciprocal to the output-based Farrell measure of technical efficiency, and then modify the directional output distance function Do ( x, y, b; y,?b) in (3.7) to accommodate time period. The computation steps are summarized below.
t t t t t t t t 1) To estimate Do ( xk ' , y k ' , bk ' ; y k ' ,?bk ' ) for airport k ' , compare ( y k ' , bk ' ) to the
?
frontier at time t , and solve the following linear program:
t t t t t t Do ( xk ' , y k ' , bk ' ; y k ' , ?bk ' ) = max ?
s.t.
k?K
?? ??
t k t k
t t y km ? (1 + ? ) y k 'm , m = 1,..........., M , t kj
k?K
?? b
t k
= (1 ? ? )bkt ' j , j = 1,............, J ,
(3.23)
k?K
t t x kn ? xk 'n , n = 1,.............., N ,
?tk ? 0, k = 1,.........., K
- 65 -
Assessment of Productive Efficiency of Airports
t +1 t +1 t +1 t +1 t +1 t +1 t t 2) To estimate Do ( xk ' , y k ' , bk ' ; y k ' ,?bk ' ) for airport k ' , compare ( y k ' , bk ' ) to
the frontier at time t + 1, and solve the following linear program:
t +1 t +1 t +1 t +1 t +1 t +1 ( xk Do ' , y k ' , bk ' ; y k ' , ?bk ' ) = max ?
s.t.
k?K
?? ?? ??
t +1 k
t +1 t +1 ? (1 + ? ) y k y km 'm , m = 1,..........., M ,
k?K
t +1 t +1 k kj
b
1 = (1 ? ? )bkt + ' j , j = 1,............, J , t +1 ? xk 'n , n = 1,.............., N ,
(3.24)
k?K
t +1 t +1 k kn
x
?tk+1 ? 0, k = 1,.........., K
t t +1 t +1 t +1 t +1 t +1 t +1 t +1 3) To estimate Do ( xk ' , y k ' , bk ' ; y k ' ,?bk ' ) for airport k ' , compare ( y k ' , bk ' )
to the frontier at time t , and solve the following linear program:
t t +1 t +1 t +1 t +1 t +1 ( xk Do ' , y k ' , bk ' ; y k ' , ?bk ' ) = max ?
s.t.
k?K
?? ??
t k t k
t +1 t ? (1 + ? ) y k y km 'm , m = 1,..........., M , t kj 1 = (1 ? ? )bkt + ' j , j = 1,............, J ,
k?K
?? b
t k
(3.25)
k?K
t t +1 ? xk x kn 'n , n = 1,.............., N ,
?tk ? 0, k = 1,.........., K
t +1 t t t t t t t 4) To estimate Do ( xk ' , y k ' , bk ' ; y k ' , ?bk ' ) for airport k ' , compare ( y k ' , bk ' ) to the
frontier at time t + 1, and solve the following linear program:
- 66 -
Assessment of Productive Efficiency of Airports
t +1 t t t t t ( xk Do ' , y k ' , bk ' ; y k ' , ?bk ' ) = max ?
s.t.
k?K
?? ?? ??
t +1 k
t +1 t ? (1 + ? ) y k y km 'm , m = 1,..........., M ,
k?K
t +1 t +1 k kj
b
= (1 ? ? )bkt ' j , j = 1,............, J ,
t ? xk 'n , n = 1,.............., N ,
(3.26)
k?K
t +1 t +1 k kn
x
?tk+1 ? 0, k = 1,.........., K
Substituting (3.23) – (3.26) for the corresponding terms in (3.17) – (3.19), the Malmquist-Luenberger productivity index and the two components can be obtained. Similarly, substituting (3.23) – (3.26) for the corresponding terms in (3.20) – (3.22), the Luenberger productivity index and the two components can be obtained. In summary, the Malmquist-Luenberger and Luenberger productivity indexes together with their components provide more insightful information regarding sources of productivity change between two time periods. Chapter 4 will describe the first case study which is summarized from three publications (Pathomsiri and Haghani, 2004; Pathomsiri, Haghani and Schonfeld, 2005; Pathomsiri, Haghani, Dresner and Windle, 2006a). The study is the first attempt to assess productivity of airports operating in multiple airport systems using the DEA model. This case study provides primary understanding on typical results when undesirable outputs are not ignored. Then another case study of U.S. airports will address the shortcomings later.
- 67 -
Assessment of Productive Efficiency of Airports
CHAPTER 4 CASE STUDY 1 PRODUCTIVITY OF AIRPORTS IN MULTIPLE AIRPORT STSYEMS
Many metropolitan regions around the world are served by multiple commercial airports. These regions are called “multiple airport systems” or MASs among aviation community. There have been many stories about functional failures in planning and managing of MASs worldwide due to over-investment or underutilization (Caves and Gosling, 1999; de Neufville, 1995; de Neufville and Odoni, 2003). It is well-documented that several airports in MAS cannot achieve sufficient traffic to economically justify capital investment. In other words, the investment is actually not sufficiently productive. However, it is very surprising that there is no productivity study that focuses on airports in multiple airport systems, though MASs involve several times more capital investment. If one is about to assess two airports comparable in both size and market, an airport in single-airport system which is enjoying its monopolistic status perhaps performs no less efficiently than an airport that is struggling with competitors in an MAS. In a productivity study of mixed samples of airports operating in single-airport and multiple-airport systems, it was found that U.S. airports in MASs were not operating more efficiently than other U.S. airports (Sarkis, 2000). This case study aims to fill this gap by looking at the productivity of airports in MASs. The results may be perceived as like-a-like comparisons among airports operating - 68 -
Assessment of Productive Efficiency of Airports
in the same market structure. Due to the unavailability of data on undesirable outputs at international airports, the study is therefore restricted to the consideration of desirable outputs only. As a result, Data Envelopment Analysis model is applicable. This chapter describes the definition of MAS, modeling of airport operation, input and output measures of airport operations and data collection. Note that the content in this chapter is based on three publications, i.e., Pathomsiri and Haghani (2004), Pathomsiri, Haghani and Schonfeld (2005) and Pathomsiri, Haghani, Dresner and Windle (2006a).
4.1 Definition of multiple airport system
Multiple airport system (MAS) is explicitly defined in few publications (de Neufville, 1995; de Neufville and Odoni, 2003; Hansen and Weidner, 1995). In one textbook (de Neufville and Odoni, 2003), the authors defined an MAS as “the set of significant airports that serve commercial transport in a metropolitan region, without regard to ownership or political control of the individual airports.” This definition involves four important points. First, MAS focuses on airports serving commercial transport. Second, MAS refers to a metropolitan region rather than a city. The region can expand to cover several cities as in the case of New York/New Jersey. Third, MAS focuses on the market, not the ownership of the airports. Although five airports in London area are owned by three different organizations, they form the London MAS since they all serve the same market. Finally, MAS focuses on significant airports. The authors suggest a threshold of more than one million passengers per year for identifying significant airports. Another paper (Hansen and Weidner, 1995) defined an MAS as two or more airports operating with scheduled passengers enplanements in a contiguous metropolitan
- 69 -
Assessment of Productive Efficiency of Airports
area in such a way as to form an integrated airport system and satisfies both of the following criteria: Each airport in the system is included in the same community by the Federal
Aviation Administration (FAA) or within 50 km. (30 miles) of the primary airport of an FAA-designated “large-hub” community, or each airport is in the same Metropolitan Statistical Area (MSA) or Consolidated MSA (CMSA). The Herfindahl Concentration Index (HCI) for the airports is less than 0.95.
HCI is the sum of squared market shares of all airports in an MAS. For example, in 2003, distribution pattern of passenger traffic in Baltimore/Washington MAS was 20,094,756 (39.34%), 16,767,767 (32.83%) and 14,214,803 (27.83%) at
Baltimore/Washington International (BWI), Washington Dulles International (IAD) and Ronald Reagan Washington National (DCA) respectively. Therefore, HCI is equal to 0.39342 + 0.32832 + 0.27832 = 0.34. Similarly, in Houston MAS, in 2002 George Bush Intercontinental (IAH), William P. Hobby (HOU) and Ellington Field (EFD) accommodated 33,905,253 (80.69%), 8,035,727 (19.112%) and 76,035 (0.18%) passengers respectively. HCI is equal to 80.692 + 19.122 + 0.182 = 0.688. By the above criteria, the Houston MAS is not so concentrated but somehow competitive. Note that for a single airport system, HCI is 1.0. For an MAS where traffic is evenly divided among N airports, HCI is 1/N (Baltimore/Washington MAS may be a close example). The above two examples of MAS definitions indeed are very similar. The difference may be the significance of the airport. One uses a threshold of passenger traffic to identify the MAS whereas the other uses HCI, regardless of traffic volume. It is
- 70 -
Assessment of Productive Efficiency of Airports
harmless to think of the latter criterion as a measure of sufficient significance in a sense that only competitive MASs are included in the analysis. Since definitions are in good agreement and well-documented in the publications, 14 MASs in the U.S. from (Hansen and Weidner, 1995) are then adopted. For more comprehensive study, the scope is expanded to other MASs worldwide. In this study, the lists of MASs in publications (Caves and Gosling, 1999; de Neufville, 1995; de Neufville and Odoni, 2003) were collected and checked if they satisfy both definitions. Eventually, we identified another 11 non-US MASs. Totally, there are 25 MASs in this study, involving 75 airports in four continents, i.e., North America, South America, Europe and Asia. Table 4.1 provides the list of all 25 MASs together with the airports in the systems along with the International Civil Aviation Organization (ICAO) airport codes. The Table shows two computed HCIs, one based on passenger and the other based on aircraft movements. This means that airports in the same region may compete for passengers or aircraft movements or both. As a result, as long as either HCI is below 0.95, the region is an MAS.
- 71 -
Assessment of Productive Efficiency of Airports
Table 4.1 List of multiple airport systems and corresponding Herfindahl Concentration Indices, 2002
Region 1 2 Chicago, IL, USA New York City, NY, USA Airport Code ORD MDW CGX EWR JFK LGA ISP HPN SWF LAX SNA ONT BUR PSP LGB OXR PMD SFO SJC OAK STS CCR SBA SMX DFW DAL IAH HOU EFD BWI IAD DCA MIA FLL PBI PNS VPS Airport Name O'Hare International Midway International Merrill C. Meigs Newark Liberty International, NJ John F. Kennedy International, NY LaGuardia, NY Long Island MacArthur, NY Westchester County, NY Stewart International, NY Los Angeles International, CA John Wayne, CA Ontario International, CA Bob Hope, CA Palm Spring International, CA Long Beach, CA Oxnard, CA Palmdale Regional, CA San Francisco International, CA Mineta San Jose International, CA Oakland International, CA Sonoma County, CA Buchanan Field, CA Santa Barbara, CA Santa Maria Public, CA Dallas/Fort Worth International, TX Dallas Love Field, TX George Bush Intercontinental, TX William P. Hobby, TX Ellington Field, TX Baltimore/Washington International, MD Washington Dulles International, VA Ronald Reagan Washington National, DC Miami International, FL Fort Lauderdale - Hollywood International, FL Palm Beach International, FL Pensacola Regional, FL Okaloosa Regional, FL Aircraft Movements 922,817 304,304 31,972 405,562 287,606 362,439 223,063 167,776 123,642 645,424 368,627 149,292 162,211 85,243 350,603 88,027 33,352 351,453 207,510 371,988 114,854 142,329 159,835 76,426 765,109 245,564 456,831 246,230 102,016 304,921 372,636 215,691 446,235 280,737 166,908 130,826 118,423 HCI-Air 0.596 0.191 Total Passengers 66,565,952 17,371,036 86,483 29,202,654 29,943,084 21,986,679 1,890,580 930,097 362,017 56,223,843 7,903,066 6,517,050 4,620,683 1,108,695 1,453,412 45,306 226 31,456,422 11,115,778 13,005,642 3,598 0 728,307 58,104 52,828,573 5,622,754 33,905,253 8,035,727 76,035 19,012,529 17,075,965 12,871,885 30,060,241 17,037,261 5,483,662 1,345,970 631,592 HCI-PAX 0.670 0.315
3
Los Angeles, CA, USA
0.209
0.543
4
San Francisco, CA, USA
0.240
0.415
5 6 7 8 9 10
Santa Barbara, CA, USA Dallas/Fort Worth, TX, USA Houston, TX, USA Washington, DC, USA Miami, FL, USA Pensacola, FL, USA
0.562 0.632 0.432 0.349 0.383 0.501
0.863 0.826 0.688 0.342 0.443 0.565
- 72 -
Assessment of Productive Efficiency of Airports
Table 4.1 List of multiple airport systems and corresponding Herfindahl Concentration Indices, 2002 (Continued)
Region 11 12 13 14 15 16 17 18 Detroit, MI, USA Cleveland, OH, USA Norfolk, VA, USA Oshkosh/Appleton, WI, USA Montreal, Canada Rio de Janeiro, Brazil Sao Paulo, Brazil Buenos Aires, Argentina Airport Code DTW DET CLE CAK ORF PHF ATW OSH YUL YMX GIG SDU GRU CGH AEP EZE DOT SFD LHR LTN LGW STN LCY GLA EDI PIK CDG ORY TXL SXF THF LIN MXP SVO VKO DME HND NRT Airport Name Detroit Metropolitan Wayne County, MI Detroit City, MI Cleveland Hopkins International, OH Arkon-Canton, OH Norfolk International, VA Newport News/Williamsburg International, VA Outagamie County Regional, WI Wittman Regional, WI Montreal-Dorval International Montreal-Mirabel International Rio De Janeiro-Galeao International Santos Dumont Sao Paulo Guarulhos International Congonhas Aeroparque Jorge Newbery Ezeiza International International Don Torcuato San Fernando London Heathrow London Luton London Gatwick London Stansted London City Glasgow Edinburgh Glasgow Prestwick International Roissy-Charles-de Gaulle Orly Tegel Schoenefeld Tempelhof Linate Malpensa Sheremetyevo Vnukovo Domodedovo Tokyo International (Haneda) New Tokyo International (Narita) Aircraft Movements 490,885 69,066 251,758 119,958 125,622 228,504 57,755 115,288 192,225 32,977 83,731 117,144 160,451 266,231 91,350 50,755 23,392 34,819 466,554 80,921 242,380 170,774 56,102 105,197 118,419 43,346 510,098 211,080 127,470 37,389 48,026 110,494 214,886 124,630 65,759 84,102 282,674 164,270 HCI-Air 0.784 0.563 0.542 0.555 0.750 0.514 0.531 0.316 Total Passengers 32,477,694 0 10,455,204 894,798 3,464,246 515,056 491,744 3,912 7,816,052 990,937 5,810,868 5,626,328 12,804,091 12,562,319 4,519,424 4,054,473 23,148 11,676 63,338,641 6,496,258 29,628,423 16,049,288 1,604,773 7,807,060 6,932,106 1,487,113 48,350,172 23,169,725 9,879,888 1,688,028 612,867 7,815,316 17,441,250 10,895,225 3,120,210 6,683,268 61,079,478 28,883,606 HCI-PAX 1.000 0.855 0.775 0.984 0.800 0.500 0.500 0.497
19
London, United Kingdom
0.305
0.379
20 21 22 23 24 25
Glasgow, United Kingdom Paris, France Berlin, Germany Milan, Italy Moscow, Russia Tokyo, Japan
0.378 0.586 0.440 0.551 0.357 0.535
0.422 0.562 0.680 0.573 0.404 0.564
- 73 -
Assessment of Productive Efficiency of Airports
4.2
Modeling airport operations
Each airport in the MAS is viewed as a production unit. As a result, airport operation can be modeled as a production process that requires some inputs for running day-to-day operations in order to produce some target outputs. Necessary inputs include production factors such as capital and labor. Most airport managers set target to maximize movement of aircrafts, passenger throughput and quantity of cargo transported. These outputs are highly desirable and the primary reason for building an airport. Due to the unavailability of data on undesirable outputs (e.g., delays, noise) at international airports, the assessment has to be restricted to the consideration of desirable outputs only. As a result, Data Envelopment Analysis (DEA) is applicable as an analytical tool.
4.3 Input and output measures of airport operations
The formulation of any DEA model given in Table 2.5 requires data on quantity of inputs and desirable outputs, ( x kn , y km ) for individual airports. The selection of inputs and outputs is an important decision issue in the assessment of airport productivity. The general suggestion is to include all important measures that are in the interest of the management. Such measures should be common for all airports so that the performance would provide meaningful interpretation. In practice, the main problem is the availability of the data across all airports rather than model limitations. After all, three common physical inputs are considered in this analysis:
x1 = Land area (LAND), acre x 2 = Number of runways (RW)
x3 = Runway area (RWA), acre - 74 -
Assessment of Productive Efficiency of Airports
These inputs are necessary infrastructure for all airports. Land area (acre) represents a considerable share of capital investment that an airport should fully utilize. Number of runways counts all existing runways at the airport, regardless of their utilization level. Runway area is the summation of product between length and width of all runways. The consideration of runway area should explains variations in productivity better than using the number of runway alone, since it takes into account the effect of size and design configuration such as length, width, and separation. Other inputs such as terminal area, number of gates, number of employees and expense cannot be included due to the lack of complete data across samples. For the set of desirable outputs, it is assumed that airport managers aim at producing the following two outputs as much as possible:
y1 = Aircraft movements y 2 = Passengers
Number of aircraft movements includes all kinds of movements, i.e., commercial aircrafts, cargo aircrafts, general aviation, and others. The number of passengers counts both arriving and departing passengers for all type of commercial passengers, i.e., international, domestic and direct transit passengers. Other desirable outputs such as cargo throughput and revenues cannot be considered due to the lack of complete data across all samples, especially for small U.S. and non-U.S. airports. Inclusion of these outputs will reduce sample size drastically; hence, it is decided to maintain all samples. Note that these input and output measures may be rather limited to partial factors of airport operations, but they have been used in previous studies such as Gillen and Lall (1997, 1998); Pels, Nijkamp and Rietveld
- 75 -
Assessment of Productive Efficiency of Airports
(2001,2003) (see Table 2.4 for details). Given the above-selected three inputs and three desirable outputs, the assessment may be perceived as the measurement of productive efficiency of airside operation.
4.4 Data collection
The study period is during 2000 – 2002. On the output side, statistics on the number of passengers and aircraft movements are collected from Airports Council International publications (ACI, 2002 - 2004). The missing data are supplemented from several sources such as FAA website (FAA, 2004a), airports’ official websites, airport newsletters, reports, airport contacts and e-mail correspondences. Collecting input data caused more trouble since there is no single source available at hand. Airport Master Record (AMS) database (FAA, 2004b) contains the most recent data on characteristics of US airports. The best effort was made to verify this recent data with airport managers whether there are runway expansions or constructions during the period 2000 – 2002. Some airports had improved their runways. For example, George Bush Intercontinental (IAH) expanded and extended runway 15R/33L to 10000’ x 150’ in 2002. Detroit Metropolitan Wayne County (DTW) opened its 6th runway on December 11, 2001. The number of runways is edited accordingly. The number of runway and runway acreage are computed precisely by the time it is in service during the year, rounding down in month. In case of DTW, for example, it is concluded that it had 5 and 6 runways in 2001 and 2002, respectively. Input data of non-US airports are more difficult to collect since there is no database such as AMS (FAA, 2004b). Inevitably, one has to rely on information from airports’ official websites, airport newsletters, reports, airport contacts and e-mail correspondences. Also, it is - 76 -
Assessment of Productive Efficiency of Airports
verified with airport managers whether there was any change in runway configuration during 2000 – 2002. Similarly to the US airport case, it is found that Narita International (NRT) opened its new second parallel runway in April 2002. As a result, it used 1.667 (= 1 + 8/12) runways in 2002. Ultimately, the study had to drop Santos Dumont/Rio de Janeiro (SDU), International Don Torcuato/Buenos Aires (DOT) and Vnukovo/Moscow (VKO) airports from the sample due to unavailable land area data. The final dataset used in this study contains 72 airports with complete input and output data. The sample size is relatively larger than most previous studies (Abbott and Wu, 2002; Adler and Berechman, 2001; Bazargan and Vasigh, 2003; Fernandes and Pacheco, 2001, 2002, 2005; Gillen and Lall, 1997, 1998; Hooper and Hensher, 1997; Martin and Roman, 2001; Martin-Cejas (2002); Murillo-Melchor, 1999; Nyshadham and Rao, 2000; Oum, Yu and Fu, 2003; Pacheco and Fernandes 2003; Parker, 1999; Pels, Nijkamp and Rietveld, 2001, 2003; Salazar de la Cruz,1999; Sarkis, 2000; Sarkis and Tulluri, 2004; Yoshida, 2004; Yu, 2004). The number of samples was checked against several applicable rules of thumb to guarantee the sufficiency and meaningful interpretation. In DEA applications, one frequent problem is a lack of discriminatory power between DMUs as a result of an excessive number of measures with respect to the total number of DMUs. The larger the number of input and output measures for a given number of airports the less discriminatory the DEA model becomes. Given a certain set of samples, this means that the addition of measures will reduce the discriminatory power of the DEA model. Essentially, this is because it is possible that an airport may dominate all others on one measure, which in turn makes it look equally efficient compared to other efficient airports. This is a major issue encountered by Parker (1999),
- 77 -
Assessment of Productive Efficiency of Airports
Adler and Berechman (2001) and Yu (2004). To avoid this problem, the straightforward way is to guarantee that there will be a sufficient number of airports for comparison, regarding any of the measures. Boussofiane, Dyson and Thanassoulis (1991) recommend that the total number of DMUs be much greater than the number of inputs times the number of outputs. Compared to this analysis where three inputs and two outputs are selected for the assessment of airport productivity; the number of samples needs to be much more than 3 x 2 or 6 airports in order to reduce the chance that an airport is too dominant compared to the others on a particular measure. According to the recommendation, the sample size of 72 airports is deemed satisfactory. To avoid losing discriminatory power, Cooper, Seiford and Tone (2000: page 103) recommend that the desired number of DMUs exceed m + s several times. They suggest a more stringent rule of thumb in the following formula (Cooper, Seiford and Tone, 2000: page 252). n ? max{m x s, 3(m + s )} (4.1)
where n is the number of DMUs, m and s are the numbers of input and output measures respectively. Substituting m and s , yields the minimum number of samples: n ? max{3 x 2, 3(3 + 2)} = max{6,15} = 15 (4.2)
Again, the sample size of 72 satisfies this recommendation. After all, it can be concluded that the sample size is sufficient for the analysis. Table 4.2 summarizes the descriptive statistics of the samples. Input measures are rather stable over time; only slight - 78 -
Assessment of Productive Efficiency of Airports
changes in number of runways and runway acreage are recorded. The variation in terms of the range between maximum and minimum is wide, indicating that airports in MAS are much different in scale of operation. Similarly, airports’ outputs are widely variable indicating that airports are much different in scope of operation. Six partial productivity ratios are also shown to provide more information on the utilization of airport. In the next chapter, results from assessing productivity of 72 airports by DEA model will be presented. The assessment will be discussed with respect to operational efficiency of individual airports. In addition, Censored Tobit regression models are also estimated for explaining variations in total productivity level.
- 79 -
Assessment of Productive Efficiency of Airports
Table 4.2 Descriptive statistics of 72 airports in MASs, 2000 - 2002
Variables Land acreage area (LAND) Statistics Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. 2000 70 18,076 2,865 3,419 1 7 2.69 1.26 9.78 295.54 79.56 52.84 30,479 908,989 234,800 188,593 0 72,144,244 14,063,577 18,120,258 0 37,316 6,561 7,678 2001 70 18,076 2,865 3,419 1 7 2.69 1.26 9.78 295.54 79.66 52.81 31,240 911,917 223,829 179,799 0 67,448,064 13,363,106 16,989,050 0 32,254 6,379 7,449 2002 70 18,076 2,865 3,419 1 7 2.72 1.28 9.78 298.60 80.98 54.41 31,972 922,817 216,093 172,890 0 66,565,952 13,240,016 16,786,913 0 32,333 6,436 7,662 2 736 157 156 16,489 184,314 76,211 40,772 299 14,983 3,021 2,052 0 21,112,880 4,502,157 5,049,647 0 572,148 142,365 141,131
Number of runways (RW)
Runway acreage area (RWA)
Annual aircraft operations (AIR)
Annual total passengers (PAX)
PAX/LAND
AIR/LAND
2 774 173 174
19,957 195,858 83,260 45,021 362 15,765 3,332 2,257 0 27,389,915 4,861,631 5,665,696 0 583,604 151,358 148,524
2 757 163 162
18,088 189,452 79,783 43,090 328 15,401 3,166 2,135 0 25,379,370 4,657,794 5,379,152 0 548,702 145,135 141,262
AIR/RW
AIR/RWA
PAX/RW
PAX/RWA
- 80 -
Assessment of Productive Efficiency of Airports
CHAPTER 5 CASE STUDY 1 RESULTS AND DISCUSSION
5.1 Selection of a DEA model
Similar to Martin and Roman (2001), it is assumed that once an airport has invested in the infrastructure, it is difficult for managers to disinvest to save costs. Consequently, airport managers are more interested to know the probable levels of outputs, given the existing infrastructure (Fernandes and Pacheco, 2002). From this viewpoint, the outputorientation DEA model is preferred. Since the analysis is focused on rather narrow study period during 2000 – 2002, the variable return to scale (VRS) frontier type is chosen to reflect the short-term view (Ganley and Cubbin, 1992). In VRS frontier type DMUs are not penalized for operating at a non-optimal scale (Banker, 1984; Banker and Thrall, 1992). After all, the applicable model is the DEA-Output-VRS. Its mathematical formulation is given below: max ? s.t.
k?K
?? ?? ??
k
+ y km ? s m = ?y km , m = 1,..........., M , ? x kn + s n = x kn , n = 1,.............., N ,
(5.1)
k?K
k
k?K
k
= 1, k = 1,.........., K
?k ? 0
For each year during 2000 – 2002, the DEA model in (5.1) is solved 72 times; i.e., one time for each airport, to determine the optimal efficiency scores ? * . ? * measures the - 81 -
Assessment of Productive Efficiency of Airports
level of inefficiency. An efficient airport will have ? * = 1 which means it does not need to increase its outputs. It is already on the efficient production frontier. In other words, the airport is more productive than others, given the same amount of inputs. All inefficient airports will have ? * > 1. The higher value of ? * shows greater inefficiencies.
5.2 Efficient scores
Table 5.1 presents the efficiency scores of 72 airports during 2000 – 2002. Bold typeface highlights airports on efficient production frontier. For example, in 2002, there are 12 efficient airports i.e. O’Hare International (ORD), Merrill C. Meigs (CGX) 1, LaGuardia (LGA), Los Angeles International (LAX), John Wayne (SNA), Oxnard (OXR), Palmdale (PMD), Congonhas/Sao Paulo (CGH), Heathrow/London (LHR), Stansted/London (STN), City/London (LCY), and Haneda/Tokyo (HND). These 12 airports form a piece-wise linear efficient production frontier under variable return-to-scale assumption. Seemingly, efficient airports can be classified into two groups i.e., the busy and the compact. The busy group is usually a primary or major airport in the region such as O’Hare International (ORD), Los Angeles International (LAX), Aeroparque Jorge Newbery/Buenos Aires (AEP), Heathrow/London (LHR), Haneda/Tokyo (HND) and Narita/Tokyo (NRT). Their land areas are relatively large, ranging from 2,000 – 8,000 acres. Annual passenger traffics are consistently among the top of the world. Another interesting observation is that they dominate the market by having more than 50% of passengers. Their high traffic flows enable airports to operate more efficiently at higher utilization level than others. The compact group includes Merrill C. Meigs (CGX), LaGuardia (LGA), John Wayne (SNA), Oxnard (OXR), Palmdale (PMD), Congonhas/Sao Paulo (CGH),
1
The airport was permanently closed in March 2003.
- 82 -
Assessment of Productive Efficiency of Airports
Stansted/London (STN) and City/London (LCY). These airports are relatively small in size (between 70 – 960 acres) with one or two runways. They are alternative airports, except Congonhas/Sao Paulo (CGH) which is a primary airport constrained in downtown area. Although traffic may not be so high, they still can be efficient airports because of their sufficiently high utilization rate. Airports with efficiency scores below two may be considered satisfactorily efficient in terms of input utilization. The airports with scores consistently higher than two should be monitored closely for improving efficiency. Some airports with consistently very high scores such as Montreal-Mirabel (YMX), Glasgow Prestwick International (GLA),
Schoenefeld/Berlin (SXF), and Tempelhof/Berlin (THF) are significantly under-utilized or over-invested. These airports tend to use comparable inputs to others but service far fewer aircraft movements and passengers.
- 83 -
Assessment of Productive Efficiency of Airports
Table 5.1 Efficiency scores, 2000 – 2002
No.
1
Multiple Airport System
Chicago, IL, USA
2
New York City, NY, USA
3
Los Angeles, CA, USA
4
San Francisco, CA, USA
5 6 7 8 9 10 11 12
Santa Barbara, CA, USA Dallas/Fort Worth, TX, USA Houston, TX, USA Washington, DC, USA Miami, FL, USA Pensacola, FL, USA Detroit, MI, USA Cleveland, OH, USA
Airport Code ORD MDW CGX EWR JFK LGA ISP HPN SWF LAX SNA ONT BUR PSP LGB OXR PMD SFO SJC OAK STS CCR SBA SMX DFW DAL IAH HOU EFD BWI IAD DCA MIA FLL PBI PNS VPS DTW DET CLE CAK - 84 -
2000
1.000 1.302 1.000 1.251 2.041 1.000 2.166 2.067 2.842 1.000 1.000 2.479 2.407 4.670 1.259 1.000 1.000 1.640 1.596 1.230 2.828 1.773 2.547 5.198 1.085 1.931 1.396 1.951 5.813 2.050 1.283 1.414 1.132 1.720 2.490 3.293 3.285 1.486 1.214 1.739 4.117
2001
1.000 1.293 1.000 1.251 2.119 1.000 2.058 2.049 3.330 1.000 1.000 2.399 2.354 4.510 1.285 1.000 1.000 1.786 1.691 1.335 2.730 2.404 2.578 4.982 1.164 1.956 1.352 1.916 6.240 1.902 1.407 1.637 1.180 1.615 2.460 3.252 3.203 1.525 1.846 1.888 3.953
2002
1.000 1.174 1.000 1.234 2.076 1.000 1.996 2.197 2.981 1.000 1.000 2.446 2.266 4.324 1.231 1.000 1.000 1.816 1.977 1.398 3.210 2.558 2.527 4.823 1.206 1.809 1.375 1.806 4.710 1.940 1.361 1.779 1.134 1.552 2.732 2.818 3.113 1.692 2.664 1.990 4.014
Assessment of Productive Efficiency of Airports
Table 5.1 Efficiency scores, 2000 – 2002 (Continued)
No. 13 14 15 16 17
18
Multiple Airport System Norfolk, VA, USA Oshkosh/Appleton, WI, USA Montreal, Canada Rio de Janeiro, Brazil Sao Paulo, Brazil
Buenos Aires, Argentina
19
London, United Kingdom
20 21 22 23 24
25
Glasgow, United Kingdom Paris, France Berlin, Germany Milan, Italy Moscow, Russia
Tokyo, Japan
Airport Code ORF PHF ATW OSH YUL YMX GIG GRU CGH AEP EZE SFD LHR LTN LGW STN LCY GLA EDI PIK CDG ORY TXL SXF THF LIN MXP SVO DME HND NRT
2000 3.107 1.844 5.871 4.478 2.792 9.679 4.592 2.016 1.082 1.000 5.630 2.810 1.000 1.191 1.113 1.000 1.000 2.899 3.756 8.583 1.395 2.192 2.798 8.022 7.784 3.260 1.480 2.887 8.025 1.070 1.000
2001 3.161 1.734 6.642 4.596 2.846 10.306 4.021 1.976 1.000 1.000 5.910 2.681 1.000 1.275 1.080 1.000 1.000 2.402 3.272 8.300 1.290 2.304 2.687 9.144 7.744 2.458 1.520 2.676 5.809 1.000 1.000
2002 2.927 1.613 6.383 3.894 2.638 11.137 4.249 2.170 1.000 1.260 6.858 2.455 1.000 1.490 1.199 1.000 1.000 2.270 3.063 8.457 1.221 2.286 2.744 9.759 7.676 2.367 1.615 2.750 4.142 1.000 1.101
Note: Bold typeface highlights efficient airports.
- 85 -
Assessment of Productive Efficiency of Airports
5.3
Determination of airport productivity
It may not be sufficient to just describe efficiency score based on historical data. For planning an airport, the understanding of factors affecting efficiency score is even more useful. In this case study, causal models for explaining the variations in efficiency score are estimated so that an airport manager can predict future productivity based on given information. In particular, this information is treated as usual exploratory variables. The information may include number of runways, land area, number of gates, noise strategies, proportion of General Aviation (GA) traffic, proportion of international passengers, type of ownership/management, etc. The dependent variable is the efficiency score that indicates the total productivity level of an airport. By the nature of the DEA-Output-VRS model, the value of efficiency scores can only be in the range of 1 to infinity. Because of this special type of limited dependent variable, simple regression is not an appropriate model. Its underlying assumptions are violated, causing inconsistency in estimated coefficients. The Censored Tobit regression model (Tobin, 1958; Maddala, 1983; Amemiya, 1984; Gillen and Lall, 1997; Greene, 2002; Greene, 2003) is more appropriate. In this case, efficiency score of airport y i is represented by Equation (5.2).
?? xi + ? i yi = ? ?1 if y i > 1 if y i ? 1
(5.2)
y i is an efficiency score that is observable for values greater than 1 and is censored for values less than or equal to 1. Efficiency scores of all efficient airports are
- 86 -
Assessment of Productive Efficiency of Airports
censored at 1, regardless of values of independent variables x i . ? and ? i are the coefficients and the error term of the Tobit model respectively. The coefficients ? can be estimated with the Maximum Likelihood (ML) method. ML estimation for the Tobit model involves dividing the observations into two sets. The first set contains uncensored observations. The second set contains censored observations. For y i > 1 , assuming y i ~ N ( ? , ? 2 ) , then the log-likelihood function is written as shown in Equation (5.3).
ln L =
Uncensored
?
ln
1 ? y i ? ? xi ? ? 1 ? ?xi ? ?? ? + ? ln ?? ? ? ? ? ? Censored ? ? ?
(5.3)
where ? and ? are the respective probability and cumulative density functions. Unlike simple regression models, the estimated coefficients cannot be interpreted as marginal effects. Equation (5.4) is used to compute marginal effect of variable k (Gillen and Lall, 1997).
?x ? 1 ?E ( y | x) = Prob(Uncensored | x) ? k = ? ( )? k ? ?x k
(5.4)
Goodness-of-fit may be measured by using R 2 ANOVA , computed by Equation (5.5) . This fit measure takes the variance of the estimated conditional mean divided by the variance of the observed variable (Greene, 2002).
1 n ? ' ?' ? ? ? yi ? y ? ? n i =1 ? ? ?
? 1 n ? ? y y ? ? ? ? i n i =1 ? ? 2
R2 ANOVA =
2
=
Var(Predicted conditional mean) Var(Dependent variable)
(5.5)
- 87 -
Assessment of Productive Efficiency of Airports
Some econometric packages can be used to estimate Censored Tobit Regression model such as EViews (QMS, 2005), LIMDEP (Greene, 2002), and STATA (StataCorp, 2005). In this case study, LIMDEP version 8.0 (Greene, 2002) is used.
5.4 Factors affecting productive efficiency of airports in MASs
The Tobit model has efficiency score ? as the dependent variable. Related literature suggests many possible exploratory variables qualify as independent variables (Gillen and Lall, 1997). In this case study, five groups of independent variables are investigated. The proxy of each group entering the model is essentially based on data availability. First, Airport characteristics are represented here by physical characteristics, basically input measures that are used in the DEA model, i.e., land area (LAND), number of runway (RW) and runway area (RWA). These inputs certainly play a major role in accommodating traffic. However, one should be aware that having more of these inputs does not necessarily mean more outputs. Second, Airport services are mainly represented by outputs of airport operations which consist of number of aircraft movements (AIR) and passengers (PAX). One would expect that more services contribute to higher efficiency. However, this is not necessarily true since efficiency takes into account both inputs and outputs. Accordingly, another group of variables is introduced, i.e., level of utilization. Third, Level of utilization may be a better determinant of operational efficiency since it takes into accounts both input and output measures. This case study considers
- 88 -
Assessment of Productive Efficiency of Airports
seven ratio variables, i.e. annual total passengers/land area (PAX/LAND), annual aircraft movements/land area (AIR/LAND), annual aircraft movements/runway (AIR/RW), annual aircraft movements/runway acreage area (AIR/RWA), annual total
passengers/number of runways (PAX/RW), annual total passengers/runway acreage area (PAX/RWA), and annual total passengers/annual aircraft movements (PAX/AIR). Intuitively, higher values of these ratios should result in more efficient operation. However the interpretation must be very careful since excessive utilization may imply undesirable congestion and delay. Whenever congested airports are classified as efficient in the results they should not be considered appropriate benchmarks. Instead, other less efficient ( ? near 1) may provide more practical benchmarks. Fourth, Market characteristics include target market (e.g., passengers, aircraft operation, cargo, general aviation and military service), market share, market focus (e.g., domestic, international, tourist, business passengers), and irregularity of time periods. Although such characteristics would be interesting to analyze, collecting them for complete cross-national sample is prohibitively expensive. For example, an attempt was made to collect the percentage of general aviation (GA) operations at airports since serving more GA operations tends to lower operational efficiency (Gillen and Lall, 1997). However, such data are not available for many airports. Similarly, the data are not available for other potential variables. To avoid discarding many airports from the analysis, the entering variables have to be limited to available data. Consequently, three variables, namely the percentage of international passengers (INTER), Y2001 and Y2002 are entered the estimation. It is unclear how this proportion affects airport efficiency.
- 89 -
Assessment of Productive Efficiency of Airports
Y2001 and Y2002 aim at capturing anomalies occurring during these two years, notably the 9-11 terrorist attacks. Fifth, Ownership/management characteristics may be another factor affecting airport efficiency. In particular, the study is interested in two contrasting types of ownership/management, namely publicly-owned and privately-owned. There are some good reasons to argue that the latter type yields higher efficiency. For example, a privately-owned entity faces higher risks. This is likely the case when there is little or no subsidy from public funds. This variable is coded as dummy variable equal to one when the airport is privately-owned or there is strong evidence that it is behaving as a commercialized profit-seeking entity. In the US, Stewart International (SWF) is one such example. It has been privatized to the National Express Group Plc. in 1998 (Steward International Airport, 2005). In London MAS, most airports are of this type. Heathrow (LHR), Gatwick (LGW) and Stansted (STN) have been privatized since 1987 and managed by BAA Plc. City/London (LCY) is owned by an Irish entrepreneur, Dermot Desmond (London City, 2005). Table 5.2 compares statistics of some candidate variables between efficient (efficiency score = 1) and inefficient airports (efficiency score > 1). It seems clear that an efficient airport uses less input to produce more output, which can be confirmed by its higher utilization variables. However, it is unclear how the proportion of international passengers associates with performance score. Both groups seem to have comparable figures around 20 - 30%. On the management style, privately-operated airports dominate
- 90 -
Assessment of Productive Efficiency of Airports
in the efficient group. For example, in 2002, five out of seven efficient airports are managed commercially. Table 5.3 shows a Censored Tobit regression model estimation results. The notation is given at the bottom of the Table. This model is a preliminary estimation. It has 11 independent variables, including the constant. Other variables are dropped off for reasons such as high correlation among themselves, being insignificant or having illogical sign. Keep in mind that the lower efficiency score is desirable because it indicates that an airport is more efficient. As a result, a negative sign of estimated coefficient, such as 0.0173 of passengers per aircraft movement (PAX/AIR), contributes to higher efficiency. In this model, passenger market share (PAXSHARE) has an illogical sign. The expected sign should be negative since higher share are more likely associated with higher efficiency. Year 2001 for which it was aimed to test whether September-11 terrorist attack had any effect on efficiency scores; turns out to be insignificant. Possibly, it did not have immediate effect in that year but propagated to the next year 2002, as Y2002 variable is significant. This means that in 2002, an airport became efficient slightly easier than normal. The marginal effects are also shown next to the right of the coefficient’s column. Basically, it indicates the change in efficiency score with respect to unit change of an independent variable. For example, if an airport were able to increase its share of aircraft movements by one percent, it would earn 1.9656 additional units on its efficiency score.
- 91 -
Assessment of Productive Efficiency of Airports
Table 5.2 Comparisons of statistics between efficient and inefficient airports
Variables Land acreage area (LAND) Number of runways (RW) Runway acreage area (RWA) Annual aircraft operations (AIR) Annual total passengers (PAX) AIR/LAND AIR/RW AIR/RWA PAX/LAND PAX/RW PAX/RWA PAX/AIR % of international passenger (INTER) Management style (MANAGE) 2000 Efficient Inefficient (N = 12) (N = 60) 2,261 2,986 (2,896) (3,523) 2.17 2.80 (1.80) (1.12) 67.13 82.05 (68.23) (49.54) 299,963 221,768 (294,845) (159,866) 23,772,828 12,121,727 (28,243,030) (14,946,553) 348 138 (252) (130) 126,123 74,688 (61,830) (35,748) 5,505 2,897 (3,833) (1,496) 12,235 5,426 (11,045) (6,351) 9,442,790 3,945,399 (9,057,704) (4,267,402) 290,332 123,563 (203,776) (118,702) 62.31 45.64 (60.97) (40.42) 33.26 22.81 (39.73) (30.15) 6/6 51/9 2001 Efficient Inefficient (N = 14) (N = 58) 2,161 3,035 (2,714) (3,567) 2.21 2.81 (1.67) (1.13) 68.66 82.32 (64.70) (49.82) 287,397 208,485 (267,031) (150,907) 23,992,367 10,797,422 (26,255,156) (12,955,868) 340 121 (248) (96) 119,697 70,148 (55,985) (33,350) 5,186 2,678 (3,559) (1,242) 13,598 4,637 (10,672) (5,220) 9,399,439 3,513,259 (8,397,929) (3,619,718) 289,295 110,337 (184,177) (103,842) 70.12 42.80 (68.16) (33.24) 27.97 23.88 (37.28) (30.59) 8/6 49/9 2002 Efficient Inefficient (N = 12) (N = 60) 1,615 3,115 (2,163) (3,579) 2.42 2.78 (1.73) (1.18) 73.43 82.49 (68.60) (51.68) 307,916 197,728 (271,078) (142,169) 25,620,505 10,763,918 (27,669,409) (12,608,237) 380 112 (223) (89) 117,094 68,035 (58,651) (30,842) 5,330 2,559 (3,665) (1,123) 15,287 4,666 (11,095) (5,346) 8,670,840 3,668,420 (7,874,927) (3,857,687) 289,748 112,888 (204,542) (104,246) 63.85 46.23 (63.97) (37.68) 24.39 25.09 (36.06) (31.30) 7/5 50/10
Note: N is the number of airports. Standard deviations are shown in parentheses. For Management style, 7/5 in 2002 represents number of noncommercial (7) and commercial airports (5) respectively.
- 92 -
Assessment of Productive Efficiency of Airports
Table 5.3 Censored Tobit regression: preliminary model estimation results
Variables Constant MANAGE PAX/AIR INTER AIRSHARE PAXSHARE Y2001 Y2002 PAX/RWA (million/acre) AIR/AREA (1,000/acre) AIR/RWA (1,000/acre)
Preliminary model 2 R ANOVA = 0.6236 Coefficient 6.8560** (17.901) -0.5964* (-2.409) -0.0173** (-2.917) 0.0236** (6.186) -2.3715** (-3.343) 0.2955 (0.583) -0.3108 (-1.451) -0.4226* (-1.973) -4.4344* (-2.260) -6.0155** (-6.731) -0.5530** (-5.237) Marginal effect 5.6827** (15.802) -0.4943* (-2.419) -0.0144** (-2.899) 0.0195** (6.040) -1.9656** (-3.327) 0.2450 (0.583) -0.2576 (-1.452) -0.3503* (-1.974) -3.6755* (-2.265) -4.9861** (-6.738) -0.4584** (-5.255)
Number of airports = 216 during 2000 - 2002
Notation: Dependent variable = Efficiency score MANAGE = 1 if privately-owned or commercially managed, otherwise = 0 PAX/AIR = Average number of passengers per aircraft movement INTER = Percentage of international passenger (%) AIRSHARE = Market share of annual aircraft movements PAXSHARE = Market share of annual total passengers Y2001 = 1 if compute performance score in year 2001, otherwise = 0 Y2002 = 1 if compute performance score in year 2002, otherwise = 0 PAX/RWA = Annual total passengers per runway area (million/acre) AIR/AREA = Annual aircraft movements per land area (103/acre) AIR/RWA = Annual aircraft movements per runway area (103/acre) ** Estimated coefficient is significant at the 0.01 level (one-tailed) * Estimated coefficient is significant at the 0.05 level (one-tailed)
- 93 -
Assessment of Productive Efficiency of Airports
Table 5.4 Censored Tobit regression: proposed model estimation results
Variables Constant MANAGE PAX/AIR INTER AIR50 PAX50 Y2001 Y2002 PAX/RWA (million/acre) AIR/AREA (1,000/acre) AIR/RWA (1,000/acre)
Proposed model 2 R ANOVA = 0.6206 Coefficient 6.5319** (18.616) -0.6494** (-2.637) -0.0169** (-2.899) 0.0241** (6.424) -0.7025** (-2.727) -0.5858* (-2.405) -0.3374 (-1.593) -0.4634* (-2.184) -4.5793* (-2.349) -5.6768** (-6.551) -0.5760** (-5.515) Marginal effect 5.4222** (16.326) -0.5391** (-2.651) -0.0140** (-2.880) 0.0200** (6.259) -0.5831** (-2.716) -0.4863* (-2.411) -0.2800 (-1.594) -0.3846* (-2.186) -3.8013* (-2.354) -4.7123** (-6.575) -0.4781** (-5.532)
Number of airports = 216 during 2000 - 2002
Notation: Dependent variable = Efficiency score MANAGE = 1 if privately-owned or commercially managed, otherwise = 0 PAX/AIR = Average number of passengers per aircraft movement INTER = Percentage of international passenger (%) AIR50 = 1 if the market share of aircraft movements > 50%, otherwise = 0 PAX50 = 1 if the market share of annual total passengers > 50%, otherwise = 0 Y2001 = 1 if compute performance score in year 2001, otherwise = 0 Y2002 = 1 if compute performance score in year 2002, otherwise = 0 PAX/RWA = Annual total passengers per runway area (million/acre) AIR/AREA = Annual aircraft movements per land area (103/acre) AIR/RWA = Annual aircraft movements per runway area (103/acre) ** Estimated coefficient is significant at the 0.01 level (one-tailed) * Estimated coefficient is significant at the 0.05 level (one-tailed)
- 94 -
Assessment of Productive Efficiency of Airports
Table 5.4 shows the final model estimation results. It is the proposed model that may be used for predicting the efficiency score. Instead of using market share like the preliminary model in Table 5.3, the proposed model considers market dominance as an exploratory variable in order to capture the effects of market characteristics. Market dominance is represented by a dummy variable, equal to 1 if an airport has market share more than 50%. There are two market dominance variables i.e. dominance by aircraft movements share (AIR50) and dominance by passenger share (PAX50). As observed previously, market dominance tends to be associated with efficient airports. They turn out to be significant, as observed. The reason may be that an airport is in a better position to utilize its inputs, given higher traffic. Most of the estimated coefficients, except Y2001, are meaningful and significant at above the 95% confidence level. Privately-operated airports (MANAGE = 0.6494) tend to be more efficient than their publicly-operated counterparts, possibly due to higher risk and higher accountability of the management. All utilization ratio variables contribute to higher efficiency, as expected. Proportion of international passengers is negatively associated with the efficiency score. A higher proportion of international passengers lead to lower efficiency. There might be some effect from anomaly in 2001 as Y2001 becomes stronger, though not yet significant enough. The model captures the anomaly in 2002, a year after September-11, where variable Y2002 is significant. The negative sign indicates that an airport becomes efficient slightly more easily in 2002. The marginal effect suggests that for every additional million passenger per acre, an airport would be more efficient by 3.8013 units. In summary, the first attempt to assess productivity of airports operating in multiple airports systems is presented in this case study. The samples consist of 72 airports in 25
- 95 -
Assessment of Productive Efficiency of Airports
MASs worldwide. The analysis period is during 2000 – 2002. The DEA-Output-VRS model is used as an analytical tool. Five indicators were considered, i.e., land area, number of runways, runway area, number of passengers and number of aircraft movements. The results indicate that there are two groups of efficient or highly productive airports, coined by the busy and the compact. The busy group consists of market leaders in large MASs such as O’Hare International (ORD), Los Angeles International (LAX) and Heathrow/London (LHR). Airports in the compact group are mostly alternative airports with relatively small land area and one or two runways. The reason that both are classified as efficient airports is mainly due to their relatively higher runway utilization. In this respect, larger size of airport does not guarantee high efficiency. It is also found that some airports are under utilized such as Montreal-Mirabel (YMX), Glasgow Prestwick International (GLA), Schoenefeld/Berlin (SXF), and Tempelhof/Berlin (THF). In fact, Montreal-Mirabel (YMX) is a case study of an unsuccessful airport in textbooks (Caves and Gosling, 1999; de Neufville, 1995; de Neufville and Odoni, 2003). Schoenefeld/Berlin (SXF), and Tempelhof/Berlin (THF) and Tegel (TXL, another airport in the Berlin MAS) are planned to be consolidated in 2011. Construction is underway (Berlin Brandenburg International, 2005). In this sense, the proposed models in this case study are useful in pointing out over-investment. Furthermore, a productivity prediction model was developed by using the Censored Tobit Regression. It is found that factors such as utilization of land area and runway area, passengers per aircraft movement, market dominance and privately-operated management style contribute to the enhancement of productivity. Meanwhile a higher proportion of international passengers tends to reduce the productivity. The model also captures anomaly effects in the year 2002 such that an airport could become efficient slightly more easily with
- 96 -
Assessment of Productive Efficiency of Airports
the same utilization rate, possibly due to a global drop in air traffic after September 11 terrorist attacks. Given some planned measures, the model can be used to predict future total productivity of an airport which should be very useful as a tool for planning airport business in a competitive market. An important observation from this case study is that an efficient airport needs to be very busy. Some efficient airports are in fact constrained and show sign of undesirable congestion. Like previous airport studies listed in Tables 2.2, 2.3 and 2.4, the downside of facilities and quality of service are still out of consideration. This may be due to the inapplicability of the DEA models to take into account such measures. In the next case study, this issue will be addressed by considering joint production of desirable and undesirable outputs from airport operations. Given that delay data are available for U.S. airports, the case study will be to assess productivity of U.S. airports using delays as a proxy of undesirable outputs. The results should provide a more complete total productivity index.
- 97 -
Assessment of Productive Efficiency of Airports
CHAPTER 6 CASE STUDY 2 PRODUCTIVITY OF U.S. AIRPORTS
Results from case study 1 indicate that ignoring the downside of facilities and quality of service from the assessment typically leads to the conclusion that efficient airports must be very busy. Such results may provide an inappropriate benchmark for managing other airports. In practice, other important output measures reflecting quality of services are always taken into account. Among them, delay is perhaps a major concern. This chapter describes the second case study where the productivity assessment will take undesirable outputs, i.e., delays into consideration.
6.1 Modeling airport operations
An airport may be viewed as a production unit. As a result, airport operation can be modeled as a production process that requires some inputs for running day-to-day operations in order to produce some target outputs. Necessary inputs include production factors such as capital and labor. Most airport managers set target to maximize movement of aircrafts, passenger throughput and quantity of cargo transported. These outputs are highly desirable and the primary reason for building an airport. However, production of these outputs is always constrained by capacity. As the air traffic volume increases, the likely by-product output is higher delays. An airport manager also wants to make sure that the undesirable byproducts from the airport operation are being kept at the minimal possible level. Furthermore, an airport is bound to comply with rules and regulations which ensure that its operation does
- 98 -
Assessment of Productive Efficiency of Airports
not create unacceptable externalities, notably noise and pollution. In this situation, where on one hand an airport manager wants to maximize desirable outputs and on the other hand minimize undesirable outputs, the directional output distance function is perfectly applicable.
6.2 Inputs and outputs of airport operations
The formulation of the directional output distance function as shown in linear programming in (3.7) requires data on quantity of inputs, desirable outputs, and undesirable outputs, ( x k , y k , b k ) for individual airports. The selection of inputs and outputs is an important decision issue in the assessment of airport productivity. The general suggestion is to include all important measures that are in the interest of the management. Such measures should be common for all airports so that the performance would provide meaningful interpretation. In practice, the main problem is the availability of the data across all airports rather than model limitations. After all, three common physical inputs are considered in this analysis:
x1 = Land area, acre x 2 = Number of runways
x3 = Runway area, acre These inputs may be rather limited due to the availability of data, but they are necessary infrastructure for all airports. Land area (acre) represent considerable share of capital investment that an airport should fully utilize. Number of runways counts all existing runways at the airport regardless of their utilization level. Runway area is the summation of the length x width product of all runways. Runway area is included to reflect the effect of design configuration such as length, width, and separation on productivity. - 99 -
Assessment of Productive Efficiency of Airports
For the set of desirable outputs, it is assumed that airport managers aim at producing the following three outputs as much as possible:
y1 = Non-delayed flights y 2 = Passengers
y 3 = Cargo throughput All outputs are considered on annual basis. A flight is counted as non-delayed if it is operated no later than 15 minutes from the scheduled time according to Federal Aviation Administration (FAA)’s definition (FAA, 2005 – 2006). The non-delayed flights include all kinds of movements, i.e., commercial aircrafts, cargo aircrafts, general aviation, and others. The number of passengers counts both arriving and departing passengers for all type of commercial passengers, i.e., international, domestic and direct transit passengers. Cargo throughput is measured in metric tones of both loaded and unloaded freight which includes international freight, domestic freight, and mail. On the set of undesirable output, it is assumed that an airport manager wants to minimize the following two outputs:
b1 = delayed flights b2 = time delays
Again, both undesirable outputs are on annual basis. Delayed flights are those movements that are operated more than 15 minutes later than the scheduled time. One might argue that delayed flights are not necessarily undesirable from the economics perspective because the passengers still can get to their destinations as they wish. However, airport
- 100 -
Assessment of Productive Efficiency of Airports
managers and passengers may have a different perspective. If they have a choice,, they probably prefer to avoid experiencing delayed flights. As a result, the delayed flights are treated as undesirable outputs. As a matter of fact, individual delayed flights may incur different delayed times; more of delayed flights does not necessarily mean lower operational efficiency. It also depends on time duration of total delays. Therefore, the time delay (b2 ) is included to reflect another perspective of delays. Time delays are the accumulation of delays experienced by individual delayed flights. Given the above-selected three inputs, three desirable outputs and two undesirable outputs, the assessment may be perceived as the measurement of productive efficiency of airside operation. A question may arise regarding the selection of land area as an input measure. One might argue that acquisition of land area is not purely for improving airside operation, but for other purposes as well. For instance, an airport may prefer to possess more land than needed as noise buffer or for future expansion. It is possible that an airport may seek investment opportunities beyond aeronautical activities such as land value appreciation, or even commercial development. Nevertheless, land area is somehow under management control and the manager can affect productivity of airside operation by managing it efficiently. More importantly, the assessment of productivity is an exploratory analysis which provides information for further judgment, not an ultimate conclusion for implementation. If an airport is found to be inefficient because of inefficient use of any measure, the manager still can reason against the findings. To account for its effect on productivity, the scenario analysis will be done for with and without consideration of land area.
- 101 -
Assessment of Productive Efficiency of Airports
6.3 6.3.1
Sample characteristics Size of sample
The case study is to assess the relative productivity of major U.S. commercial airports and examine the impact of the inclusion of undesirable outputs on the productivity, ranking and productivity index. Due to readily available data, part of samples are taken from previous airport productivity studies (Pathomsiri and Haghani, 2004; Pathomsiri, Haghani and Schonfeld, 2005; Pathomsiri, Haghani, Dresner and Windle, 2006a). Additional samples are collected in order to increase the sample size. Overall, there are 56 airports in the dataset. This is a relatively high number compared to most previous studies (Abbott and Wu, 2002; Adler and Berechman, 2001; Bazargan and Vasigh, 2003; Fernandes and Pacheco, 2001, 2002, 2005; Gillen and Lall, 1997, 1998; Hooper and Hensher, 1997; Martin and Roman, 2001; Martin-Cejas (2002); Murillo-Melchor, 1999; Nyshadham and Rao, 2000; Oum, Yu and Fu, 2003; Pacheco and Fernandes 2003; Parker, 1999; Pels, Nijkamp and Rietveld, 2001, 2003; Salazar de la Cruz,1999; Sarkis, 2000; Sarkis and Tulluri, 2004; Yoshida, 2004; Yu, 2004). Table 6.1 lists all 56 airports along with the International Civil Aviation Organization (ICAO) airport codes. Nevertheless, the number of samples is checked against several applicable rules of thumb to guarantee the sufficiency and meaningful interpretation. For non-parametric approach, DEA provides a very good guideline. As mentioned earlier in Chapter 4, an excessive number of measures with respect to the total number of DMUs may deteriorate the discriminatory power of DEA model. The larger the number of input and output measures for a given number of airports the less discriminatory the DEA model becomes. Given a certain set of samples, this means that the addition of measures will reduce the discriminatory power
- 102 -
Assessment of Productive Efficiency of Airports
of the DEA model. Essentially, this is because it is possible that an airport may dominate all others on one measure, which in turn makes it looked equally efficient to other efficient airports. This is a major issue encountered by Parker (1999), Adler and Berechman (2001) and Yu (2004). To avoid this problem, the straightforward way is to guarantee that there will be sufficient number of airports for comparison with, regarding any measures. Boussofiane, Dyson and Thanassoulis (1991) recommend that the total number of DMUs be much greater than the number of inputs times the number of outputs. Compared to this analysis where three inputs and five (three desirable plus two undesirable) outputs are selected for the assessment of airport productivity; the number of samples needs to be much more than 3 x 5 or 15 airports in order to reduce the chance that an airport may be too dominant compared to the others on a particular measure. According to the recommendation, the sample size of 56 airports is deemed satisfactory. To avoid losing discriminatory power, Cooper, Seiford and Tone (2000: page 103) recommend that the desired number of DMUs exceed m + s several times. They suggest a more stringent rule of thumb in the following formula (Cooper, Seiford and Tone, 2000: page 252). n ? max{m x s, 3(m + s )} (6.1)
where n is the number of DMUs, while m and s are the numbers of input and output measures, respectively. Substituting m and s , yields the minimum number of samples: n ? max{3 x 5, 3(3 + 5)} = max{15,24} = 24 (6.2)
- 103 -
Assessment of Productive Efficiency of Airports
Again, the sample size of 56 satisfies this recommendation. After all, it can be concluded that the sample size is sufficient for the analysis. Figure 6.1 shows the locations of the sample. Note that these 56 airports are major U.S. commercial airports which regularly appeared at the top of the published statistics from trade publications (ACI, 2002 – 2004). Many of them are in the top twenty according to statistics on annual aircraft movement, passengers, and cargo throughput.
Alaska
Hawaii
Figure 6.1 Locations of 56 airports 6.3.2 Analysis period
In order to obtain timely results, a recent 4-year panel data for the years 2000 through 2003 are collected. Coincidently, the period spans the critical time before and after September 11 terrorist attacks which severely affected the aviation industry worldwide. As a result, it allows for analyzing airport productivity to understand its effect on productivity of U.S. airports.
- 104 -
Assessment of Productive Efficiency of Airports
6.3.3 Data source and definition
Input measures were collected from the Airport Master Record database (FAA, 2004) which records physical characteristics of U.S. airports. The database is revised on a regular basis to reflect input changes at airports (e.g., the addition of runways). The data was verified with airport managers, airports’ websites and reports, to determine if there were major changes in runway characteristics during the analysis period. There was no change in land area during the analysis period. However, it was found that some airports had improved their runways. For example, George Bush Intercontinental (IAH) expanded and extended runway 15R/33L from 6038’ x 100’ to 10000’ x 150’ in 2002. Detroit Metropolitan Wayne County (DTW) opened its 6th runway on December 11, 2001. The number of runways and runway area were then edited accordingly. They were computed precisely by the time the runway improvement was in service during the year, rounding down in month. In the case of DTW, it was recorded that the airport has 5 and 6 runways in 2001 and 2002 respectively. All airports in the dataset are well-established, having been built and served their respective markets for a number of years. This knowledge helps to relieve concerns about possible sudden productivity drops during the early years after initial lumpy investments. It may be assumed, with caution, that any temporal changes in productivity that might be observed result from operational performance. Data on the three desirable outputs, i.e., annual statistics on number of passengers, cargo throughput, and aircraft movements are published by the Airports Council International (ACI, 2002 - 2004). Note that aircraft movements include both delayed and non-delayed flights. Given the assumption that an airport manager wishes to maximize only non-delayed flights, the data on delayed flights needed to be collected as well.
- 105 -
Assessment of Productive Efficiency of Airports
There are two available delay database; the Airline Service Quality Performance (ASQP) database maintained by the Bureau of Transportation Statistics (BTS) and the FAA’s Operational Network (OPSNET) database (FAA, 2005). In both databases, a flight is counted as delayed if it is operated more than 15 minutes later than the scheduled time, according to the FAA’s definition, The ASQP database contains delays reported by 18 certified U.S. air carriers that have at least one percent of total domestic scheduled-service passenger revenues, plus other carriers that report on a voluntarily basis. The reports cover non-stop scheduled-service flights between points within the U.S., including territories. So far, the carriers report monthly on operations at 31 U.S. airports that account for at least one percent of the nation’s total domestic scheduled-service enplanements. The up-to-date list of reportable airlines and airports can be viewed at BTS’s website (BTS, 2006). In June 2003, the airlines began reporting the causes of delays in five broad categories, namely (BTS 2004 - 2006; FAA 2006): 1) Air carrier: the cause of the cancellation and delay was due to circumstances within the airline’s control (e.g., maintenance or crew problems, aircraft cleaning, baggage loading, fueling, etc.). 2) Extreme weather: significant meteorological conditions (actual and forecast) that, in the judgment of the carrier, delays or prevents the operation of a flight (e.g., tornado, blizzard, hurricane, etc.). 3) National aviation system (NAS): delays and cancellations attributable to the national aviation system that refer to a broad set of conditions – non-extreme weather conditions, airport operations, heavy traffic volume, air traffic control, etc. - 106 -
Assessment of Productive Efficiency of Airports
4) Late-arriving aircraft: a previous flight with the same aircraft arrived late, causing the present flight to depart late. 5) Security: delays or cancellations caused by evacuation of a terminal or concourse, re-boarding of aircraft because of security breach, inoperative screening equipment and/or long lines in excess of 29 minutes at screening areas. In broad classification, delays caused by 1) air carrier, 2) extreme weather and 4) latearriving aircraft are attributable to operation of airlines. Meanwhile, 3) NAS and 5) security delays are from airport operation. According to the recent statistics (BTS, 2005), the proportion of delay causes are 25.6, 4.8, 39.6, 29.7 and 0.3 % for air carrier, extreme weather, NAS, late-arriving aircraft and security respectively. This means that about 60% of delayed flights are caused by airline operation and the rest of 40% are from airport operation. The other delay database, OPSET is an official source of historical National Aviation System (NAS) air traffic delays and covers a broad set of causes, such as non-extreme weather conditions, airport operations, heavy traffic volume, terminal volume, air traffic control, runway, equipment and others. The OPSNET database is chosen because of two main reasons. First, the analysis focuses on airport operation, rather than airline operations being the source of the delay as in the BTS database. Second, the OPSET database is more complete, covering all flights, all flight types (both domestic and international), all airlines (U.S. and non-U.S. carriers), and all airports in the sample. On the contrary, if ASQP database was used, there would be a large number of under-reported flights since it is based on the sampling report. The numbers of under-reported flights are varying depending on airports. For instance, ASQP database reported only 79.97% of flights at O'Hare International airport in 2003. The missing 21.03% of flights were resulted from non-reportable airlines,
- 107 -
Assessment of Productive Efficiency of Airports
international flights, unclear report, incomplete data, missing value and so on. The figure is quite different at the Baltimore/Washington International airport where 50.86% of total movements are not reported. Inherently, the ASQP database contained biased delay data due to different sampling rates across airports.
- 108 -
Assessment of Productive Efficiency of Airports
Table 6.1 List of 56 US airports under consideration and their outputs in 2003
Airport name 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Hartsfield-Jackson Atlanta International O'Hare International Los Angeles International, CA Dallas/Fort Worth International, TX Denver International, CO Phoenix Sky Harbor International, AZ McCarran International, NV George Bush Intercontinental, TX Minneapolis/St. Paul International, MN Detroit Metropolitan Wayne County, MI John F. Kennedy International, NY Miami International, FL Newark Liberty International, NJ San Francisco International, CA Orlando International, FL Seattle Tacoma International, WA Philadelphia International, PA Charlotte/Douglas International, NC Boston Logan International, MA LaGuardia, NY Covington/Cincinnati/Northern Kentucky International, KY Lambert-St. Louis International, MO Baltimore/Washington International, MD Honolulu International, HI Salt Lake City International, UT Midway International, IL Fort Lauderdale - Hollywood International, FL Washington Dulles International, VA Tampa International, FL San Diego International, CA
Airport code ATL ORD LAX DFW DEN PHX LAS IAH MSP DTW JFK MIA EWR SFO MCO SEA PHL CLT BOS LGA CVG STL BWI HNL SLC MDW FLL IAD TPA SAN
Total passengers 79,086,792 69,508,672 54,982,838 53,253,607 37,505,138 37,412,165 36,285,932 34,154,574 33,201,860 32,664,620 31,732,371 29,595,618 29,431,061 29,313,271 27,319,223 26,755,888 24,671,075 23,062,570 22,791,169 22,482,770 21,228,402 20,427,317 20,094,756 19,732,556 18,466,756 18,426,397 17,938,046 16,767,767 15,523,568 15,260,791
Cargo (tons) 798,501 1,510,746 1,833,300 667,574 325,350 288,350 82,153 381,926 315,987 220,246 1,626,722 1,637,278 874,641 573,523 193,037 351,418 524,485 140,085 363,082 28,402 392,695 115,574 235,576 421,930 216,870 23,266 156,449 285,352 93,457 135,547
Aircraft movements 911,723 928,691 622,378 765,296 499,794 541,771 501,029 474,913 510,382 491,073 280,302 417,423 405,808 334,515 295,542 354,770 446,529 443,394 373,304 374,952 505,557 379,772 299,469 319,989 400,452 328,035 287,593 335,397 233,601 203,285
Non delayed flights 874,203 859,506 620,178 755,873 498,469 529,971 494,332 458,924 503,049 486,231 274,217 412,559 381,159 325,205 294,300 352,786 432,902 440,079 369,452 357,054 498,577 374,984 297,733 319,976 399,680 323,041 283,700 329,552 232,471 202,506
Delayed flights 37,520 69,185 2,200 9,423 1,325 11,800 6,697 15,989 7,333 4,842 6,085 4,864 24,649 9,310 1,242 1,984 13,627 3,315 3,852 17,898 6,980 4,788 1,736 13 772 4,994 3,893 5,845 1,130 779
- 109 -
Assessment of Productive Efficiency of Airports
Table 6.1 List of 56 US airports under consideration and their outputs in 2003 (Continued)
Airport name 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56
Pittsburg International, PA Ronald Reagan Washington National, DC Oakland International, CA Portland International, OR Memphis International, TN Mineta San Jose International, CA Cleveland Hopkins International, OH Kansas City International, MO Louis Armstrong New Orleans International, LA John Wayne, CA William P. Hobby, TX Ontario International, CA Port Columbus International, OH Albuquerque International Sunport Airport, NM Palm Beach International, FL Jacksonville International, FL Anchorage International, AK Bob Hope, CA Norfolk International, VA Long Beach, CA Birmingham International, AL Pensacola Regional, FL Palm Spring International, CA Jackson International, MS Santa Barbara, CA Stewart International, NY Total
Airport Code PIT DCA OAK PDX MEM SJC CLE MCI MSY SNA HOU ONT CMH ABQ PBI JAX ANC BUR ORF LGB BHM PNS PSP JAN SBA SWF
Total passengers 14,266,984 14,214,803 13,548,363 12,395,938 11,437,307 10,677,903 10,555,387 9,715,411 9,275,690 8,535,130 7,803,330 6,547,877 6,252,061 6,051,879 6,010,820 4,883,329 4,791,431 4,729,936 3,436,391 2,875,703 2,672,637 1,361,758 1,246,842 1,215,093 752,762 393,530 1,094,725,865
Cargo (tons) 121,536 5,774 597,383 239,265 3,390,515 108,622 95,761 136,687 80,831 12,050 5,775 518,710 10,766 71,599 18,300 70,650 2,102,025 44,654 32,283 50,873 34,184 4,569 103 10,957 2,825 19,024 22,599,243
Aircraft movements 361,329 250,802 342,871 267,052 402,258 198,082 258,460 170,758 137,312 350,074 242,635 146,413 237,979 221,003 171,692 121,143 277,361 178,079 121,373 338,807 154,849 127,197 93,068 79,377 152,485 112,284 18,781,482
Non delayed flights 360,619 249,056 342,567 266,872 400,683 197,855 256,993 170,722 137,094 348,475 242,084 146,212 237,915 220,962 169,836 121,043 277,165 177,902 121,330 338,727 154,781 127,195 93,032 79,376 152,434 112,277 18,485,876
Delayed flights 710 1,746 304 180 1,575 227 1,467 36 218 1,599 551 201 64 41 1,856 100 196 177 43 80 68 2 36 1 51 7 295,606
- 110 -
Assessment of Productive Efficiency of Airports
The OPSNET database recorded both the number of delayed flights and time delays, which are two undesirable outputs in the analysis. Given the number of delayed flights from OPSNET database, the number of non-delayed flights (one of desirable outputs) is simply the difference between aircraft movements and number of delayed flights. Table 6.1 presents output measures of 56 airports in 2003. The figures are ordered by number of annual passengers. At the top of the list, Hartsfield-Jackson Atlanta (ATL) is the busiest airport in terms of passengers. O’Hare International (ORD) serviced the highest number of aircraft movements. Memphis International (MEM), the FedEx hub, had the highest cargo throughput. On the downside, ORD experienced the highest number of delayed flights. As shown in Figure 6.2, the number of delayed flights tends to increase with number of passengers serviced at the airport. In fact, there are always externalities inherent in airport operations, notably delay and noise that increase, ceteris paribus, with airport volume. These externalities are also outputs from the production process, although undesirable. In Figure 6.3 density of aircraft movement (number of flights per runway area) is plotted against average delay per passenger, computed for 56 airports in the dataset during 2000 – 2003. This graph shows that higher density of traffic is associated with higher average delay. According to these results, airport efficiency may come at the cost of numerous delays. This situation may be undesirable from the viewpoints of airports, regulators, airlines, and passengers.
- 111 -
Assessment of Productive Efficiency of Airports
Delayed Flights V.S. Passengers
120,000 100,000 Delayed Flights 80,000 60,000 40,000 20,000 0 0 20,000,000 40,000,000 60,000,000 80,000,000 Passengers
Figure 6.2 Scatter plot between number of delayed flights and number of passengers, 2003
Delay/Passenger V.S. Movements/Runway Area
80,000 Delay/Passenger (minutes) 70,000 60,000 50,000 40,000 30,000 20,000 10,000 0 0 2,500 5,000 7,500 10,000 12,500 15,000 17,500 Movements/Runway Area (flights/acre)
Figure 6.3 Scatter plot between delay/passenger and density of movements, 2000 – 2003
- 112 -
Assessment of Productive Efficiency of Airports
Table 6.2 summarizes the descriptive statistics on input and output measures. All measures show large standard deviations suggesting that airports in the sample vary in both scale and scope of operations. During the study period, all airports experienced at least some flight delays and delays are positively associated with air traffic volume, as scatter plot shows in Figure 6.2. This suggests that delays are important undesirable byproducts that should be taken into consideration when assessing airport productivity. In the next chapter, results from assessing productivity of 56 airports by the directional output distance function will be presented. The assessment will be discussed with respect to operational efficiency of individual airports. In addition, changes of productivity over analysis period, as well as the impact of the inclusion of undesirable outputs (i.e., delayed flights and time delays) on productivity and ranking of airports will be discussed.
- 113 -
Assessment of Productive Efficiency of Airports
Table 6.2 Descriptive statistics of samples 2000 – 2003
Input Statistics Land area (acre) 501 33,422 32,921 4,381 2,650 5,298 Number of runways 1.00 7.00 6.00 3.35 3.00 1.21 Runway area (acre) 24.60
Desirable outputs Total passengers 362,017 # of non delayed flights 79,376 874,203 794,827 343,324 326,086 176,881 Cargo throughput (tons) 74 3,390,800 3,390,726 401,667 171,349 591,702
Undesirable outputs # of delayed flights 1 96,346 96,345 5,818 1,355 11,917 Time delays (minutes) 20 5,398,921 5,398,901 259,558 57,200 611,968
Minimum Maximum Range Mean Median Standard deviation
305.87 80,162,407 281.26 79,800,390 104.21 20,009,558 99.56 16,225,655 51.65 16,924,416
- 114 -
Assessment of Productive Efficiency of Airports
CHAPTER 7 CASE STUDY 2 RESULTS AND DISCUSSION
7.1 Impact of the inclusion of undesirable outputs
For each year (2000 – 2003), the directional output distance function in formulation (3.7) is solved 56 times; i.e., one time for each airport, to determine the optimal efficiency score ? * . ? * , or the distance from the efficient frontier, measures the level of inefficiency. An efficient airport will have ? * = 0. The higher value of ? * shows greater inefficiencies. There are several cases in the case study. Case 1 is for the case that considers delayed flights and time delays as undesirable outputs. The sets of inputs and outputs are as follows: Case 1: with consideration of undesirable outputs Input = {land area, number of runway, runway area} Desirable outputs = {non-delayed flights, passengers, cargo} Undesirable outputs = {delayed flights, time delays} In order to analyze the effect of the inclusion of undesirable outputs on productive efficiency, a model that ignores undesirable outputs is also solved. Suppose that Case 2 is for the case that does not consider undesirable outputs. The sets of inputs and outputs are as follows: - 115 -
Assessment of Productive Efficiency of Airports
Case 2: without undesirable outputs or Case 1 – {delayed flights, time delays} Input = {land area, number of runway, runway area} Desirable output = {aircraft movements, passengers, cargo} Undesirable output = {none} Aircraft movements are the total number of operations both landings and takeoffs, regardless of delay status. To assess airport productivity for this case, the model (3.7) needs to be modified accordingly. Specifically, the model in (3.7) is modified by taking out the constraints associated with undesirable outputs, resulting in the following model: max ? s.t.
k?K
?? ??
k
y km ? (1 + ? ) y km , m = 1,..........., M , x kn ? x kn , n = 1,.............., N ,
(7.1)
k?K
k
? k ? 0, k = 1,.........., K
The model described in (7.1) is also solved 56 times, each time for an individual airport. Table 7.1 shows efficiency scores for the two cases annually during the period 2000 – 2003. In the Table, airports are ordered alphabetically by their corresponding airport codes. An efficient airport must yield a score of zero, implying that increases in desirable outputs or decreases in undesirable outputs and inputs from current levels are not necessary. In Table 7.1 the efficient airports are highlighted with bold typeface. Several observations can be made from the results. They are discussed in the following sections.
- 116 -
Assessment of Productive Efficiency of Airports
7.1.1
Classification of efficient airports
When delayed flights and time delays are ignored (Case 2), the results are typical of those reported in past studies which suggest that operational efficiency is associated with busy airports (Adler and Berechman, 2001; Bazargan and Vasigh, 2003; Fernandes and Pacheco, 2002; Gillen and Lall, 1997, 1998; Martin and Roman, 2001; Oum and Yu, 2003; Pacheco and Fernandes, 2003; Pathomsiri and Haghani, 2004; Pathomsiri, Haghani and Schonfeld, 2005; Pathomsiri, Haghani, Dresner and Windle, 2006; Pels, Nijkamp and Rietveld, 2003; Sarkis, 2000; as well as Sarkis and Talluri, 2004. As is evident from the 2003 data, six efficient airports are also very busy. For examples, Hartsfield-Jackson Atlanta (ATL) and Memphis (MEM), respectively, are the busiest airports in the world in terms of number of passengers and cargo throughput. LaGuardia (LGA) is one of the most chronically congested airports in the U.S. (CRA, 2001). John Wayne airport (SNA) constrains the number of passengers using its facilities. Other well-known busy airports such as Anchorage International (ANC), Newark Liberty International (EWR), John F. Kennedy International (JFK), Midway International (MDW), Miami International (MIA), O'Hare International (ORD), Seattle Tacoma International (SEA) and Lambert-St. Louis International (STL), though not classified as efficient, show very low inefficiency level. They all earn relatively low efficient scores (less than 0.5). The implication of these results is that an airport that is very busy or constrained is generally determined to be efficient.
- 117 -
Assessment of Productive Efficiency of Airports
Table 7.1 Efficiency scores for Case 1 and Case 2
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM 2000 Case 1 Case 2 0.2034 2.6026 0.2314 0.0000 0.0000 0.0000 2.0490 0.0000 0.5044 0.9495 0.0108 1.4154 1.7999 0.0000 0.4323 1.4892 0.3502 0.6097 1.0697 0.0000 0.4044 0.5496 0.0102 0.7610 0.2956 1.2381 0.7486 0.9718 0.6938 1.2993 0.0892 0.0000 0.5941 0.8739 1.6907 0.0000 0.3367 2.0979 0.4777 0.6289 0.5179 0.8696 4.2904 0.0000 0.5331 2.2758 0.3099 0.3469 0.5919 0.6444 0.0000 0.0000 0.0000 0.0000 1.1960 0.0000 0.2827 2.4069 0.2619 0.9268 0.0118 0.3364 0.0000 0.0000 2001 Case 1 Case 2 2.3279 0.0000 0.0506 0.0000 0.0000 0.0000 1.8211 0.0000 0.5901 1.0133 1.3685 0.0000 1.6086 0.0000 0.4745 1.7279 0.1170 0.4057 0.7945 0.0000 0.3123 0.7237 0.3270 0.9666 0.0263 1.2805 0.6133 0.9455 0.4017 1.1319 0.2055 0.0000 0.1505 0.7538 1.6235 0.0000 0.3014 2.0753 0.2531 0.6822 0.8391 0.0000 3.6530 0.0000 0.3636 2.2219 0.2767 0.4730 0.0045 0.6896 0.0000 0.0000 0.0000 0.0000 1.2235 0.0000 0.1933 2.1817 1.0137 0.0000 0.2882 0.0000 0.0000 0.0000 2002 Case 1 Case 2 2.1394 0.0000 0.2719 0.0000 0.0000 0.0000 1.8425 0.0000 0.5707 1.1142 1.2516 0.0000 1.6914 0.0000 0.6814 2.1009 0.2693 0.4162 0.7010 0.0000 0.2028 0.3720 0.0831 1.0714 0.0723 1.2264 0.7380 1.0190 0.6783 1.6990 0.0362 0.1541 0.3577 0.7183 1.5242 0.0000 0.4408 2.0687 0.4155 0.7778 0.5056 0.9481 4.1563 0.0000 0.5007 2.4574 0.3642 0.4246 0.2138 0.6718 0.0000 0.0000 0.0000 0.0000 1.2579 0.0000 0.2980 2.4846 0.0383 1.1632 0.1665 0.0000 0.0000 0.0000 2003 Case 1 Case 2 0.1459 2.5656 0.1607 0.0000 0.0000 0.0000 1.7062 0.0000 0.5366 1.1994 0.9675 0.0000 1.7879 0.0000 0.5329 1.9452 0.2139 0.4722 0.8564 0.0000 0.1445 0.3526 0.8962 0.0000 0.0370 1.4323 0.5632 1.0848 0.5520 1.7536 0.1417 0.0000 0.3649 0.7067 1.5904 0.0000 0.3063 2.0455 0.4017 1.0388 0.6074 0.9998 4.4356 0.0000 0.4274 2.6094 0.2737 0.3759 0.4974 0.6483 0.0000 0.0000 0.0000 0.0000 1.2281 0.0000 0.3422 2.9981 0.1094 1.1712 0.1177 0.0000 0.0000 0.0000
- 118 -
Assessment of Productive Efficiency of Airports
Table 7.1 Efficiency scores for Case 1 and Case 2 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Average score Number of efficient airports 2000 Case 1 Case 2 0.0301 0.0000 0.1833 0.3488 0.1553 2.6039 0.5276 0.0000 1.2123 0.0000 0.2600 0.5309 0.2944 2.5854 0.6936 2.2824 1.1367 0.0000 0.7166 0.6512 0.0000 0.0000 0.2264 1.2968 0.1525 2.7395 0.4099 4.0613 0.0000 0.0000 2.4880 0.0000 0.0821 0.0000 0.5650 0.7716 0.8666 0.0000 0.6842 1.6591 0.0000 0.0000 0.3992 0.0000 0.0147 2.5810 0.4392 1.5896 0.2208 1.1813 23 7 2001 Case 1 Case 2 0.0000 0.0000 0.0393 0.2947 2.4819 0.0000 0.5457 0.0000 0.2388 1.2000 0.1223 0.4958 2.4231 0.0000 0.2441 2.1645 1.1653 0.0000 0.5415 0.6293 0.1491 0.0000 0.1214 0.9713 2.4767 0.0000 0.7247 3.7163 0.0000 0.0000 2.5483 0.0000 0.0873 0.0000 0.4686 0.9830 0.8341 0.0000 0.2175 1.3735 0.0000 0.0000 0.2189 0.3213 0.0512 2.8170 0.0294 1.4895 0.1326 1.1296 29 7 2002 Case 1 Case 2 0.0229 0.0000 0.2002 0.2750 2.4938 0.0000 0.6508 0.0000 1.0546 0.0000 0.4646 0.0000 0.0331 2.2075 0.6013 2.5192 1.2548 0.0000 0.5628 0.0000 0.1590 0.0000 0.3194 1.0942 2.0497 0.0000 3.5666 0.0000 0.0000 0.0000 2.4674 0.0000 0.1816 0.0000 0.7539 1.1199 0.3169 1.1352 0.1123 1.1670 0.0000 0.0000 0.3945 0.4253 2.4968 0.0000 0.1627 1.6490 0.1672 1.1591 29 6 2003 Case 1 Case 2 0.1650 0.0000 0.1709 0.2884 0.1366 2.6035 0.7313 0.0000 0.1085 1.0359 0.4776 0.0000 0.1202 2.2866 0.6686 2.4350 1.3765 0.0000 0.6064 0.5875 0.1739 0.0000 0.3257 1.5232 2.0909 0.0000 3.0955 0.0000 0.0000 0.0000 2.4552 0.0000 0.2607 0.0000 0.7767 1.2892 1.3895 0.0000 0.1898 1.2490 0.0000 0.0000 0.5103 0.6358 2.9040 0.0000 0.3630 1.8012 0.1792 1.2168 28 6
Note: An efficient airport has a zero score as highlighted by bold typeface. The input set of both cases are the same. The
output set of Case 2 consist of passengers, aircraft movements, and cargo throughput. The output set of Case 1 include passengers, non-delayed flights, cargo throughput, delayed flights, and time delays.
- 119 -
Assessment of Productive Efficiency of Airports
On the contrary, when delayed flights and time delays are also considered (Case 1), the results show a greater number of efficient airports, including less-congested airports. In 2003, totally 28 airports are identified as efficient, as shown in Table 7.1. The additional 22 airports received credit due to their relatively low numbers of delayed flights and total time delays. The different classification is a result of overemphasis on increasing desirable outputs and not giving credit to airports with good performance on controlling undesirable outputs. The results indicate that there may be a balance between quantity and quality of outputs in the achievement of efficient outcomes; i.e., airports can trade-off utilization levels for reduced flight and time delays. For certain stakeholders, this option may be an optimal strategy. Passengers and shippers receive services with fewer flight delays. The FAA, as the regulator, has less concern over congestion and safety. Meanwhile, airport managers are able to balance traffic volume with customer satisfaction. By all accounts, the inclusion of undesirable outputs in the analysis appears to provide a fairer assessment of airport productivity.
7.1.2 The number of efficient airports
The results also show that the number of efficient airports increases as the number of measured outputs increases. The results are in line with Salazar de la Cruz (1999) who also found that as the number of variables (inputs and outputs) increases, the number of efficient airports is likely to be more. In 2000, without consideration of delayed flights and time delays (Case 2), only seven efficient airports are identified i.e., HartsfieldJackson Atlanta (ATL), Los Angeles International (LAX), LaGuardia (LGA), Phoenix Sky Harbor International (PHX), San Diego International (SAN) and John Wayne (SNA).
- 120 -
Assessment of Productive Efficiency of Airports
With the consideration of delayed flights and time delays as undesirable outputs (Case 1), there are 23 efficient airports or 16 more. The increase in the number of efficient airports as outputs are added is partly due to the way efficient units are calculated using nonparametric linear programming methods. The greater the number of outputs, the less likely an airport is to be dominated on all outputs; thus the more likely it is to be on the efficient frontier. As pointed out earlier in Chapter 3 the best way to avoid domination is to have a sufficiently large sample size so that an airport at least has some peers for comparison. This is of course a case in this dissertation.
7.1.3 Difference in efficiency scores
In order to show that efficiency scores are significantly different between Cases 1 and 2, several statistical tests are applied to the results. Tests are performed on both yearly basis and all years. Table 7.2 provides the results from paired-sample t-tests by treating efficiency scores as random variables. The results strongly support the assertion of differences. The efficiency scores in Case 2 are statistically higher than in Case 1. To avoid restricted assumptions of t-test, the non-parametric Wilcoxon signed-rank test and sign test are also performed. The results are shown in Table 7.3. They confirm that the difference in efficiency scores between cases with and without consideration of undesirable outputs is significant.
7.1.4 Ranking
The efficiency scores can be used to rank the performance of airports. All efficient airports (score = 0) are equally efficient since they all are on the efficient frontier. The inefficient airports are ranked in descending order by their efficiency scores. Table 7.4 shows ranking of airport productivity during 2000 – 2003. As shown in the - 121 -
Assessment of Productive Efficiency of Airports
Table rankings are drastically different when cases with and without consideration of undesirable outputs are compared. For example, by accounting for delays (Case 1) the operational efficiency of Boston Logan International (BOS) is ranked 25th, 54th, 46th, and 49th during 2000 – 2003 as compared to 29th, 31st, 29th, and 31st when ignoring delays (Case 2). This indicates that by not accounting for delayed flights and time delays, the performance ranking of airports can be distorted.
Table 7.2
Comparisons of efficiency scores between Cases 1 and 2 by paired sample t-test Paired differences Cases 1 and 2 95% confidence interval of the Std. Std. error difference deviation mean Lower Upper 1.0109 0.1351 0.6898 1.2312 0.9211 0.1231 0.7504 1.2437 0.9571 0.1279 0.7356 1.2482 0.9872 0.1319 0.7733 1.3020 0.9635 0.0644 0.8699 1.1236
Paired-sample t-test Pair 1: year 2000 Pair 2: year 2001 Pair 3: year 2002 Pair 4: year 2003 Pair 5: 2000 – 2003
Mean 0.9605 0.9971 0.9919 1.0377 0.9968
t 7.110 8.100 7.755 7.866 15.484
- 122 -
Assessment of Productive Efficiency of Airports
Table 7.3
Comparisons of efficiency scores between Cases 1 and 2 by nonparametric paired tests Asymptotic significance (2-tailed) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Nonparametric paired test A. Wilcoxon Signed-Rank test Pair 1: year 2000 Pair 2: year 2001 Pair 3: year 2002 Pair 4: year 2003 Pair 5: 2000 – 2003 Pair 1: year 2000 Pair 2: year 2001 Pair 3: year 2002 Pair 4: year 2003 Pair 5: 2000 – 2003
Z -6.053a -6.093a -6.154a -6.144a -12.189a -6.571 -6.857 -6.930 -6.647 -13.716
B. Sign test
a
Based on positive ranks.
- 123 -
Assessment of Productive Efficiency of Airports
Table 7.4 Ranking of airport productivity
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW
2000 2001 2002 2003 Case 1 Case 2 Case 1 Case 2 Case 1 Case 2 Case 1 Case 2 1 52 1 49 1 46 34 53 1 11 1 8 1 11 1 8 1 1 1 1 1 1 1 1 46 44 1 44 1 44 1 44 25 29 54 31 46 29 49 31 1 38 1 38 26 38 1 38 43 43 1 41 1 43 1 41 40 39 52 43 43 39 48 43 1 19 35 15 40 19 37 15 41 31 1 24 1 31 1 24 24 18 47 22 41 18 33 22 37 23 48 28 25 23 1 28 56 35 31 37 37 35 29 37 54 30 55 27 56 30 51 27 1 37 50 33 54 37 50 33 51 10 1 11 1 10 1 11 1 27 38 23 51 27 43 23 39 42 1 42 1 42 1 42 45 45 46 45 39 45 39 45 47 20 44 20 45 20 44 20 1 26 1 26 47 26 53 26 48 56 1 55 1 56 1 55 38 46 49 48 48 47 45 48 50 13 45 16 38 13 38 16 1 21 30 21 50 21 46 21 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 35 33 1 36 1 33 1 36 34 48 39 47 35 49 41 47 26 28 1 32 34 28 30 32 1 12 1 12 27 12 1 12
- 124 -
Assessment of Productive Efficiency of Airports
Table 7.4 Ranking of airport productivity (Continued)
Airport code MEM MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Sum Average 56 airports Average inefficiency # of efficient airports
2000 2001 2002 2003 Case 1 Case 2 Case 1 Case 2 Case 1 Case 2 Case 1 Case 2 1 1 1 1 1 1 1 1 30 8 1 1 1 8 1 1 29 14 33 13 31 14 35 13 1 53 1 52 30 53 32 51 1 16 1 18 1 16 1 18 33 34 42 35 1 34 29 35 36 17 37 17 33 17 1 17 53 51 1 50 36 52 31 49 1 47 43 46 53 48 54 46 55 32 1 34 1 32 1 34 1 22 53 19 55 22 52 19 32 1 1 10 1 1 1 10 28 36 36 29 32 36 40 29 42 54 1 51 29 54 1 50 1 55 56 56 42 55 1 56 1 1 1 1 1 1 1 1 1 49 1 53 1 50 1 52 49 9 1 9 1 9 1 9 1 24 51 30 49 24 55 30 52 25 1 25 1 25 1 25 1 41 40 39 52 41 36 39 1 1 1 1 1 1 1 1 27 15 41 14 1 15 47 14 44 50 34 54 28 51 1 54 31 40 32 40 44 40 42 40 1,343 1,575 1,190 1575 1320 1,575 1191 1575 0.2208 1.1813 0.1326 1.1296 0.2172 1.1731 0.1792 1.1339 0.3747 1.3501 0.2749 1.2910 0.3801 1.3406 0.3135 1.2958 23 7 29 7 24 7 28 7
Note: An efficient airport has a ranking = 1.
At the bottom of Table 7.4, the average efficiency scores across inefficient airports are shown. In 2003, the average scores are 0.3135 and 1.2958 for Cases 1 and 2 respectively. These average scores show that the performance of the inefficient airports is about four times poorer if delays are ignored. This result casts a doubt if these airports are really performing that poorly. While these figures may be a result of using incomplete
- 125 -
Assessment of Productive Efficiency of Airports
output measures, they also may be due to ignoring delays that cause unrealistic inefficiency level. This observation leads to another interesting insight next.
7.1.5 Maximum possible production outputs
Another interesting and insightful observation is that by ignoring delayed flights and time delays as outputs, the level of inefficiency may be overestimated. Recall that the terms (1 + ? ) y km plus the corresponding output slacks and (1 ? ? )bkj in (3.7) give the projection of desirable and undesirable outputs onto the frontier. For inefficient airports, these terms represent the maximum possible production outputs or highest potential outputs that an airport could have produced. For an efficient airport with ? = 0 , the terms are simply ( y km , bkj ) or the current level of outputs. It can be seen in Table 7.1 that efficiency scores in Case 2 are much higher than in Case 1. For example, in 2003, Albuquerque International Sunport (ABQ) has a score of 2.5656 in Case 2; implying that ABQ could accommodate at least 2 256.56% more passenger trips, aircraft movements and cargo throughput. Meanwhile, in Case 1, ABQ receives a relatively lower score of 0.1459, implying that ABQ would only need to increase all outputs by 14.59% in order to be on the efficient frontier. Overall, in 2003, the average score for these 56 airports suggest that the U.S. airport system should increase all outputs by 17.92% according to the calculations in Case 1 in order to achieve maximum possible production. On the contrary, the system would have had to produce as high as 121.68% more in Case 2. Tables 7.5, 7.6, 7.7 and 7.8 compare estimated maximum possible production of each airport during 2000 – 2003, for Cases 1 and 2. In each case, the percentage increase
2
The maximum possible production outputs may be higher depending on the value of output slacks.
- 126 -
Assessment of Productive Efficiency of Airports
from current levels of outputs is also computed. Let’s look at results of the most recent year 2003 (Table 7.8) for explanation. In Case 2, Albuquerque International Sunport (ABQ) had the potential to produce 54,411,318 passengers rather than 6,051,879 that was actually produced in 2003, a 799% increase in passengers. In practice, if ABQ were to produce this high output, it is likely that the number of delayed flights and time delays would be very high and unacceptable. However, after consideration of delayed flights and time delays (Case 1), the maximum possible output at ABQ is just 6,935,011 passengers, or a 14.59% percent increase over the current level. In general, ignoring undesirable outputs may yield unrealistic maximum possible production outputs. The unrealistic figures occur mainly because the model neglects the relationship between traffic volume, capacity and delay. In practice, delays play a major role in determining acceptable traffic volume and vice versa. The joint consideration of capacity and delays is therefore necessary (de la Cruz, 1999). In certain situations, the capacity of airside operation is limited by environmental considerations (Pels, Nijkamp and Rietveld, 2003). Based on the results of Case 1, it is suggested that the 56 airports grossly have the potential to increase passengers, aircraft movements (delayed plus non-delayed flights) and cargo throughput by 23.03%, 20.19%, and 34.54%, respectively. If the undesirable outputs are not considered (Case 2), the increases are 133.50%, 90.98% and 363.68%, respectively. The numbers are shown at the end of Table 7.8. The difference of the estimation between Cases 1 and 2 may be interpreted as amount of output loss due to cleaning up delayed flights and time delays or keeping them at relatively low levels.
- 127 -
Assessment of Productive Efficiency of Airports
Table 7.5 Maximum possible passengers, aircraft movements and cargo throughput in 2000
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM Case 2
31,337,627 11,630,465 80,162,407 23,268,112 54,053,755 11,470,266 54,885,529 33,076,352 39,402,365 27,704,471 46,087,079 27,978,866 86,731,133 119,828,724 81,705,167 39,629,328 29,719,867 65,387,989 28,205,634 48,445,074 65,906,186 25,459,430 26,717,442 44,252,913 60,622,286 67,303,182 25,374,868 18,625,264 44,451,860 59,390,735 20,945,697 11,769,213
Total passengers % add Case 1
398 131 0 658 95 142 180 149 71 303 104 76 124 97 130 16 87 184 210 141 87 1,772 405 35 64 0 0 2,820 260 93 34 0 7,572,239 5,030,557 80,162,407 3,067,777 41,712,092 4,800,070 19,602,609 24,396,829 31,173,199 6,873,998 41,581,182 16,050,706 50,206,299 106,262,017 70,620,630 34,188,468 25,282,740 23,016,542 12,170,935 42,098,078 53,509,212 1,360,280 8,106,485 43,038,655 58,688,340 67,303,182 25,374,868 637,853 15,841,801 38,897,365 15,857,218 11,769,213
% add
20 0 0 0 50 1 0 84 35 0 84 1 30 75 99 0 59 0 34 109 52 0 53 31 59 0 0 0 28 26 1 0
Case 2
841,169 355,766 915,454 469,295 951,353 388,327 886,740 826,156 727,592 492,609 740,464 524,560 1,144,754 1,651,938 1,276,964 490,392 548,365 930,378 788,203 743,506 904,086 480,811 487,422 645,887 857,227 783,433 383,325 833,152 743,773 691,833 398,415 388,412
Aircraft movements % add Case 1
260 23 0 205 95 142 180 149 61 107 55 76 124 97 130 9 87 169 210 63 87 429 228 87 64 0 0 120 241 93 34 0 280,925 288,919 915,454 153,917 726,532 162,504 316,703 474,220 607,944 238,011 665,523 300,889 739,636 1,438,268 931,597 450,229 459,262 345,771 339,653 657,193 714,538 90,883 227,940 648,506 815,812 783,433 383,325 379,399 279,970 540,853 301,469 388,412
% add
20 0 0 0 49 1 0 43 34 0 39 1 45 72 68 0 57 0 33 44 48 0 53 88 56 0 0 0 28 51 1 0
Case 2
310,571 2,221,661 868,286 198,815 925,906 90,332 660,899 297,305 351,435 251,273 605,615 76,112 1,055,298 1,780,819 782,621 1,178,964 443,491 1,187,052 95,890 625,271 688,946 224,726 239,602 2,449,730 524,591 2,038,784 71,149 108,514 513,023 639,250 56,849 2,489,078
Cargo (tons) % add Case 1
260 23 0 388 95 142 180 149 78 1,013 55 102 124 97 163 9 87 169 1,136 63 87 1,236 293 35 426 0 0 120 241 136 169 0 130,936 1,804,221 868,286 40,722 714,502 62,433 236,043 211,276 287,627 22,572 644,908 177,857 610,884 1,579,199 1,325,558 1,082,407 515,706 441,163 150,134 567,217 1,146,512 16,815 93,435 2,382,512 1,060,137 2,038,784 71,149 49,415 244,619 545,822 99,464 2,489,078
% add
52 0 0 0 50 67 0 77 46 0 65 371 30 75 345 0 118 0 1,836 48 211 0 53 31 963 0 0 0 62 101 371 0
- 128 -
Assessment of Productive Efficiency of Airports
Table 7.5 Maximum possible passengers, aircraft movements and cargo throughput in 2000 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Total Case 2
41,331,587 49,589,802 35,585,902 17,638,778 27,630,613 110,445,560 19,038,966 22,028,846 34,705,402 41,289,840 36,044,635 64,079,351 17,784,319 13,963,990 14,868,547 12,042,199 32,471,210 72,724,391 24,447,399 56,770,120 7,772,801 42,762,111 26,957,370 41,546,269 2,275,049,295
Total passengers % add Case 1
23 35 260 61 309 53 524 277 152 66 0 223 1,572 990 0 1,450 14 77 87 185 0 40 4,958 159 91 33,621,273 43,503,104 11,407,888 10,963,802 6,757,398 90,903,872 3,946,234 9,894,942 13,790,115 45,168,996 36,044,635 26,063,445 3,769,595 3,991,725 14,868,547 776,904 28,408,553 64,240,596 13,097,259 34,339,104 7,772,801 30,561,387 8,144,375 23,090,278 1,551,380,674
% add
0 18 16 0 0 26 29 69 0 81 0 32 254 212 0 0 0 56 0 73 0 0 1,428 44 30
Case 2
533,000 705,627 575,108 685,977 344,014 1,391,570 447,069 636,289 671,733 799,671 579,816 1,030,784 440,476 420,399 206,289 583,806 482,282 839,091 535,850 975,732 387,862 673,060 488,683 719,560 38,445,502
Aircraft movements % add Case 1
3 35 260 53 121 53 259 228 114 65 0 130 274 406 0 249 8 95 87 166 0 40 258 159 86 517,440 616,656 184,325 449,050 155,501 1,156,282 161,378 324,876 314,378 792,038 579,816 550,159 135,749 117,090 206,289 167,376 445,677 778,647 287,072 614,373 387,862 481,025 138,470 399,174 26,008,394
% add
0 18 16 0 0 27 29 68 0 64 0 23 15 41 0 0 0 81 0 67 0 0 1 44 26
Case 2
1,692,142 498,911 311,060 1,047,067 1,026,868 2,248,207 148,807 146,000 602,591 923,560 340,352 605,070 133,971 88,797 139,107 27,912 494,449 1,545,324 276,125 682,341 15,589 277,661 242,439 383,839 37,950,051
Cargo (tons) % add Case 1
3 35 260 53 121 53 414 596 114 65 0 312 2,196 67,684 0 840 8 77 87 166 0 113 647 273 59 1,642,744 1,045,794 164,659 685,425 464,164 2,437,829 57,061 434,258 282,019 960,161 340,352 180,292 77,276 80,433 139,107 2,970 456,920 1,707,269 147,929 432,166 15,589 130,152 156,004 421,808 34,173,774
% add
0 183 91 0 0 66 97 1,971 0 72 0 23 1,224 61,299 0 0 0 96 0 68 0 0 381 309 43
Note: Cases 1 and 2 are with and without consideration of undesirable outputs, respectively. Aircraft movements include both delayed
and non-delayed flights. % add is the percentage increase from current level of the corresponding output.
- 129 -
Assessment of Productive Efficiency of Airports
Table 7.6 Maximum possible passengers, aircraft movements and cargo throughput in 2001
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM Case 2
47,465,694 9,209,894 75,858,500 27,516,240 48,722,818 10,628,250 57,724,171 36,686,611 47,920,880 33,859,280 56,893,875 25,901,064 90,510,082 129,575,647 94,823,125 36,838,382 28,775,415 71,693,282 26,562,043 56,870,278 64,963,714 30,870,493 32,545,265 43,231,654 59,438,371 61,606,204 21,933,000 15,548,312 56,581,449 56,893,875 20,201,325 11,808,247
Total passengers % add Case 1
668 80 0 813 101 137 183 209 107 407 229 97 151 135 194 21 75 256 208 218 87 2,304 541 47 69 0 0 2,547 370 101 29 0 6,183,606 5,107,311 75,858,500 3,012,729 38,481,290 4,487,335 20,369,923 17,510,135 25,890,177 6,680,897 22,664,832 17,476,832 37,043,730 88,971,994 45,266,906 30,558,000 18,877,562 20,151,936 11,240,147 29,224,387 34,803,580 1,284,311 6,926,081 37,470,750 35,338,310 61,606,204 21,933,000 587,473 14,358,821 28,253,061 15,681,966 11,808,247
% add
0 0 0 0 59 0 0 47 12 0 31 33 3 61 40 0 15 0 30 64 0 0 36 28 0 0 0 0 19 0 0 0
Case 2
807,793 298,831 890,494 419,971 915,317 378,261 845,347 795,772 648,419 436,424 667,871 479,876 1,103,768 1,524,400 1,113,118 526,114 508,702 857,904 766,691 667,618 866,084 429,945 433,576 616,485 834,169 738,114 365,716 797,136 667,193 667,871 359,062 394,826
Aircraft movements % add Case 1
438 5 0 597 101 137 161 173 154 2,015 72 147 144 95 284 21 75 162 2,031 68 84 1,868 406 47 562 0 0 122 287 148 205 0 242,733 284,441 890,494 148,869 703,723 159,705 324,065 428,372 514,675 243,201 506,019 321,934 540,903 1,243,721 725,372 436,420 333,261 327,006 323,801 495,596 470,916 92,402 183,398 598,329 530,328 738,114 365,716 358,508 250,145 315,752 278,734 394,826
% add
0 0 0 0 55 0 0 47 12 0 31 32 12 59 39 0 15 0 30 25 0 0 36 105 7 0 0 0 19 0 0 0
Case 2
392,202 1,968,545 739,927 246,803 795,525 77,872 587,145 278,267 450,829 322,739 554,945 62,278 874,863 1,525,449 924,909 959,095 319,020 885,779 128,895 556,658 621,341 288,059 307,492 2,107,489 530,144 1,774,402 52,148 118,267 551,320 554,945 47,809 2,631,631
Cargo (tons) % add Case 1
438 5 0 597 101 137 161 173 154 2,015 72 147 144 95 284 21 75 162 2,031 68 84 1,868 406 47 562 0 0 122 287 148 205 0 72,876 1,873,750 739,927 35,433 628,306 32,878 225,083 150,411 292,366 15,260 493,936 132,195 420,832 1,378,600 1,442,812 795,584 209,287 337,631 74,212 414,667 337,842 14,634 82,896 1,826,652 517,649 1,774,402 52,148 53,190 236,252 223,545 15,684 2,631,631
% add
0 0 0 0 59 0 0 47 65 0 53 425 17 76 499 0 15 0 1,127 25 0 0 36 28 547 0 0 0 66 0 0 0
- 130 -
Assessment of Productive Efficiency of Airports
Table 7.6 Maximum possible passengers, aircraft movements and cargo throughput in 2001 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Total Case 2
31,668,450 50,445,737 33,313,785 18,105,026 25,557,671 100,892,176 21,572,361 24,791,783 44,798,994 42,669,451 45,702,538 75,858,500 19,880,293 16,548,642 15,184,332 11,496,714 37,026,987 68,676,642 24,006,781 72,645,066 7,324,557 43,567,735 32,864,676 48,378,795 2,402,635,134
Total passengers % add Case 1
0 50 248 55 281 50 628 317 253 74 29 280 1,781 1,309 0 1,485 37 98 83 286 0 63 8,052 204 116 31,668,450 35,058,695 9,567,651 11,713,225 8,302,839 75,694,280 2,963,223 7,388,932 12,703,676 37,847,971 35,439,051 22,365,971 1,057,150 2,025,856 15,184,332 725,140 27,036,073 50,861,183 13,088,997 22,911,625 7,324,557 36,378,058 1,918,917 16,355,131 1,280,691,015
% add
0 4 0 0 24 12 0 24 0 54 0 12 0 72 0 0 0 47 0 22 0 36 376 3 15
Case 2
471,008 649,297 506,842 611,557 340,376 1,364,091 408,409 598,280 630,363 760,845 635,807 890,494 405,037 396,255 206,988 569,460 435,600 813,708 468,616 883,528 378,903 626,503 434,268 649,412 35,958,516
Aircraft movements % add Case 1
0 40 248 55 120 50 522 786 117 63 48 432 3,107 129,894 0 857 9 98 83 224 0 219 1,463 471 67 471,008 520,651 145,564 395,653 191,525 1,063,376 119,309 234,992 291,117 699,371 553,310 506,271 116,501 144,738 206,988 160,486 400,635 730,017 255,499 452,824 378,903 574,174 119,598 268,481 22,772,468
% add
0 4 0 0 24 17 0 24 0 50 0 12 0 72 0 0 0 88 0 22 0 21 5 3 16
Case 2
1,639,760 476,155 263,581 917,575 921,894 1,944,048 178,978 182,456 526,102 873,728 418,347 739,927 159,600 119,595 134,689 28,190 435,452 1,261,208 263,955 702,641 14,849 389,536 311,198 456,142 35,596,398
Cargo (tons) % add Case 1
0 40 248 55 120 50 522 786 117 63 48 432 3,107 129,894 0 857 9 98 83 224 0 219 1,463 471 67 1,639,760 353,018 75,700 593,634 519,100 1,567,465 28,786 31,619 242,967 826,639 283,337 155,931 4,976 21,966 134,689 2,946 400,499 934,037 143,914 313,835 14,849 395,215 20,924 217,161 26,455,537
% add
0 4 0 0 24 21 0 53 0 54 0 12 0 23,776 0 0 0 47 0 45 0 223 5 172 24
Note: Cases 1 and 2 are with and without consideration of undesirable outputs, respectively. Aircraft movements include both delayed
and non-delayed flights. % add is the percentage increase from current level of the corresponding output.
- 131 -
Assessment of Productive Efficiency of Airports
Table 7.7 Maximum possible passengers, aircraft movements and cargo throughput in 2002
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM Case 2
47,082,564 15,526,628 76,876,128 27,861,486 47,983,915 10,403,788 56,209,470 36,782,412 48,704,535 34,377,363 57,657,096 26,662,417 91,791,939 132,833,476 112,639,482 33,702,666 29,275,757 65,692,610 24,937,323 56,988,954 76,876,128 31,395,381 33,066,340 42,657,320 58,997,163 56,223,843 21,986,679 17,041,398 57,345,382 57,657,096 20,262,756 11,141,594
Total passengers % add Case 1
665 216 0 891 111 125 196 252 106 410 177 107 157 151 247 15 72 233 210 234 127 2,471 568 42 69 0 0 1,073 458 116 17 0 6,151,129 4,914,539 76,876,128 2,810,791 35,647,876 4,620,683 19,012,529 17,578,887 29,952,090 6,741,354 37,620,386 13,941,695 38,229,992 91,816,824 54,507,100 30,258,481 23,131,236 19,749,905 11,577,573 31,569,526 51,048,711 1,221,138 7,432,943 40,849,751 42,492,287 56,223,843 21,986,679 1,453,412 13,342,966 27,674,633 17,371,036 11,141,594
% add
0 0 0 0 57 0 0 68 27 0 81 8 7 74 68 4 36 0 44 85 51 0 50 36 21 0 0 0 30 4 0 0
Case 2
800,139 352,657 889,966 416,577 828,929 365,229 820,678 780,673 645,088 434,829 667,474 446,776 1,101,697 1,544,739 1,324,897 468,057 482,401 817,160 755,604 662,482 889,966 427,372 431,550 571,199 830,639 645,424 362,439 791,612 666,695 667,475 354,961 398,769
Aircraft movements % add Case 1
214 27 0 184 111 125 169 210 42 70 37 107 123 102 170 15 72 152 207 78 95 416 246 99 67 0 0 126 248 131 17 0 254,874 277,267 889,966 146,555 610,900 162,211 304,921 420,585 576,384 255,630 582,466 252,786 558,511 1,302,154 815,124 485,889 379,747 323,726 354,136 524,229 743,794 82,883 187,219 584,615 602,418 645,424 362,439 350,603 248,280 403,408 304,304 398,769
% add
0 0 0 0 56 0 0 67 27 0 20 17 13 70 66 20 35 0 44 41 63 0 50 103 21 0 0 0 30 39 0 0
Case 2
439,749 2,253,301 734,083 251,104 820,220 89,502 676,506 314,417 446,975 320,056 550,562 45,457 867,812 1,353,342 1,081,237 981,039 283,600 1,047,424 147,858 577,566 734,083 285,553 304,887 2,264,634 575,288 1,779,855 32,223 120,470 546,955 550,562 30,689 3,390,800
Cargo (tons) % add Case 1
491 27 0 676 111 125 169 210 179 2,891 57 675 161 102 364 15 72 152 2,654 78 123 1,960 342 42 602 0 0 126 303 178 17 0 74,460 1,771,595 734,083 32,353 609,352 39,751 251,354 170,482 280,897 10,700 420,995 131,868 506,884 1,165,008 764,041 880,784 224,076 414,947 99,629 459,860 514,101 13,863 103,415 2,168,672 983,493 1,779,855 32,223 53,356 257,344 358,923 26,309 3,390,800
% add
0 0 0 0 57 0 0 68 75 0 20 2,150 52 74 228 4 36 0 1,756 42 56 0 50 36 1,100 0 0 0 89 81 0 0
- 132 -
Assessment of Productive Efficiency of Airports
Table 7.7 Maximum possible passengers, aircraft movements and cargo throughput in 2002 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Total Case 2
30,748,013 50,905,293 32,323,453 21,470,312 18,244,259 100,910,331 22,022,803 25,628,109 45,742,898 38,756,685 45,950,621 76,876,128 20,355,106 16,923,015 14,931,854 12,363,316 36,563,308 66,684,332 23,734,248 73,670,012 7,903,066 43,496,722 33,385,024 49,161,407 2,427,391,404
Total passengers % add Case 1
2 56 249 65 180 52 536 367 274 56 29 326 1,412 1,426 0 1,598 37 112 114 295 0 70 9,122 217 123 30,060,241 39,838,014 9,251,773 13,005,642 6,517,050 66,565,952 3,579,015 8,780,743 12,241,975 24,799,470 35,547,167 23,785,515 1,345,970 1,108,695 14,931,854 728,307 26,690,843 55,169,980 14,638,695 20,741,521 7,903,066 38,825,764 362,017 18,015,387 1,323,382,375
% add
0 22 0 0 0 0 3 60 0 0 0 32 0 0 0 0 0 75 32 11 0 52 0 16 22
Case 2
456,445 647,262 486,649 614,095 306,737 1,351,570 402,935 587,383 624,291 723,839 632,547 889,966 398,979 389,267 206,380 554,213 430,974 769,286 443,072 881,949 368,627 623,023 432,347 646,230 35,512,220
Aircraft movements % add Case 1
2 27 249 65 105 46 221 252 125 56 16 109 205 357 0 247 18 119 114 117 0 43 250 165 86 446,235 605,830 139,291 371,988 149,292 922,817 129,782 265,888 276,877 463,167 545,771 559,932 130,826 85,243 206,380 159,835 364,735 756,932 296,477 452,571 368,627 604,053 123,642 283,457 23,101,862
% add
0 19 0 0 0 0 3 59 0 0 0 32 0 0 0 0 0 115 43 11 0 38 0 16 21
Case 2
1,661,404 483,388 294,000 1,047,698 1,020,211 2,158,811 180,394 179,964 552,719 845,537 433,703 734,083 160,392 124,395 151,644 26,481 442,811 1,250,166 299,251 696,986 13,730 416,451 308,574 452,261 37,832,866
Cargo (tons) % add Case 1
2 51 249 65 105 46 449 906 125 56 45 424 3,450 168,002 0 835 18 112 114 222 0 223 2,228 393 70 1,624,242 383,425 84,150 634,643 496,547 1,473,980 52,005 54,643 245,134 541,039 298,945 184,943 4,518 74 151,644 2,832 374,753 1,034,300 184,570 297,407 13,730 352,447 13,257 268,010 27,496,714
% add
0 20 0 0 0 0 58 205 0 0 0 32 0 0 0 0 0 75 32 38 0 173 0 192 24
Note: Cases 1 and 2 are with and without consideration of undesirable outputs, respectively. Aircraft movements include both delayed
and non-delayed flights. % add is the percentage increase from current level of the corresponding output.
- 133 -
Assessment of Productive Efficiency of Airports
Table 7.8 Maximum possible passengers, aircraft movements and cargo throughput in 2003
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM Case 2
54,411,318 11,266,457 79,086,792 31,529,895 50,127,482 12,125,357 59,299,554 38,512,358 50,223,775 35,419,757 59,315,094 26,953,931 105,449,056 138,401,886 115,815,897 33,602,832 30,614,102 67,258,823 38,225,686 59,315,094 82,382,075 32,391,557 34,088,414 43,660,391 62,644,718 54,982,838 22,482,770 18,447,560 58,998,549 59,315,094 20,595,835 11,437,307
Total passengers % add Case 1
799.08 135.14 0.00 1,079.73 119.94 156.35 195.10 264.86 117.77 466.53 179.41 89.62 181.16 159.89 254.56 14.17 70.67 240.85 389.86 253.74 141.20 2,565.77 598.06 37.59 72.64 0.00 0.00 541.50 507.27 117.12 11.77 0.00 6,935,011 4,791,431 79,086,792 2,672,637 35,020,347 4,729,936 20,094,756 16,180,712 31,635,406 6,252,061 38,539,652 14,214,803 38,894,115 83,244,552 50,696,543 29,431,061 24,484,401 19,732,556 10,193,688 28,154,745 54,898,411 1,215,093 6,970,535 40,417,649 54,332,887 54,982,838 22,482,770 2,875,703 13,040,110 30,308,286 18,426,397 11,437,307
% add
14.59 0.00 0.00 0.00 53.66 0.00 0.00 53.29 37.17 0.00 81.55 0.00 3.70 56.32 55.20 0.00 36.49 0.00 30.63 67.91 60.74 0.00 42.74 27.37 49.74 0.00 0.00 0.00 34.22 10.94 0.00 0.00
Case 2
788,011 321,940 911,723 419,055 821,054 350,368 834,883 761,211 652,776 441,793 683,792 475,568 1,215,631 1,595,515 1,352,208 463,330 490,823 828,895 738,946 683,792 949,711 431,462 437,251 565,144 825,834 622,378 374,952 754,909 682,712 683,792 366,656 402,258
Aircraft movements % add Case 1
256.56 16.07 0.00 170.62 119.94 96.75 178.79 194.52 47.22 85.64 35.26 89.62 143.23 108.48 175.36 14.17 70.67 159.04 204.55 103.88 99.98 443.56 260.94 101.62 64.83 0.00 0.00 122.81 299.81 131.37 11.77 0.00 253,241 277,361 911,723 154,849 569,476 178,079 299,469 394,639 536,801 237,979 576,589 250,802 541,268 1,185,676 756,815 405,808 389,706 319,989 316,623 465,444 743,930 79,377 172,836 570,852 743,556 622,378 374,952 338,807 229,168 380,232 328,035 402,258
% add
14.59 0.00 0.00 0.00 52.55 0.00 0.00 52.69 21.07 0.00 14.05 0.00 8.30 54.93 54.11 0.00 35.51 0.00 30.49 38.77 56.65 0.00 42.67 103.66 48.41 0.00 0.00 0.00 34.21 28.66 0.00 0.00
Case 2
331,992 2,439,876 798,501 233,382 798,572 87,856 656,757 282,033 485,352 347,759 598,876 37,820 1,064,668 1,397,377 1,157,774 998,619 267,005 1,092,962 56,298 598,876 831,772 309,945 331,134 2,238,198 523,744 1,833,300 28,402 113,352 594,923 598,876 26,160 3,390,515
Cargo (tons) % add Case 1
363.68 16.07 0.00 582.72 119.94 96.75 178.79 194.52 246.47 3,130.16 52.50 555.01 227.24 109.32 425.67 14.17 70.67 159.04 874.86 109.87 117.78 2,728.74 368.70 37.59 537.52 0.00 0.00 122.81 335.24 210.24 12.44 0.00 133,954 2,102,025 798,501 34,184 943,554 44,654 235,576 272,944 334,987 10,766 449,437 5,774 884,820 1,043,533 710,246 874,641 226,674 421,930 155,183 399,990 683,510 10,957 104,553 2,071,962 510,720 1,833,300 28,402 50,873 265,216 860,080 23,266 3,390,515
% add
87.09 0.00 0.00 0.00 159.87 0.00 0.00 185.03 139.13 0.00 14.45 0.00 171.96 56.32 222.48 0.00 44.89 0.00 2,587.15 40.17 78.96 0.00 47.99 27.37 521.67 0.00 0.00 0.00 94.03 345.55 0.00 0.00
- 134 -
Assessment of Productive Efficiency of Airports
Table 7.8 Maximum possible passengers, aircraft movements and cargo throughput in 2003 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Total Case 2
34,479,734 54,502,729 33,425,071 23,456,948 17,639,338 103,836,235 24,266,049 31,753,180 47,653,087 39,166,142 49,730,463 79,086,792 22,273,756 20,361,442 15,260,791 13,319,297 38,400,771 67,104,325 25,514,563 75,830,984 8,535,130 47,057,418 34,412,038 50,687,729 2,556,136,266
Total passengers % add Case 1
16.50 64.16 260.35 73.13 169.39 49.39 606.15 428.27 284.43 58.75 32.93 454.33 1,535.66 1,533.04 0.00 1,669.39 43.52 128.92 138.95 310.64 0.00 130.37 8,644.45 226.52 133.50 29,595,618 41,741,976 10,542,505 13,548,363 7,258,006 69,508,672 3,849,419 10,029,678 12,395,938 39,631,312 37,412,165 18,913,688 1,361,758 1,246,842 15,260,791 752,762 26,755,888 52,079,506 10,677,903 21,972,152 8,535,130 35,870,254 393,530 21,158,079 1,346,865,126
% add
0.00 25.72 13.66 0.00 10.85 0.00 12.02 66.86 0.00 60.64 0.00 32.57 0.00 0.00 0.00 0.00 0.00 77.67 0.00 18.98 0.00 75.60 0.00 36.30 23.03
Case 2
486,310 657,595 494,806 593,629 298,085 1,372,264 398,907 589,762 634,652 708,879 635,974 911,723 393,148 381,163 203,285 526,873 447,256 767,885 473,312 900,616 350,074 621,218 438,355 654,359 35,868,503
Aircraft movements % add Case 1
16.50 28.84 260.35 73.13 103.59 47.76 228.66 243.50 137.65 58.75 17.39 152.32 209.09 309.55 0.00 245.52 26.07 129.55 138.95 124.90 0.00 63.58 290.40 180.12 90.98 417,423 595,117 156,040 342,871 162,248 928,691 135,951 284,004 267,052 700,772 541,771 478,550 127,197 93,068 203,285 152,485 354,770 600,665 198,082 476,174 350,074 568,680 112,284 317,570 22,573,542
% add
0.00 16.60 13.64 0.00 10.82 0.00 12.01 65.41 0.00 56.94 0.00 32.44 0.00 0.00 0.00 0.00 0.00 79.56 0.00 18.91 0.00 49.74 0.00 35.95 20.19
Case 2
1,907,475 529,522 291,276 1,034,279 1,056,052 2,232,328 174,884 254,710 568,616 832,637 415,223 798,501 157,215 95,380 135,547 24,526 443,031 1,312,916 259,549 757,846 12,050 378,835 335,175 491,146 39,051,395
Cargo (tons) % add Case 1
16.50 67.58 260.35 73.13 103.59 47.76 441.72 1,291.86 137.65 58.75 44.00 557.01 3,340.92 92,501.67 0.00 768.17 26.07 128.92 138.95 249.45 0.00 227.79 1,661.85 425.53 363.68 1,637,278 370,000 253,278 597,383 574,965 1,510,746 36,821 175,200 239,265 842,526 288,350 294,139 4,569 103 135,547 2,825 351,418 1,682,768 108,622 402,086 12,050 299,115 19,024 649,238 30,404,043
% add
0.00 17.09 213.34 0.00 10.85 0.00 14.06 857.38 0.00 60.64 0.00 142.02 0.00 0.00 0.00 0.00 0.00 193.41 0.00 85.40 0.00 158.81 0.00 594.69 34.54
Note: Cases 1 and 2 are with and without consideration of undesirable outputs, respectively. Aircraft movements include both delayed
and non-delayed flights. % add is the percentage increase from current level of the corresponding output.
- 135 -
Assessment of Productive Efficiency of Airports
7.2
Lumpiness of airport investment
Due to the fact that airport inputs are relatively fixed and air traffic tends to grow over a long period of time, one may expect a rather stable operational efficiency during the analysis. In most cases, this is true as long as there is no asset selling or drastic change in air traffic. Nevertheless, it is possible to see sharp decline in efficiency at some point in time due to the opening of a new facility. In the early years of an asset’s life it is likely that excess capacity will prevail and hence show up as a contributing factor to low annual productivity. In the later years of an asset’s life it might show up through an impact on high levels of congestion and hence a shortage of capacity which can reduce output and hence affect productivity in a different way (Hooper and Hensher, 1997). In his study, Parker (1999) explained that the sharp decline during 1991/92 at Stansted, London was associated with the opening of a new terminal in that year, leading to further excess capacity. The technical efficiency can be expected to rise over time and favor later airports over the earlier ones. Similarly, a newly delivered runway may therefore have a capacity that far exceeds realized demand. The lumpiness of runway investments can signify a productivity drop in the early years after the investment, as results indicate in the case of Detroit Metropolitan Wayne County (DTW) which opened its 6th runway on December 11, 2001. During 2000 – 2001, its productivity scores in Table 7.1 are rather stable (i.e., 1.2993 and 1.1319), but downgraded significantly in 2002 and 2003 (i.e., 1.6990 and 1.7536) after the new runway was completed. The same situation occurs at George Bush Intercontinental (IAH) which expanded and extended runway 15R/33L from 6038’ x 100’ to 10000’ x 150’. The scores are rather stable during
- 136 -
Assessment of Productive Efficiency of Airports
2000 – 2001, i.e., 0.8696 and 0.8391) and downgraded afterwards (i.e., 0.9481 and 0.9998 in 2002 and 2003 respectively). Therefore, once the sharp drop is detected, it should not be presumed that it is due to poor management.
7.3 Changes in productivity over time
In the past only Gillen and Lall (1998) studied the productivity growth of U.S. airports during 1989 – 1993 by computing the Malmquist index. As reviewed in Chapter 2, the study did not consider any undesirable outputs. It was explained in Chapter 3 that Malmquist index is not appropriate when there is an undesirable output. Here, the Luenberger (L) productivity index (equation 3.20) is computed with more comprehensive output measures during recent years. Table 7.9 and 7.10 shows the results for Cases 1 and 2 respectively. Table 7.9 shows computed changes in productivity for each airport and the overall average for 56 airports. Along with Luenberger index, its two components, i.e., efficiency change (LEFFCH) and technical change (LTECHCH) are also shown. Note that these indexes signal improvements with values greater than zero, and declines in productivity with values less than zero. The zero value indicates no productivity change between two years.
- 137 -
Assessment of Productive Efficiency of Airports
Table 7.9 Luenberger productivity indexes, Case 1
Airport code
ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM
L
0.079 0.000 0.310 -0.002 -0.079 -0.001 0.015 -0.127 0.040 0.000 -0.170 -0.220 -0.019 0.023 0.060 0.011 0.102 0.000 -0.073 -0.048 0.104 0.077 0.042 -0.116 0.235 0.000 0.475 -0.014 -0.025 -0.015 -0.002 0.000
2000 – 2001 LEFFCH LTECHCH
0.203 0.000 0.000 0.000 -0.086 0.011 0.000 -0.042 0.233 0.000 0.092 -0.317 0.269 0.135 0.292 0.000 0.444 0.000 0.035 0.225 0.518 0.000 0.170 0.033 0.587 0.000 0.000 0.000 0.089 0.262 0.012 0.000 -0.124 0.000 0.310 -0.002 0.007 -0.012 0.015 -0.085 -0.193 0.000 -0.262 0.096 -0.288 -0.112 -0.232 0.011 -0.342 0.000 -0.108 -0.273 -0.414 0.077 -0.127 -0.149 -0.353 0.000 0.475 -0.014 -0.114 -0.277 -0.014 0.000
L
0.009 0.000 0.000 0.000 -0.023 0.000 0.096 -0.068 -0.003 0.012 0.219 0.091 0.033 -0.001 -0.086 0.124 -0.045 0.000 0.001 -0.064 -0.102 0.026 -0.029 0.089 0.017 0.000 0.001 0.000 -0.062 -0.009 0.000 0.000
2001 - 2002 LEFFCH LTECHCH
0.000 0.000 0.000 0.000 0.019 0.000 0.000 -0.207 -0.152 0.000 0.110 0.244 -0.046 -0.125 -0.277 -0.036 -0.207 0.000 -0.139 -0.162 -0.506 0.000 -0.137 -0.088 -0.209 0.000 0.000 0.000 -0.105 -0.038 0.000 0.000 0.009 0.000 0.000 0.000 -0.042 0.000 0.096 0.139 0.150 0.012 0.109 -0.153 0.079 0.124 0.190 0.160 0.162 0.000 0.140 0.098 0.404 0.026 0.108 0.176 0.226 0.000 0.001 0.000 0.042 0.029 0.000 0.000
L
-0.090 0.000 0.000 0.029 -0.029 0.000 0.000 0.035 -0.021 -0.012 0.042 0.060 -0.008 0.055 -0.052 0.050 0.009 0.000 0.007 -0.055 -0.032 0.000 -0.029 0.043 -0.111 0.000 0.173 -0.001 -0.072 -0.021 0.000 0.000
2002 - 2003 LEFFCH LTECHCH
-0.146 0.000 0.000 0.000 0.034 0.000 0.000 0.148 0.055 0.000 0.058 0.083 0.035 0.175 0.126 0.036 -0.007 0.000 0.134 0.014 -0.102 0.000 0.073 0.091 -0.284 0.000 0.000 0.000 -0.044 -0.071 0.000 0.000 0.056 0.000 0.000 0.029 -0.063 0.000 0.000 -0.114 -0.076 -0.012 -0.016 -0.023 -0.043 -0.120 -0.178 0.014 0.016 0.000 -0.128 -0.069 0.070 0.000 -0.103 -0.048 0.173 0.000 0.173 -0.001 -0.028 0.050 0.000 0.000
- 138 -
Assessment of Productive Efficiency of Airports
Table 7.9 Luenberger productivity indexes, Case 1 (Continued)
Airport code
MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Average index Number of regress Number of no change Number of progress
L
0.000 -0.001 -0.012 -0.013 -0.077 0.264 0.162 -0.078 -0.015 0.204 0.265 -0.016 0.023 -0.056 0.000 -0.010 -0.026 -0.105 0.000 -0.061 0.000 -0.109 -0.070 0.066 0.018 28 9 19
2000 - 2001 LEFFCH LTECHCH
0.000 0.144 0.155 0.000 -0.239 0.138 0.294 0.450 0.000 0.175 0.000 0.105 0.152 -0.315 0.000 0.000 0.000 0.096 0.000 0.467 0.000 -0.219 -0.037 0.410 0.088 7 21 28 0.000 -0.145 -0.168 -0.013 0.162 0.126 -0.132 -0.527 -0.015 0.029 0.265 -0.121 -0.130 0.259 0.000 -0.010 -0.026 -0.202 0.000 -0.528 0.000 0.109 -0.034 -0.344 -0.070 34 9 13
L
0.000 -0.030 0.046 0.000 0.069 0.170 -0.017 -0.138 0.000 0.047 0.020 -0.066 0.001 0.314 0.000 -0.002 0.000 0.014 -0.158 0.096 0.000 -0.087 0.191 -0.015 0.012 19 15 22
2001 - 2002 LEFFCH LTECHCH
0.000 -0.161 0.000 0.000 0.239 0.122 -0.033 -0.357 0.000 0.541 0.000 -0.198 0.000 0.725 0.000 0.000 0.000 -0.285 -0.317 0.105 0.000 -0.176 0.051 -0.133 -0.035 23 24 9 0.000 0.131 0.046 0.000 -0.170 0.047 0.017 0.219 0.000 -0.495 0.020 0.132 0.001 -0.411 0.000 -0.002 0.000 0.299 0.158 -0.009 0.000 0.089 0.139 0.119 0.047 7 15 34
L
0.000 -0.002 -0.068 0.000 -0.054 0.134 -0.047 -0.009 -0.003 -0.116 0.008 -0.138 0.002 0.000 0.000 -0.018 0.000 -0.056 0.019 -0.056 0.000 -0.171 -0.005 -0.093 -0.013 27 15 14
2002 - 2003 LEFFCH LTECHCH
0.000 0.029 -0.137 0.000 -0.108 0.000 -0.087 -0.067 0.000 -0.606 0.000 -0.006 0.000 0.000 0.000 0.000 0.000 -0.023 0.317 -0.078 0.000 -0.116 0.000 -0.200 -0.012 16 25 15 0.000 -0.031 0.068 0.000 0.054 0.134 0.040 0.058 -0.003 0.491 0.008 -0.132 0.002 0.000 0.000 -0.018 0.000 -0.033 -0.298 0.022 0.000 -0.056 -0.005 0.107 -0.001 23 15 18
Note: Negative index indicates regressed productivity. Zero value means that there is no change in productivity between two
years. Positive index indicates productivity growth.
- 139 -
Assessment of Productive Efficiency of Airports
According to the results, the airport system on average had productivity gains in two periods during 2000 – 2002 and productivity loss in 2003. The amount of changes regardless of progress or regress is rather low. Between 2000 and 2001, the overall average rise in efficiency was 1.8 percent; and continued to increase by 1.2 percent in the next period before falling down slightly 1.3 percent during 2002 – 2003. The results are suggesting that productivity of U.S. airports during 2000 – 2003 is more or less the same. At the airport level, only five airports, i.e., Newark Liberty International (EWR), LaGuardia (LGA), O'Hare International (ORD), Phoenix Sky Harbor International (PHX), and Pensacola Regional (PNS) show progress in all periods. For all airports and all periods, it is found that there were productivity losses in 74 cases; productivity remains the same in 39 cases and there were productivity gains in 55 cases. About 32% of all cases show productivity gains. The productivity loss at many airports, especially between 2000 and 2001, may be associated with the September 11 terrorist attacks which shook aviation industry worldwide. Decomposition of Luenberger index into efficiency change (LEFFCH) and technical change (LTECHCH) can help explain source of productivity gain or loss. The equations for computing LEFFCH and LTECHCH are given in (3.21) and (3.22) respectively. An airport which has been efficient in time period t and t + 1 , will naturally show no change in relative efficiency. Only Covington/Cincinnati/Northern Kentucky International (CVG) achieved productivity gains in all time periods. For the sample as a whole, efficiency gains occur between 2000 and 2001; then drop afterwards in 2002 and 2003. Between 2000 and 2001 the overall average rise in efficiency was 8.8 percent. This
- 140 -
Assessment of Productive Efficiency of Airports
was followed by a 3.5 % and 1.2 % drops in the two subsequent periods. For all airports and all periods, it is found that there were productivity losses in 46 cases; productivity remains the same in 70 cases and there were productivity gains in 52 cases. About 33% of all cases show productivity gains. The computed LTECHCH is shown next to LEFFCH column in Table 7.9. LTECHCH measures the average shifts in the efficient frontier from time period t to time period t + 1 . This corresponds to the term in equation (3.22). The results show productivity gain in one period, i.e., 2001 – 2002; and two periods, i.e., 2000 - 2001 and 2002 - 2003 with productivity loss. Between 2000 and 2001, high negative value of LTECHCH (-0.070) indicates that the efficient frontier shifted backward. In other words, for a given level of inputs in 2000, the airport system produces lower outputs in 2001 than in 2000. For all airports and all periods, it is found that there were productivity losses in 64 cases; productivity remains the same in 39 cases and there were productivity gains in 65 cases. About 33% of all cases show productivity gains. About 39% of all cases show productivity gains. In conclusion, the productivity gains between 2000 and 2001 are mainly from efficiency change (LEFFCH = 0.088) which compensates the productivity loss from frontier shifted backward (LTECHCH = -0.070). The situation is opposite in 2001/2002 period where productivity gains resulted from frontier shift (LTECHCH = 0.047). Between 2002 and 2003 both efficiency loss and frontier backward shift collectively contribute to the overall productivity loss (L = -0.013).
- 141 -
Assessment of Productive Efficiency of Airports
Productivity indexes for Case 2 are also computed in order to analyze the impact of considering undesirable outputs in the assessment. The results are shown in Table 7.10. For convenient comparisons, productivity indexes in Cases 1 and 2 are presented side by side in Table 7.11. For individual airports, the results do not show any recognizable pattern. Moreover, the overall picture of the airport system indicates a different conclusion. Instead of showing productivity gains between 2000 and 2001 (L = 0.018), Case 2 shows the opposite result, i.e., productivity loss (L = -0.041). The classification of productivity changes is also very different in all periods. It clearly shows the two sets of results are drastically different. Again, this confirms previous findings that ignoring undesirable outputs in the assessment really creates problematic results. Several statistical tests are performed on the Luenberger indexes to check if the difference between Cases 1 and 2 are statistical different. Table 7.12 provides the results from paired-sample t-tests which strongly support the assertion of differences. To avoid the restricted assumptions of the t-test, the non-parametric Wilcoxon signed-rank test and sign test are also performed. The results are shown in Table 7.13. They confirm that the difference in Luenberger productivity indexes between Cases 1 and 2 is significant. There is one exception in period 2002/03 where Z-statistics from both Wilcoxon signed rank test (Z = -2.213) and sign test (Z = 0.099) are not significant at 95%.
- 142 -
Assessment of Productive Efficiency of Airports
Table 7.10 Luenberger productivity indexes, Case 2
Airport code
ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM
L
0.125 0.014 0.000 -0.098 -0.201 -0.041 0.040 -0.333 0.031 0.042 -0.325 -0.319 -0.165 -0.137 -0.135 -0.119 -0.032 -0.203 -0.068 -0.224 -0.053 0.082 -0.327 -0.177 -0.078 0.000 0.000 -0.095 -0.132 -0.139 -0.055 0.000
2000 – 2001 LEFFCH LTECHCH
0.275 0.181 0.000 0.228 -0.064 0.047 0.191 -0.239 0.204 0.275 -0.174 -0.206 -0.042 0.026 0.167 -0.116 0.120 0.067 0.023 -0.053 0.030 0.637 0.054 -0.126 -0.045 0.000 0.000 -0.028 0.225 -0.087 0.048 0.000 -0.150 -0.167 0.000 -0.326 -0.138 -0.088 -0.151 -0.094 -0.173 -0.234 -0.151 -0.113 -0.122 -0.163 -0.302 -0.003 -0.152 -0.270 -0.090 -0.170 -0.084 -0.555 -0.381 -0.051 -0.033 0.000 0.000 -0.067 -0.358 -0.052 -0.104 0.000
L
0.158 -0.029 0.000 -0.044 -0.159 0.061 -0.112 -0.416 -0.018 0.087 0.350 -0.106 0.050 -0.073 -0.569 -0.017 -0.005 0.039 -0.047 -0.098 -0.114 -0.533 -0.251 0.070 0.010 0.000 0.000 -0.043 -0.305 -0.092 0.096 0.078
2001 – 2002 LEFFCH LTECHCH
0.189 -0.221 0.000 -0.021 -0.101 0.117 -0.083 -0.373 -0.010 0.093 0.352 -0.105 0.054 -0.073 -0.567 0.051 0.035 0.099 0.007 -0.096 -0.109 -0.503 -0.235 0.048 0.018 0.000 0.000 -0.034 -0.303 -0.149 0.122 0.000 -0.031 0.192 0.000 -0.023 -0.059 -0.056 -0.029 -0.043 -0.007 -0.006 -0.002 -0.002 -0.004 0.001 -0.002 -0.068 -0.041 -0.060 -0.054 -0.003 -0.005 -0.029 -0.016 0.022 -0.008 0.000 0.000 -0.009 -0.002 0.058 -0.026 0.078
L
-0.477 0.073 0.012 0.153 -0.046 0.189 -0.061 0.066 -0.039 -0.127 0.052 0.191 -0.154 0.000 0.001 0.009 0.038 -0.018 -0.045 -0.194 -0.004 -0.229 -0.106 0.034 0.014 0.000 0.008 -0.077 -0.425 0.041 0.062 0.000
2002 - 2003 LEFFCH LTECHCH
-0.426 0.111 0.000 0.136 -0.085 0.284 -0.096 0.156 -0.056 -0.155 0.019 0.175 -0.206 -0.066 -0.055 0.012 0.012 -0.066 0.023 -0.261 -0.052 -0.279 -0.152 0.049 0.024 0.000 0.000 0.030 -0.514 -0.008 0.049 0.000 -0.051 -0.038 0.012 0.016 0.039 -0.095 0.036 -0.090 0.017 0.028 0.033 0.015 0.052 0.066 0.056 -0.003 0.027 0.048 -0.068 0.066 0.048 0.050 0.046 -0.015 -0.010 0.000 0.008 -0.107 0.089 0.049 0.013 0.000
- 143 -
Assessment of Productive Efficiency of Airports
Table 7.10 Luenberger productivity indexes, Case 2 (Continued)
Airport code
MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Average index Number of regress Number of no change Number of progress
L
-0.023 -0.051 -0.195 -0.146 -0.132 -0.066 -0.155 -0.081 -0.178 -0.058 -0.201 0.014 -0.040 0.056 0.000 -0.148 -0.093 -0.264 -0.124 0.017 -0.007 -0.021 -0.675 -0.141 -0.104 42 4 10
2000 – 2001 LEFFCH LTECHCH
0.030 0.054 0.122 -0.018 0.012 0.035 0.162 0.118 -0.029 0.022 -0.149 0.326 0.263 0.345 0.000 -0.060 -0.005 -0.211 0.032 0.286 0.000 0.078 -0.236 0.100 0.052 18 6 32 -0.054 -0.105 -0.317 -0.128 -0.144 -0.101 -0.317 -0.199 -0.149 -0.080 -0.052 -0.312 -0.303 -0.289 0.000 -0.088 -0.088 -0.053 -0.156 -0.268 -0.007 -0.099 -0.439 -0.242 -0.156 51 4 1
L
-0.011 0.016 -0.063 0.011 0.088 0.038 0.171 -0.416 -0.093 -0.009 -0.016 -0.124 0.378 0.067 0.000 -0.014 -0.097 -0.139 -0.263 0.202 -0.007 -0.112 0.304 -0.172 -0.041 33 4 19
2001 - 2002 LEFFCH LTECHCH
-0.023 0.020 -0.012 -0.105 0.145 0.031 0.216 -0.355 -0.089 0.066 -0.010 -0.123 0.427 0.150 0.000 0.081 -0.094 -0.137 -0.301 0.207 0.000 -0.104 0.320 -0.160 -0.029 28 6 22 0.011 -0.004 -0.051 0.116 -0.057 0.007 -0.045 -0.061 -0.003 -0.075 -0.006 -0.001 -0.049 -0.083 0.000 -0.095 -0.003 -0.002 0.038 -0.004 -0.007 -0.008 -0.016 -0.013 -0.012 43 4 9
L
-0.132 0.002 -0.024 -0.064 0.001 0.013 -0.112 0.012 -0.078 -0.050 -0.009 -0.373 -0.086 0.380 0.000 -0.163 -0.036 -0.098 -0.184 -0.036 -0.012 -0.215 -0.356 -0.118 -0.050 32 3 21
2002 - 2003 LEFFCH LTECHCH
-0.142 -0.013 -0.110 -0.081 0.019 -0.013 -0.079 0.084 -0.122 -0.025 -0.015 -0.429 -0.041 0.471 0.000 0.012 -0.079 -0.169 -0.254 -0.082 0.000 -0.210 -0.407 -0.152 -0.058 33 6 17 0.010 0.016 0.086 0.017 -0.018 0.026 -0.033 -0.072 0.044 -0.026 0.006 0.056 -0.045 -0.091 0.000 -0.175 0.043 0.072 0.071 0.046 -0.012 -0.004 0.051 0.034 0.008 18 3 35
Note: The negative index indicates regressed productivity. Zero value means that there is no change in the productivity
between two years. The positive index indicates productivity growth.
- 144 -
Assessment of Productive Efficiency of Airports
Table 7.11 Comparisons of Luenberger productivity indexes between Cases 1 and 2
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM
2000 - 2001 Case 1 Case 2 0.079 0.125 0.000 0.014 0.310 0.000 -0.002 -0.098 -0.079 -0.201 -0.001 -0.041 0.015 0.040 -0.127 -0.333 0.040 0.031 0.000 0.042 -0.170 -0.325 -0.220 -0.319 -0.019 -0.165 0.023 -0.137 0.060 -0.135 0.011 -0.119 0.102 -0.032 0.000 -0.203 -0.073 -0.068 -0.048 -0.224 0.104 -0.053 0.077 0.082 0.042 -0.327 -0.116 -0.177 0.235 -0.078 0.000 0.000 0.475 0.000 -0.014 -0.095 -0.025 -0.132 -0.015 -0.139 -0.002 -0.055 0.000 0.000
2001 - 2002 Case 1 Case 2 0.009 0.158 0.000 -0.029 0.000 0.000 0.000 -0.044 -0.023 -0.159 0.000 0.061 0.096 -0.112 -0.068 -0.416 -0.003 -0.018 0.012 0.087 0.219 0.350 0.091 -0.106 0.033 0.050 -0.001 -0.073 -0.086 -0.569 0.124 -0.017 -0.045 -0.005 0.000 0.039 0.001 -0.047 -0.064 -0.098 -0.102 -0.114 0.026 -0.533 -0.029 -0.251 0.089 0.070 0.017 0.010 0.000 0.000 0.001 0.000 0.000 -0.043 -0.062 -0.305 -0.009 -0.092 0.000 0.096 0.000 0.078
2002 - 2003 Case 1 Case 2 -0.090 -0.477 0.000 0.073 0.000 0.012 0.029 0.153 -0.029 -0.046 0.000 0.189 0.000 -0.061 0.035 0.066 -0.021 -0.039 -0.012 -0.127 0.042 0.052 0.060 0.191 -0.008 -0.154 0.055 0.000 -0.052 0.001 0.050 0.009 0.009 0.038 0.000 -0.018 0.007 -0.045 -0.055 -0.194 -0.032 -0.004 0.000 -0.229 -0.029 -0.106 0.043 0.034 -0.111 0.014 0.000 0.000 0.173 0.008 -0.001 -0.077 -0.072 -0.425 -0.021 0.041 0.000 0.062 0.000 0.000
- 145 -
Assessment of Productive Efficiency of Airports
Table 7.11 Comparisons of Luenberger productivity indexes (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Average index Number of regress Number of no change Number of progress
2000 - 2001 Case 1 Case 2 0.000 -0.023 -0.001 -0.051 -0.012 -0.195 -0.013 -0.146 -0.077 -0.132 0.264 -0.066 0.162 -0.155 -0.078 -0.081 -0.015 -0.178 0.204 -0.058 0.265 -0.201 -0.016 0.014 0.023 -0.040 -0.056 0.056 0.000 0.000 -0.010 -0.148 -0.026 -0.093 -0.105 -0.264 0.000 -0.124 -0.061 0.017 0.000 -0.007 -0.109 -0.021 -0.070 -0.675 0.066 -0.141 0.018 -0.104 28 42 9 4 19 10
2001 - 2002 Case 1 Case 2 0.000 -0.011 -0.030 0.016 0.046 -0.063 0.000 0.011 0.069 0.088 0.170 0.038 -0.017 0.171 -0.138 -0.416 0.000 -0.093 0.047 -0.009 0.020 -0.016 -0.066 -0.124 0.001 0.378 0.314 0.067 0.000 0.000 -0.002 -0.014 0.000 -0.097 0.014 -0.139 -0.158 -0.263 0.096 0.202 0.000 -0.007 -0.087 -0.112 0.191 0.304 -0.015 -0.172 0.012 -0.041 19 33 15 4 22 19
2002 - 2003 Case 1 Case 2 0.000 -0.132 -0.002 0.002 -0.068 -0.024 0.000 -0.064 -0.054 0.001 0.134 0.013 -0.047 -0.112 -0.009 0.012 -0.003 -0.078 -0.116 -0.050 0.008 -0.009 -0.138 -0.373 0.002 -0.086 0.000 0.380 0.000 0.000 -0.018 -0.163 0.000 -0.036 -0.056 -0.098 0.019 -0.184 -0.056 -0.036 0.000 -0.012 -0.171 -0.215 -0.005 -0.356 -0.093 -0.118 -0.013 -0.050 27 32 15 3 14 21
Note: The negative index indicates regressed productivity. Zero value means that there is
no change in the productivity between two years. The positive index indicates productivity growth.
- 146 -
Assessment of Productive Efficiency of Airports
Table 7.12 Comparisons of Luenberger productivity indexes by paired sample t-test
Paired-sample t-test Pair 1: 2000/01 Pair 2: 2001/02 Pair 3: 2002/03 Pair 4: 2000 – 03 Mean 0.1219 0.0531 0.0373 0.0708
Paired differences Cases 1 and 2 95% confidence interval of the Std. Std. error difference deviation mean Lower Upper 0.1458 0.0194 0.0829 0.1610 0.1527 0.0204 0.0122 0.0940 0.1287 0.0172 0.0029 0.0718 0.1466 0.0113 0.0484 0.0931
t 6.2605 2.6014 2.1741 6.2607
Table 7.13 Comparisons of Luenberger productivity indexes by nonparametric tests
Nonparametric paired test A. Wilcoxon Signed-Rank test Pair 1: years 2000/01 Pair 2: years 2001/02 Pair 3: years 2002/03 Pair 4: years 2000 - 03 Pair 1: years 2000/01 Pair 2: years 2001/02 Pair 3: years 2002/03 Pair 4: years 2000 - 03
Z -5.307a -2.598a -2.213a -6.027a -4.395 -2.747 -1.648 -5.234
B. Sign test
Asymptotic significance (2-tailed) 0.000 0.009 0.026 0.000 0.000 0.006 0.099 0.000
a
Based on positive ranks.
- 147 -
Assessment of Productive Efficiency of Airports
7.4
Scenario analysis
Having shown that ignoring undesirable outputs while assessing productivity of airports can cause drastically different interpretations (sometimes unrealistic), it is interesting to analyze the impact of other measures on productivity. The results will provide more insights on the sensitivity of estimated productivity. Many different sets of possible input and output measures can be analyzed. Pathomsiri, Haghani, Windle and Dresner (2006) have analyzed the impact of cargo throughput and a single undesirable output – delayed flights, on the productivity of 56 U.S. airports (the same dataset that is used in this dissertation). Their work notes that the addition of the undesirable output (delayed flights) into the model made much more difference than the addition of another desirable output (cargo throughput); i.e., higher number of efficient airports are identified. The results suggest that consideration of undesirable output is at least as important as the consideration of additional desirable outputs in determining relative productivity of airports. Here let’s consider the effects of an input measure, i.e., land area. Recall that the preceding sections consider land area as an operational input along with number of runways and runway area. The resulting estimated productivity indicates the operational efficiency of airside operation from utilizing these three inputs. One might argue that land area is not being used solely for airside operation which is probably true. Nowadays many airports are more enthusiastic to provide non-aeronautical services including concessions, rentals and car parking (Francis and Humphreys, 2001; Francis, Humphreys and Fry (2002); Hooper and Hensher, 1997; Humphreys, 1999; Humphreys and Francis, 2002; Nyshadham and Rao, 2000; Oum and Yu, 2004; Oum, Yu and Fu,
- 148 -
Assessment of Productive Efficiency of Airports
2003). In some airports such as Honolulu, Vancouver and Sydney, non-aeronautical revenues account for as much as 70% of their total revenues (Oum and Yu, 2004). Airport management has become more diversified (Francis and Humphreys, 2001). Land area is being utilized beyond just aeronautical services. Furthermore, airport may own more excessive land for other reasons such as noise abatement, planned future expansion, land appreciation, and other investment. In such cases, inclusion of land area in the set of input may give biased results in favor of airports with less land area and limited nonaeronautical activities. To analyze the impact of land area, Case 3 is then set up as follow: Case 3: without land area as an input Input = {number of runway, runway area} Desirable outputs = {non-delayed flights, passengers, cargo} Undesirable outputs = {delayed flights, time delays} The directional output distance function in (3.7) is then solved again 56 times, each for an individual airport to estimate the efficiency score, ? . Table 7.14 shows the results. The estimated efficiency scores in Case 1 are copied from Table 7.1 for comparison purposes. There are several interesting observations for discussion. First, on the efficiency scores, average scores from both cases (at the bottom of the Table) are not much different. The difference is relatively much smaller than the comparison between Cases 1 (with undesirable outputs) and 2 (without desirable outputs). This is because at the individual airports, scores are not different. Several
- 149 -
Assessment of Productive Efficiency of Airports
airports even earned identical scores regardless of cases. In 2003, 45 airports have identical pairs of scores. Only 11 airports (i.e., ABQ, BOS, DCA, FLL, HOU, LGB, MDW, MSY, ONT, and ORF) show different scores. For all of them, the productivity is downgraded. These 11 airports are relatively smaller in land size than those 45. The range of area is between 650 (at MDW) and 2039 (at ABQ) acres. Inclusion of land area as an input clearly favors airports with relatively smaller size. Second, on the classification of efficient airports, the results show that all efficient airports in Case 3 are also efficient in Case 1. In other words, the set of efficient airports in Case 3 is a subset of efficient airports in Case 1. Excluding land area never decreases inefficiency level in Case 1. Inefficient airports in Case 1 are still identified as (more) inefficient in Case 3. Note that one should not expect this finding to be always true in other applications. It is possible that excluding (or including) an input measure from the consideration may decrease (or increase) inefficiency levels if that input measure is a dominant one. Third, on the number of efficient airports, the annual figures are not much different. Case 1 identifies 23, 29, 29 and 28 efficient airports whereas Case 3 identifies 22, 24, 24 and 25 airports in 2000, 2001, 2002 and 2003 respectively. Both cases identify almost the same set of efficient airports. Case 1 with more total measures (number of inputs plus outputs) identifies more efficient airports. This finding is in line with previous observations (Parker, 1999; Salazar de la Cruz, 1999; Pathomsiri, Haghani, Dresner and Windle, 2006b, 2006c; Pathomsiri, Haghani, Windle and Dresner, 2006).
- 150 -
Assessment of Productive Efficiency of Airports
Table 7.14 Efficiency scores for Cases 1 and 3
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM 2000 Case 1 Case 3 0.2034 0.5771 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5044 0.5048 0.0108 0.1478 0.0000 0.0000 0.4323 0.4763 0.3502 0.3502 0.0000 0.0000 0.4044 0.4044 0.0102 0.0570 0.2956 0.2956 0.7486 0.7486 0.6938 0.6938 0.0000 0.0000 0.5941 0.5941 0.0000 0.0000 0.3367 0.4391 0.4777 0.4777 0.5179 0.5179 0.0000 0.0000 0.5331 0.5331 0.3099 0.3099 0.5919 0.6252 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.2827 0.2827 0.2619 0.2619 0.0118 0.6070 0.0000 0.0000 2001 Case 1 Case 3 0.0193 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5901 0.5901 0.0500 0.0000 0.0000 0.0000 0.4745 0.5047 0.1170 0.1170 0.0000 0.0000 0.3123 0.3123 0.3270 0.3445 0.0263 0.0263 0.6133 0.6133 0.4017 0.4017 0.0000 0.0000 0.1505 0.1505 0.0000 0.0000 0.3014 0.4138 0.2531 0.2531 0.0000 0.0000 0.0000 0.0000 0.3636 0.3636 0.2767 0.2767 0.0045 0.0429 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.1933 0.1933 0.0000 0.0000 0.1707 0.0000 0.0000 0.0000 2002 Case 1 Case 3 0.1729 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5707 0.5713 0.0000 0.0000 0.0000 0.0000 0.6814 0.6958 0.2693 0.2693 0.0000 0.0000 0.2028 0.2028 0.0831 0.1696 0.0723 0.0723 0.7380 0.7380 0.6783 0.6783 0.0362 0.1011 0.3577 0.3577 0.0000 0.0000 0.4408 0.4880 0.4155 0.4155 0.5056 0.5056 0.0000 0.0000 0.5007 0.5007 0.3642 0.3642 0.2138 0.2595 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.2980 0.2980 0.0383 0.0383 0.4767 0.0000 0.0000 0.0000 2003 Case 1 Case 3 0.1459 0.4162 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5366 0.5407 0.0000 0.0000 0.0000 0.0000 0.5329 0.6372 0.2139 0.2139 0.0000 0.0000 0.1445 0.1445 0.0428 0.0000 0.0370 0.0370 0.5632 0.5632 0.5520 0.5520 0.0000 0.0000 0.3649 0.3654 0.0000 0.0000 0.3063 0.4503 0.4017 0.4017 0.6074 0.6074 0.0000 0.0000 0.4274 0.4274 0.2737 0.2737 0.4974 0.4974 0.0000 0.0000 0.0000 0.0000 0.0306 0.0000 0.3422 0.3422 0.1094 0.1094 0.5131 0.0000 0.0000 0.0000
- 151 -
Assessment of Productive Efficiency of Airports
Table 7.14 Efficiency scores for Cases 1 and 3 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Average score Number of efficient airports 2000 Case 1 Case 3 0.0000 0.0000 0.1833 0.1833 0.1553 0.2757 0.0000 0.0000 0.1934 0.0000 0.2600 0.2600 0.2944 0.2944 0.6936 0.6936 0.0000 0.0000 0.7166 0.7526 0.0000 0.0000 0.2264 0.2264 0.1525 0.1525 0.4099 0.4372 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5650 0.5650 0.0000 0.0000 0.6842 0.6842 0.0000 0.0000 0.0000 0.0000 0.0147 0.0147 0.4392 0.4392 0.2208 0.2514 23 22 2001 Case 1 Case 3 0.0000 0.0000 0.0393 0.0438 0.1037 0.0000 0.0000 0.0000 0.2388 0.3104 0.1223 0.1223 0.0000 0.0000 0.2441 0.2441 0.0000 0.0000 0.5415 0.5620 0.0000 0.0000 0.1214 0.1214 0.0000 0.0000 0.7247 0.7255 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.4686 0.4686 0.0729 0.0000 0.2175 0.2175 0.0000 0.0000 0.2189 0.2189 0.0512 0.0512 0.0294 0.0294 0.1326 0.1453 29 24 2002 Case 1 Case 3 0.0000 0.0000 0.2002 0.2002 0.0468 0.0000 0.0000 0.0000 0.1477 0.0000 0.0000 0.0000 0.0331 0.0408 0.6013 0.6013 0.0000 0.0000 0.6188 0.0000 0.0000 0.0000 0.3194 0.3194 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.7539 0.7539 0.3169 0.4290 0.1123 0.1123 0.0000 0.0000 0.3945 0.3945 0.0000 0.0000 0.1627 0.1627 0.1672 0.2001 29 24 2003 Case 1 Case 3 0.0000 0.0000 0.1709 0.1709 0.1366 0.2220 0.0000 0.0000 0.1085 0.2135 0.0000 0.0000 0.1202 0.1316 0.6686 0.6686 0.0000 0.0000 0.6064 0.6064 0.0000 0.0000 0.3257 0.3257 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.7767 0.7767 0.0000 0.0000 0.1898 0.1898 0.0000 0.0000 0.5103 0.5103 0.0000 0.0000 0.3630 0.3630 0.1792 0.2026 28 25
Note: An efficient airport has a zero score as highlighted by bold typeface. The output sets of Cases 1 and 3 are the same. The input
set of Case 3 is {number of runways, runway area}. The input set of Case 1 is {land area, number of runways, runway area}
- 152 -
Assessment of Productive Efficiency of Airports
Both parametric and nonparametric statistical tests are performed to determine whether the score differences between Cases 1 and 3 are significant. Table 7.15 shows results from paired-sample t-test whereas results from nonparametric Wilcoxon signed rank test and sign test are shown in Table 7.16. In brief, paired-sample t-tests infer that efficiency scores in Cases 1 and 3 are still significantly different. However, the level of significance is not as strong as in the comparison between Cases 1 and 2. The t-statistics are as low as -2.1708 (Pair 4: year 2003). Results from nonparametric tests (Table 7.16) also provide the same statistical inference. Both Wilcoxon signed-rank test and sign test indicate the difference in scores in Cases 1 and 3 are statistically significant at 99% level.
Table 7.15
Comparisons of efficiency scores between Cases 1 and 3 by paired sample t-test Paired differences Cases 1 and 3 95% confidence interval of the Std. Std. error difference deviation mean Lower Upper 0.0981 0.0131 -0.0568 -0.0042 0.0330 0.0044 -0.0215 -0.0038 0.1072 0.0143 -0.0616 -0.0041 0.0807 0.0107 -0.0450 -0.0018 0.0845 0.0056 -0.0360 -0.0137
Paired-sample t-test Pair 1: year 2000 Pair 2: year 2001 Pair 3: year 2002 Pair 4: year 2003 Pair 5: 2000 – 2003 Mean -0.0305 -0.0127 -0.0328 -0.0234 -0.0248
t -2.3273 -2.8848 -2.2956 -2.1708 -4.4057
- 153 -
Assessment of Productive Efficiency of Airports
Table 7.16
Comparisons of efficiency scores between Cases 1 and 3 by nonparametric paired tests Asymptotic significance (2-tailed) 0.0022 0.0014 0.0014 0.0033 0.0000 .0004 b .0002 b .0002 b .0009 b .0000 b
Nonparametric paired test A. Wilcoxon Signed-Rank test Pair 1: year 2000 Pair 2: year 2001 Pair 3: year 2002 Pair 4: year 2003 Pair 5: 2000 – 2003 Pair 1: year 2000 Pair 2: year 2001 Pair 3: year 2002 Pair 4: year 2003 Pair 5: 2000 – 2003
Z -3.0594a -3.1798a -3.1798a -2.9340a -6.0927a -6.8571
B. Sign test
a b
Based on negative ranks. Binomial distribution used. Fourth, on the maximum possible production, Tables 7.17, 7.18, 7.19 and 7.20
compare estimated potential outputs from Cases 1 and 3 in 2000, 2001, 2002, and 2003 respectively. Since the two cases tend to identify similar set of efficient airports, it is not surprising to see that the figures are very similar between cases. The difference between two cases only occurs at some airports. For example, in 2003, 11 airports (i.e., ABQ, BOS, DCA, FLL, HOU, LGB, MDW, MSY, ONT, and ORF - those with different efficiency scores) show different maximum possible production. Consequently, the figures for overall system are slightly different. Without land area (Case 3), the system may have potential to accommodate 25, 22, and 42% of passengers, movements and cargo throughput as compared to 23, 20 and 35% when having land area as another input (Case 1). Seemingly, ignoring land area does not drastically change the results. The assessment of 56 U.S. airports tends to be robust. The addition of undesirable outputs
- 154 -
Assessment of Productive Efficiency of Airports
(i.e., delayed flights and time delays) into the model made a much greater difference than the addition of another input (i.e., land area). In the latter case, the model only identifies very few more efficient airports. This finding suggests that consideration of undesirable outputs is at least as important as the consideration of additional inputs in determining the relative productivity of airports. Fifth, on the productivity growth indexes, Table 7.21 compares computed Luenberger productivity indexes between Cases 1 and 3. Since inputs to the computation (i.e., efficiency scores) are rather similar, the resulting indexes are therefore only slightly different. Between 2000 and 2001, the overall system growth is 0.9% in Case 3 as compared to 1.8% in Case 1. The gaps are narrower in the the two subsequent periods where Case 1 shows growth of 1.2% and -1.3% as compared to 1.3% and -1.0% in Case 3. The significant difference of efficiency scores between two cases are largely hidden in the computation of Luenberger indexes. A caution should be raised here if one were to perform statistical tests on the difference of Luenberger indexes.
- 155 -
Assessment of Productive Efficiency of Airports
Table 7.17 Maximum possible passengers, aircraft movements and cargo throughput in 2000, Cases 1 and 3
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM Case 1 7,572,239 5,030,557 80,162,407 3,067,777 41,712,092 4,800,070 19,602,609 24,396,829 31,173,199 6,873,998 41,581,182 16,050,706 50,206,299 106,262,017 70,620,630 34,188,468 25,282,740 23,016,542 12,170,935 42,098,078 53,509,212 1,360,280 8,106,485 43,038,655 58,688,340 67,303,182 25,374,868 637,853 15,841,801 38,897,365 15,857,218 11,769,213 Total passengers % add Case 3 20 17,571,012 0 5,030,557 0 80,162,407 0 3,067,777 50 41,722,480 1 5,508,388 0 19,602,609 84 21,725,198 35 31,173,199 0 6,873,998 84 41,581,182 1 16,793,926 30 50,206,299 75 106,262,017 99 70,620,630 0 34,188,468 59 25,282,740 0 23,016,542 34 14,351,543 109 42,098,078 52 53,509,212 0 1,360,280 53 8,106,485 31 43,038,655 59 59,913,693 0 67,303,182 0 25,374,868 0 637,853 28 15,841,801 26 38,897,365 1 25,185,292 0 11,769,213 % add 179 0 0 0 50 16 0 63 35 0 84 6 30 75 99 0 59 0 58 109 52 0 53 31 63 0 0 0 28 26 61 0 Case 1 280,925 288,919 915,454 153,917 726,532 162,504 316,703 474,220 607,944 238,011 665,523 300,889 739,636 1,438,268 931,597 450,229 459,262 345,771 339,653 657,193 714,538 90,883 227,940 648,506 815,812 783,433 383,325 379,399 279,970 540,853 301,469 388,412 Aircraft movements % add Case 3 20 368,092 0 288,919 0 915,454 0 153,917 49 726,709 1 184,494 0 316,703 43 488,684 34 607,944 0 238,011 39 665,523 1 314,654 45 739,636 72 1,438,268 68 931,597 0 450,229 57 459,262 0 345,771 33 365,560 44 657,193 48 714,538 0 90,883 53 227,940 88 648,506 56 832,349 0 783,433 0 383,325 0 379,399 28 279,970 51 540,853 1 471,021 0 388,412 % add 58 0 0 0 49 15 0 47 34 0 39 6 45 72 68 0 57 0 44 44 48 0 53 88 60 0 0 0 28 51 58 0 Case 1 130,936 1,804,221 868,286 40,722 714,502 62,433 236,043 211,276 287,627 22,572 644,908 177,857 610,884 1,579,199 1,325,558 1,082,407 515,706 441,163 150,134 567,217 1,146,512 16,815 93,435 2,382,512 1,060,137 2,038,784 71,149 49,415 244,619 545,822 99,464 2,489,078 Cargo (tons) % add Case 3 52 288,345 0 1,804,221 0 868,286 0 40,722 50 714,680 67 49,499 0 236,043 77 407,440 46 287,627 0 22,572 65 644,908 371 295,216 30 610,884 75 1,579,199 345 1,325,558 0 1,082,407 118 515,706 0 441,163 1,836 145,628 48 567,217 211 1,146,512 0 16,815 53 93,435 31 2,382,512 963 1,195,525 0 2,038,784 0 71,149 0 49,415 62 244,619 101 545,822 371 428,643 0 2,489,078 % add 234 0 0 0 50 32 0 241 46 0 65 682 30 75 345 0 118 0 1,778 48 211 0 53 31 1,099 0 0 0 62 101 1,932 0
- 156 -
Assessment of Productive Efficiency of Airports
Table 7.17 Maximum possible passengers, aircraft movements and cargo throughput in 2000, Cases 1 and 3 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Total Case 1 33,621,273 43,503,104 11,407,888 10,963,802 6,757,398 90,903,872 3,946,234 9,894,942 13,790,115 45,168,996 36,044,635 26,063,445 3,769,595 3,991,725 14,868,547 776,904 28,408,553 64,240,596 13,097,259 34,339,104 7,772,801 30,561,387 8,144,375 23,090,278 1,551,380,674 Total passengers % add Case 3 0 33,621,273 18 43,503,104 16 12,596,173 0 10,963,802 0 8,064,476 26 90,903,872 29 3,946,234 69 9,894,942 0 13,790,115 81 47,507,035 0 36,044,635 32 26,063,445 254 3,769,595 212 5,016,855 0 14,868,547 0 776,904 0 28,408,553 56 64,240,596 0 13,097,259 73 34,339,104 0 7,772,801 0 30,561,387 1,428 8,144,375 44 23,090,278 30 1,578,762,310 % add 0 18 28 0 19 26 29 69 0 91 0 32 254 292 0 0 0 56 0 73 0 0 1,428 44 33 Case 1 517,440 616,656 184,325 449,050 155,501 1,156,282 161,378 324,876 314,378 792,038 579,816 550,159 135,749 117,090 206,289 167,376 445,677 778,647 287,072 614,373 387,862 481,025 138,470 399,174 26,008,394 Aircraft movements % add Case 3 0 517,440 18 616,656 16 203,499 0 449,050 0 185,542 27 1,156,282 29 161,378 68 324,876 0 314,378 64 807,482 0 579,816 23 550,159 15 135,749 41 119,352 0 206,289 0 167,376 0 445,677 81 778,647 0 287,072 67 614,373 0 387,862 0 481,025 1 138,470 44 399,174 26 26,424,875 % add 0 18 28 0 19 27 29 68 0 67 0 23 15 44 0 0 0 81 0 67 0 0 1 44 28 Case 1 1,642,744 1,045,794 164,659 685,425 464,164 2,437,829 57,061 434,258 282,019 960,161 340,352 180,292 77,276 80,433 139,107 2,970 456,920 1,707,269 147,929 432,166 15,589 130,152 156,004 421,808 34,173,774 Cargo (tons) % add Case 3 0 1,642,744 183 1,045,794 91 204,213 0 685,425 0 553,947 66 2,437,829 97 57,061 1,971 434,258 0 282,019 72 980,276 0 340,352 23 180,292 1,224 77,276 61,299 92,445 0 139,107 0 2,970 0 456,920 96 1,707,269 0 147,929 68 432,166 0 15,589 0 130,152 381 156,004 309 421,808 43 35,253,474 % add 0 183 137 0 19 66 97 1,971 0 75 0 23 1,224 70,468 0 0 0 96 0 68 0 0 381 309 48
Note: Case 3 differs from case 1 in that it drops land area from the set of inputs. Aircraft movements include both delayed and non-
delayed flights. % add is the percentage increase from current level of the corresponding output.
- 157 -
Assessment of Productive Efficiency of Airports
Table 7.18 Maximum possible passengers, aircraft movements and cargo throughput in 2001, Cases 1 and 3
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM Case 1 6,183,606 5,107,311 75,858,500 3,012,729 38,481,290 4,487,335 20,369,923 17,510,135 25,890,177 6,680,897 22,664,832 17,476,832 37,043,730 88,971,994 45,266,906 30,558,000 18,877,562 20,151,936 11,240,147 29,224,387 34,803,580 1,284,311 6,926,081 37,470,750 35,338,310 61,606,204 21,933,000 587,473 14,358,821 28,253,061 15,681,966 11,808,247 Total passengers % add Case 3 0 6,302,898 0 5,107,311 0 75,858,500 0 3,012,729 59 38,481,290 0 4,711,491 0 20,369,923 47 17,868,321 12 25,890,177 0 6,680,897 31 22,664,832 33 17,707,265 3 37,043,730 61 88,971,994 40 45,266,906 0 30,558,000 15 18,877,562 0 20,151,936 30 12,211,045 64 29,224,387 0 34,803,580 0 1,284,311 36 6,926,081 28 37,470,750 0 36,688,573 0 61,606,204 0 21,933,000 0 587,473 19 14,358,821 0 28,253,061 0 18,358,628 0 11,808,247 % add 2 0 0 0 59 5 0 50 12 0 31 34 3 61 40 0 15 0 41 64 0 0 36 28 4 0 0 0 19 0 17 0 Case 1 242,733 284,441 890,494 148,869 703,723 159,705 324,065 428,372 514,675 243,201 506,019 321,934 540,903 1,243,721 725,372 436,420 333,261 327,006 323,801 495,596 470,916 92,402 183,398 598,329 530,328 738,114 365,716 358,508 250,145 315,752 278,734 394,826 Aircraft movements % add Case 3 0 247,415 0 284,441 0 890,494 0 148,869 55 703,723 0 167,673 0 324,065 47 437,059 12 514,675 0 243,201 31 506,019 32 326,103 12 540,903 59 1,243,721 39 725,372 0 436,420 15 333,261 0 327,006 30 351,588 25 495,596 0 470,916 0 92,402 36 183,398 105 598,329 7 564,386 0 738,114 0 365,716 0 358,508 19 250,145 0 315,752 0 325,549 0 394,826 % add 2 0 0 0 55 5 0 50 12 0 31 34 12 59 39 0 15 0 41 25 0 0 36 105 14 0 0 0 19 0 17 0 Case 1 72,876 1,873,750 739,927 35,433 628,306 32,878 225,083 150,411 292,366 15,260 493,936 132,195 420,832 1,378,600 1,442,812 795,584 209,287 337,631 74,212 414,667 337,842 14,634 82,896 1,826,652 517,649 1,774,402 52,148 53,190 236,252 223,545 15,684 2,631,631 Cargo (tons) % add Case 3 0 87,662 0 1,873,750 0 739,927 0 35,433 59 628,306 0 47,052 0 225,083 47 153,488 65 292,366 0 15,260 53 493,936 425 155,913 17 420,832 76 1,378,600 499 1,442,812 0 795,584 15 209,287 0 337,631 1,127 63,929 25 414,667 0 337,842 0 14,634 36 82,896 28 1,826,652 547 699,128 0 1,774,402 0 52,148 0 53,190 66 236,252 0 223,545 0 187,012 0 2,631,631 % add 20 0 0 0 59 43 0 50 65 0 53 519 17 76 499 0 15 0 957 25 0 0 36 28 773 0 0 0 66 0 1,092 0
- 158 -
Assessment of Productive Efficiency of Airports
Table 7.18 Maximum possible passengers, aircraft movements and cargo throughput in 2001, Cases 1 and 3 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Total Case 1 31,668,450 35,058,695 9,567,651 11,713,225 8,302,839 75,694,280 2,963,223 7,388,932 12,703,676 37,847,971 35,439,051 22,365,971 1,057,150 2,025,856 15,184,332 725,140 27,036,073 50,861,183 13,088,997 22,911,625 7,324,557 36,378,058 1,918,917 16,355,131 1,280,691,015 Total passengers % add Case 3 0 31,668,450 4 35,211,933 0 10,559,396 0 11,713,225 24 8,782,872 12 75,694,280 0 2,963,223 24 7,388,932 0 12,703,676 54 38,352,450 0 35,439,051 12 22,365,971 0 1,057,150 72 2,026,702 0 15,184,332 0 725,140 0 27,036,073 47 50,861,183 0 14,043,080 22 22,911,625 0 7,324,557 36 36,378,058 376 1,918,917 3 16,355,131 15 1,289,705,328 % add 0 4 10 0 31 12 0 24 0 56 0 12 0 73 0 0 0 47 7 22 0 36 376 3 16 Case 1 471,008 520,651 145,564 395,653 191,525 1,063,376 119,309 234,992 291,117 699,371 553,310 506,271 116,501 144,738 206,988 160,486 400,635 730,017 255,499 452,824 378,903 574,174 119,598 268,481 22,772,468 Aircraft movements % add Case 3 0 471,008 4 522,863 0 213,523 0 395,653 24 202,566 17 1,063,376 0 119,309 24 234,992 0 291,117 50 708,189 0 553,310 12 506,271 0 116,501 72 144,798 0 206,988 0 160,486 0 400,635 88 730,017 0 273,871 22 452,824 0 378,903 21 574,174 5 119,598 3 268,481 16 23,015,097 % add 0 4 47 0 31 17 0 24 0 52 0 12 0 72 0 0 0 88 7 22 0 21 5 3 17 Case 1 1,639,760 353,018 75,700 593,634 519,100 1,567,465 28,786 31,619 242,967 826,639 283,337 155,931 4,976 21,966 134,689 2,946 400,499 934,037 143,914 313,835 14,849 395,215 20,924 217,161 26,455,537 Cargo (tons) % add Case 3 0 1,639,760 4 354,561 0 184,203 0 593,634 24 549,112 21 1,567,465 0 28,786 53 31,619 0 242,967 54 837,658 0 283,337 12 155,931 0 4,976 23,776 19,408 0 134,689 0 2,946 0 400,499 47 934,037 0 514,508 45 313,835 0 14,849 223 395,215 5 20,924 172 217,161 24 27,372,929 % add 0 4 143 0 31 21 0 53 0 56 0 12 0 20,996 0 0 0 47 258 45 0 223 5 172 29
Note: Case 3 differs from case 1 in that it drops land area from the set of inputs. Aircraft movements include both delayed and non-
delayed flights. % add is the percentage increase from current level of the corresponding output.
- 159 -
Assessment of Productive Efficiency of Airports
Table 7.19 Maximum possible passengers, aircraft movements and cargo throughput in 2002, Cases 1 and 3
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM Case 1 6,151,129 4,914,539 76,876,128 2,810,791 35,647,876 4,620,683 19,012,529 17,578,887 29,952,090 6,741,354 37,620,386 13,941,695 38,229,992 91,816,824 54,507,100 30,258,481 23,131,236 19,749,905 11,577,573 31,569,526 51,048,711 1,221,138 7,432,943 40,849,751 42,492,287 56,223,843 21,986,679 1,453,412 13,342,966 27,674,633 17,371,036 11,141,594 Total passengers % add Case 3 0 16,538,636 0 4,914,539 0 76,876,128 0 2,810,791 57 35,662,137 0 4,620,683 0 19,012,529 68 17,729,702 27 29,952,090 0 6,741,354 81 37,620,386 8 15,055,396 7 38,229,992 74 91,816,824 68 54,507,100 4 32,154,783 36 23,131,236 0 19,749,905 44 11,957,427 85 31,569,526 51 51,048,711 0 1,221,138 50 7,432,943 36 40,849,751 21 44,094,440 0 56,223,843 0 21,986,679 0 1,453,412 30 13,342,966 4 27,674,633 0 25,651,538 0 11,141,594 % add 169 0 0 0 57 0 0 70 27 0 81 17 7 74 68 10 36 0 49 85 51 0 50 36 26 0 0 0 30 4 48 0 Case 1 254,874 277,267 889,966 146,555 610,900 162,211 304,921 420,585 576,384 255,630 582,466 252,786 558,511 1,302,154 815,124 485,889 379,747 323,726 354,136 524,229 743,794 82,883 187,219 584,615 602,418 645,424 362,439 350,603 248,280 403,408 304,304 398,769 Aircraft movements % add Case 3 0 298,929 0 277,267 0 889,966 0 146,555 56 611,140 0 162,211 0 304,921 67 424,159 27 576,384 0 255,630 20 582,466 17 251,933 13 558,511 70 1,302,154 66 815,124 20 492,992 35 379,747 0 323,726 44 365,708 41 524,229 63 743,794 0 82,883 50 187,219 103 584,615 21 623,900 0 645,424 0 362,439 0 350,603 30 248,280 39 403,408 0 446,523 0 398,769 % add 17 0 0 0 56 0 0 68 27 0 20 17 13 70 66 22 35 0 49 41 63 0 50 103 26 0 0 0 30 39 47 0 Case 1 74,460 1,771,595 734,083 32,353 609,352 39,751 251,354 170,482 280,897 10,700 420,995 131,868 506,884 1,165,008 764,041 880,784 224,076 414,947 99,629 459,860 514,101 13,863 103,415 2,168,672 983,493 1,779,855 32,223 53,356 257,344 358,923 26,309 3,390,800 Cargo (tons) % add Case 3 0 340,854 0 1,771,595 0 734,083 0 32,353 57 609,596 0 39,751 0 251,354 68 171,945 75 280,897 0 10,700 20 420,995 2,150 157,961 52 506,884 74 1,165,008 228 764,041 4 935,983 36 224,076 0 414,947 1,756 59,656 42 459,860 56 514,101 0 13,863 50 103,415 36 2,168,672 1,100 621,355 0 1,779,855 0 32,223 0 53,356 89 257,344 81 358,923 0 249,193 0 3,390,800 % add 358 0 0 0 57 0 0 70 75 0 20 2,595 52 74 228 10 36 0 1,011 42 56 0 50 36 658 0 0 0 89 81 847 0
- 160 -
Assessment of Productive Efficiency of Airports
Table 7.19 Maximum possible passengers, aircraft movements and cargo throughput in 2002, Cases 1 and 3 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Total Case 1 30,060,241 39,838,014 9,251,773 13,005,642 6,517,050 66,565,952 3,579,015 8,780,743 12,241,975 24,799,470 35,547,167 23,785,515 1,345,970 1,108,695 14,931,854 728,307 26,690,843 55,169,980 14,638,695 20,741,521 7,903,066 38,825,764 362,017 18,015,387 1,323,382,375 Total passengers % add Case 3 0 30,060,241 22 39,838,014 0 9,684,859 0 13,005,642 0 7,579,751 0 66,565,952 3 3,605,579 60 8,780,743 0 12,241,975 0 40,145,233 0 35,547,167 32 23,785,515 0 1,345,970 0 1,108,695 0 14,931,854 0 728,307 0 26,690,843 75 55,169,980 32 15,884,555 11 20,741,521 0 7,903,066 52 38,825,764 0 362,017 16 18,015,387 22 1,365,321,443 % add 0 22 5 0 16 0 4 60 0 62 0 32 0 0 0 0 0 75 43 11 0 52 0 16 26 Case 1 446,235 605,830 139,291 371,988 149,292 922,817 129,782 265,888 276,877 463,167 545,771 559,932 130,826 85,243 206,380 159,835 364,735 756,932 296,477 452,571 368,627 604,053 123,642 283,457 23,101,862 Aircraft movements % add Case 3 0 446,235 19 605,830 0 156,839 0 371,988 0 171,318 0 922,817 3 130,745 59 265,888 0 276,877 0 729,445 0 545,771 32 559,932 0 130,826 0 85,243 0 206,380 0 159,835 0 364,735 115 756,932 43 295,878 11 452,571 0 368,627 38 604,053 0 123,642 16 283,457 21 23,637,472 % add 0 19 13 0 15 0 4 59 0 57 0 32 0 0 0 0 0 115 43 11 0 38 0 16 24 Case 1 1,624,242 383,425 84,150 634,643 496,547 1,473,980 52,005 54,643 245,134 541,039 298,945 184,943 4,518 74 151,644 2,832 374,753 1,034,300 184,570 297,407 13,730 352,447 13,257 268,010 27,496,714 Cargo (tons) % add Case 3 0 1,624,242 20 383,425 0 196,986 0 634,643 0 569,902 0 1,473,980 58 50,334 205 54,643 0 245,134 0 875,831 0 298,945 32 184,943 0 4,518 0 74 0 151,644 0 2,832 0 374,753 75 1,034,300 32 234,260 38 297,407 0 13,730 173 352,447 0 13,257 192 268,010 24 28,235,881 % add 0 20 134 0 15 0 53 205 0 62 0 32 0 0 0 0 0 75 67 38 0 173 0 192 27
Note: Case 3 differs from case 1 in that it drops land area from the set of inputs. Aircraft movements include both delayed and non-
delayed flights. % add is the percentage increase from current level of the corresponding output.
- 161 -
Assessment of Productive Efficiency of Airports
Table 7.20 Maximum possible passengers, aircraft movements and cargo throughput in 2003, Cases 1 and 3
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM Case 1 6,935,011 4,791,431 79,086,792 2,672,637 35,020,347 4,729,936 20,094,756 16,180,712 31,635,406 6,252,061 38,539,652 14,214,803 38,894,115 83,244,552 50,696,543 29,431,061 24,484,401 19,732,556 10,193,688 28,154,745 54,898,411 1,215,093 6,970,535 40,417,649 54,332,887 54,982,838 22,482,770 2,875,703 13,040,110 30,308,286 18,426,397 11,437,307 Total passengers % add Case 3 15 15,625,565 0 4,791,431 0 79,086,792 0 2,672,637 54 35,114,859 0 4,729,936 0 20,094,756 53 17,281,283 37 31,635,406 0 6,252,061 82 38,539,652 0 14,823,901 4 38,894,115 56 83,244,552 55 50,696,543 0 29,431,061 36 24,493,287 0 19,732,556 31 11,316,924 68 28,154,745 61 54,898,411 0 1,215,093 43 6,970,535 27 40,417,649 50 54,332,887 0 54,982,838 0 22,482,770 0 5,103,201 34 13,040,110 11 30,308,286 0 27,880,730 0 11,437,307 % add 158 0 0 0 54 0 0 64 37 0 82 4 4 56 55 0 37 0 45 68 61 0 43 27 50 0 0 77 34 11 51 0 Case 1 253,241 277,361 911,723 154,849 569,476 178,079 299,469 394,639 536,801 237,979 576,589 250,802 541,268 1,185,676 756,815 405,808 389,706 319,989 316,623 465,444 743,930 79,377 172,836 570,852 743,556 622,378 374,952 338,807 229,168 380,232 328,035 402,258 Aircraft movements % add Case 3 15 312,946 0 277,361 0 911,723 0 154,849 53 570,992 0 178,079 0 299,469 53 421,281 21 536,801 0 237,979 14 576,589 0 261,399 8 541,268 55 1,185,676 54 756,815 0 405,808 36 389,845 0 319,989 30 351,390 39 465,444 57 743,930 0 79,377 43 172,836 104 570,852 48 743,556 0 622,378 0 374,952 0 349,176 34 229,168 29 380,232 0 491,221 0 402,258 % add 42 0 0 0 53 0 0 63 21 0 14 4 8 55 54 0 36 0 45 39 57 0 43 104 48 0 0 3 34 29 50 0 Case 1 133,954 2,102,025 798,501 34,184 943,554 44,654 235,576 272,944 334,987 10,766 449,437 5,774 884,820 1,043,533 710,246 874,641 226,674 421,930 155,183 399,990 683,510 10,957 104,553 2,071,962 510,720 1,833,300 28,402 50,873 265,216 860,080 23,266 3,390,515 Cargo (tons) % add Case 3 87 305,789 0 2,102,025 0 798,501 0 34,184 160 1,001,459 0 44,654 0 235,576 185 458,278 139 334,987 0 10,766 14 449,437 0 403,062 172 884,820 56 1,043,533 222 710,246 0 874,641 45 227,763 0 421,930 2,587 333,060 40 399,990 79 683,510 0 10,957 48 104,553 27 2,071,962 522 510,720 0 1,833,300 0 28,402 0 52,431 94 265,216 346 860,080 0 744,163 0 3,390,515 % add 327 0 0 0 176 0 0 379 139 0 14 6,881 172 56 222 0 46 0 5,667 40 79 0 48 27 522 0 0 3 94 346 3,098 0
- 162 -
Assessment of Productive Efficiency of Airports
Table 7.20 Maximum possible passengers, aircraft movements and cargo throughput in 2003, Cases 1 and 3 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Total Case 1 29,595,618 41,741,976 10,542,505 13,548,363 7,258,006 69,508,672 3,849,419 10,029,678 12,395,938 39,631,312 37,412,165 18,913,688 1,361,758 1,246,842 15,260,791 752,762 26,755,888 52,079,506 10,677,903 21,972,152 8,535,130 35,870,254 393,530 21,158,079 1,346,865,126 Total passengers % add Case 3 0 29,595,618 26 41,741,976 14 11,335,042 0 13,548,363 11 7,945,867 0 69,508,672 12 3,888,687 67 10,029,678 0 12,395,938 61 39,631,312 0 37,412,165 33 18,913,688 0 1,361,758 0 1,246,842 0 15,260,791 0 752,762 0 26,755,888 78 52,079,506 0 10,677,903 19 21,972,152 0 8,535,130 76 35,870,254 0 393,530 36 21,158,079 23 1,371,693,477 % add 0 26 22 0 21 0 13 67 0 61 0 33 0 0 0 0 0 78 0 19 0 76 0 36 25 Case 1 417,423 595,117 156,040 342,871 162,248 928,691 135,951 284,004 267,052 700,772 541,771 478,550 127,197 93,068 203,285 152,485 354,770 600,665 198,082 476,174 350,074 568,680 112,284 317,570 22,573,542 Aircraft movements % add Case 3 0 417,423 17 595,117 14 167,701 0 342,871 11 177,587 0 928,691 12 137,337 65 284,004 0 267,052 57 700,772 0 541,771 32 478,550 0 127,197 0 93,068 0 203,285 0 152,485 0 354,770 80 600,665 0 198,082 19 476,174 0 350,074 50 568,680 0 112,284 36 317,570 20 22,908,850 % add 0 17 22 0 21 0 13 65 0 57 0 32 0 0 0 0 0 80 0 19 0 50 0 36 22 Case 1 1,637,278 370,000 253,278 597,383 574,965 1,510,746 36,821 175,200 239,265 842,526 288,350 294,139 4,569 103 135,547 2,825 351,418 1,682,768 108,622 402,086 12,050 299,115 19,024 649,238 30,404,043 Cargo (tons) % add Case 3 0 1,637,278 17 370,000 213 276,540 0 597,383 11 629,456 0 1,510,746 14 36,532 857 175,200 0 239,265 61 842,526 0 288,350 142 294,139 0 4,569 0 103 0 135,547 0 2,825 0 351,418 193 1,682,768 0 108,622 85 402,086 0 12,050 159 299,115 0 19,024 595 649,238 35 32,195,292 % add 0 17 242 0 21 0 13 857 0 61 0 142 0 0 0 0 0 193 0 85 0 159 0 595 42
Note: Case 3 differs from case 1 in that it drops land area from the set of inputs. Aircraft movements include both delayed and non-
delayed flights. % add is the percentage increase from current level of the corresponding output.
- 163 -
Assessment of Productive Efficiency of Airports
Table 7.21 Luenberger productivity indexes, Cases 1 and 3
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM
2000 - 2001 Case 1 Case 3 0.136 0.079 0.000 0.000 0.310 0.310 0.006 -0.002 -0.078 -0.079 -0.002 -0.001 0.015 0.015 -0.095 -0.127 0.040 0.040 0.000 0.000 -0.170 -0.170 -0.216 -0.220 -0.019 -0.019 0.023 0.023 0.060 0.060 0.011 0.011 0.074 0.102 0.000 0.000 0.002 -0.073 -0.048 -0.048 0.104 0.104 0.077 0.077 0.042 0.042 -0.116 -0.116 0.189 0.235 0.000 0.000 0.202 0.475 -0.016 -0.014 -0.025 -0.025 -0.015 -0.015 -0.103 -0.002 -0.062 0.000
2001 - 2002 Case 1 Case 3 0.093 0.009 0.000 0.000 0.000 0.000 0.000 0.000 -0.027 -0.023 0.079 0.000 0.096 0.096 -0.090 -0.068 -0.003 -0.003 0.012 0.012 0.219 0.219 0.041 0.091 0.033 0.033 -0.001 -0.001 -0.086 -0.086 0.167 0.124 -0.045 -0.045 0.000 0.000 -0.007 0.001 -0.064 -0.064 -0.102 -0.102 0.026 0.026 -0.029 -0.029 0.089 0.089 -0.002 0.017 0.000 0.000 0.030 0.001 0.000 0.000 -0.062 -0.062 -0.009 -0.009 -0.002 0.000 0.000 0.000
2002 - 2003 Case 1 Case 3 -0.079 -0.090 0.000 0.000 0.000 0.000 0.028 0.029 -0.020 -0.029 0.000 0.000 0.000 0.000 0.002 0.035 -0.021 -0.021 -0.012 -0.012 0.042 0.042 0.093 0.060 -0.008 -0.008 0.055 0.055 -0.052 -0.052 0.023 0.050 0.021 0.009 0.000 0.000 -0.014 0.007 -0.055 -0.055 -0.032 -0.032 0.000 0.000 -0.029 -0.029 0.043 0.043 -0.036 -0.111 0.000 0.000 0.009 0.173 -0.018 -0.001 -0.072 -0.072 -0.021 -0.021 0.045 0.000 0.000 0.000
- 164 -
Assessment of Productive Efficiency of Airports
Table 7.21 Luenberger productivity indexes, Cases 1 and 3 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Average index Number of regress Number of no change Number of progress
2000 - 2001 Case 1 Case 3 0.000 0.000 -0.003 -0.001 -0.049 -0.012 -0.013 -0.013 -0.032 -0.077 0.555 0.264 0.162 0.162 -0.078 -0.078 -0.015 -0.015 -0.022 0.204 0.075 0.265 -0.016 -0.016 0.023 0.023 -0.027 -0.056 0.000 0.000 -0.010 -0.010 -0.026 -0.026 -0.105 -0.105 -0.047 0.000 -0.061 -0.061 0.000 0.000 -0.109 -0.109 -0.070 -0.070 0.066 0.066 0.009 0.996 28 29 9 7 19 20
2001 - 2002 Case 1 Case 3 0.000 0.000 -0.029 -0.030 0.036 0.046 0.000 0.000 0.043 0.069 0.170 0.170 -0.020 -0.017 -0.138 -0.138 0.000 0.000 0.016 0.047 0.019 0.020 -0.066 -0.066 0.003 0.001 0.270 0.314 0.000 0.000 -0.002 -0.002 -0.003 0.000 0.013 0.014 -0.151 -0.158 0.096 0.096 0.000 0.000 -0.087 -0.087 0.191 0.191 -0.015 -0.015 0.013 0.681 19 24 15 11 22 21
2002 - 2003 Case 1 Case 3 -0.011 0.000 -0.002 -0.002 -0.081 -0.068 0.000 0.000 0.008 -0.054 0.134 0.134 -0.049 -0.047 -0.009 -0.009 -0.003 -0.003 -0.014 -0.116 0.008 0.008 -0.138 -0.138 0.003 0.002 0.000 0.000 0.000 0.000 -0.021 -0.018 0.000 0.000 -0.056 -0.056 0.106 0.019 -0.056 -0.056 -0.001 0.000 -0.171 -0.171 -0.005 -0.005 -0.093 -0.093 -0.010 -0.013 27 29 15 12 14 15
Note: The negative index indicates regressed productivity. Zero value means that there is
no change in the productivity between two years. The positive index indicates productivity growth.
- 165 -
Assessment of Productive Efficiency of Airports
Tables 7.22 and 7.23 show the test results from paired-sample t-test and the two nonparametric tests. The t-statistics in Table 7.22 (0.8926, -0.1367, -0.5812, and 0.5096) indicate that the differences in Luenberger indexes between two cases are not statistically significant. Results from nonparametric tests in Table 7.23 support the same inference. The differences in Luenberger indexes are not statistically significant, although the efficiency scores themselves are statistically different (see results in Tables 7.15 and 7.16). The computation of Luenberger indexes really can conceal the difference in efficiency level and in turn can provide a misleading interpretation.
Table 7.22
Comparisons of Luenberger productivity indexes by paired sample t-test Cases 1and 3 Paired differences Cases 1 and 3 95% confidence interval of the Std. Std. error difference deviation mean Lower Upper 0.0708 0.0094 -0.0105 0.0274 0.0205 0.0027 -0.0058 0.0051 0.0331 0.0044 -0.0114 0.0062 0.0466 0.0036 -0.0052 0.0089
Paired-sample t-test Pair 1: 2000/01 Pair 2: 2001/02 Pair 3: 2002/03 Pair 4: 2000 – 03 Mean 0.0084 -0.0003 -0.0025 0.0018
t 0.8926 -0.1367 -0.5812 0.5096
- 166 -
Assessment of Productive Efficiency of Airports
Table 7.23
Comparisons of Luenberger productivity indexes by nonparametric paired tests Asymptotic significance (2-tailed) 0.4140 0.3477 0.6264 0.5791 0.6636b 0.1892b 1.0000a 0.2077
Nonparametric paired test A. Wilcoxon Signed-Rank test Pair 1: years 2000/01 Pair 2: years 2001/02 Pair 3: years 2002/03 Pair 4: years 2000 - 03 Pair 1: years 2000/01 Pair 2: years 2001/02 Pair 3: years 2002/03 Pair 4: years 2000 – 03
Z -0.8168a -0.9388a -0.4867c -0.5546a -1.2598
B. Sign test
a b c
Based on positive rank Binomial distribution used Based on negative rank
Determination of airport productivity
7.5
A causal model is developed to explain the variations in the efficiency score. For planning and managing an airport, the model will be very useful for predicting future productivity based on given information. The information is treated as exploratory variables which may include number of passengers per runway, passengers per movement, average delay, percentage of international passengers, etc. The dependent variable is the efficiency score. The model presented here is for Case 1 (with consideration of delays) which is considered as a more complete assessment of productive efficiency. By the nature of the directional output distance function, the value of efficiency scores can only be in the range of zero to infinity. Because of this special type of limited dependent variable,
- 167 -
Assessment of Productive Efficiency of Airports
simple regression is not an appropriate model. The issue was discussed earlier in chapter 5. Censored Tobit regression is employed. In this case, efficiency score of airport y i is represented by Equation (7.2)
?? xi + ? i yi = ? ?0
if y i > 0 if y i ? 0
(7.2)
y i is an efficiency score that is observable for values greater than 0 and is censored for values less than or equal to 0. Efficiency scores of all efficient airports are censored at 0, regardless of values of independent variables xi . ? and ? i are the coefficients and the error term of the Tobit model respectively. Coefficients ? can be estimated using Maximum Likelihood (ML) method. ML estimation for the Tobit model involves dividing the observations into two sets. The first set contains uncensored observations. The second set contains censored observations. The log-likelihood function is given in Equation (5.3). Meanwhile the marginal effect with respect to an exploratory variable can be computed using Equation (5.4). To measure the goodness-of-fit, the R2 ANOVA given in Equation (5.5) is computed. LIMDEP version 8.0 (Greene, 2002) is used to estimate the model.
7.6 Factors affecting productive efficiency of U.S. airports
The airport operation is a complex process involving a large number of activities. There are many variables that can affect the operational efficiency. Five groups of variables are investigated. The proxy of each group entering the model is essentially based on data availability.
- 168 -
Assessment of Productive Efficiency of Airports
First, Airport characteristics are represented here by physical characteristics. These are basically input measures that are used in the directional output distance function, i.e., land area (LAND), number of runways (RW) and runway area (RWA). These inputs certainly play a major role in accommodating traffic. However, one should be aware that having more of these inputs does not necessarily mean more outputs. Second, Airport services are mainly represented by outputs of airport operations which consist of number of aircraft movements (AIR), passengers (PAX) and cargo throughput (CARGO). One would expect that more services contribute to higher efficiency. However, this is not necessarily true since efficiency takes into account both inputs and outputs. Accordingly, another group of variables is introduced, i.e., level of utilization. Third, Level of utilization may be a better determinant of operational efficiency since it takes into accounts both input and output measures. This case study considers many ratio variables, such as non-delayed flights/land area, non-delayed flights/ /runway, non-delayed flights/ runway acreage area, annual total passengers/land area, annual total passengers/number of runways, annual total passengers/runway acreage area, annual cargo throughput/runway acreage area etc. Intuitively, higher values of these ratios should result in more efficient operation. Fourth, Market characteristics include target market (e.g., passengers, aircraft operation, cargo, general aviation and military service), market share, market dominance, market focus (e.g., domestic, international, tourist, business passengers), whether the airport is an airport in a multiple airport system, whether the airport is a hub airport
- 169 -
Assessment of Productive Efficiency of Airports
according to FAA definition, and irregularity of time periods. An attempt was made to collect these variables as much as possible. After all, six variables are entered in the estimation. These include the percentage of international passengers, percentage of general aviation, whether the airport is an airport in MAS, whether the airport is any kind of FAA hub (i.e., large, medium, small, or non-hub), whether the airport dominates in its corresponding MAS, and irregularity of years. In addition, several interaction variables (e.g., whether the airport is an airport in an MAS and also dominates the market etc.) are also tested. The fifth group of variables is Service characteristics. Not only the model
considers quantity of airport services, but it also aims to investigate the effects of service quality on the productive efficiency. As pointed out from the results in case study 1, unless quality of services is taken into account, only busy and congested airports will be classified as efficient. This case study takes delays as a proxy to represent service characteristics. Different ratios are computed and entered as exploratory variables. These include percentage of delayed flights, delayed-flights per runway, average delay per passenger and average delay per movement. It is expected that the lower values of these ratios should indicate the higher productive efficiency. In other words, the sign of the coefficients should be positive. Note that ownership/management characteristic, which is one significant variable in case study 1, is not considered here because there is no difference across airports in the dataset. Every airport, except Stewart International (SWF) is publicly owned and
- 170 -
Assessment of Productive Efficiency of Airports
operated. The inclusion of a dummy variable to represent private ownership does not give the meaningful results. The model is estimated from the pooled 4-year data (i.e., the years 2000 – 2003). Airports with incomplete independent variables are taken out. Totally, there are 211 complete observations or 13 samples shorter than the full sample size of 224 (i.e., 56 airports x 4 years). Several models consisting of different combinations of exploratory variable were estimated. Table 7.24 shows final model estimation results. It has nine independent variables, including the constant. Other variables are dropped off for reasons such as high correlation among themselves, being insignificant or having illogical sign. Recall that the lower efficiency score is desirable because it indicates that an airport is more efficient. As a result, a negative sign of the three utilization ratio variables in the model, i.e., non-delayed flights per land area (-0.1451 x 10-2), non-delayed-flights per runway area (-0.5986 x 10-4) and cargo throughput per runway area (-0.7212 x 10-4) contribute to higher productive efficiency. They are statistically significant at above the 95% confidence level. The marginal effects in the last column indicate changes in efficiency scores with respect to the changes in the corresponding exploratory variables. For instance, an increment of one non-delayed flight per acre of runway would result in an airport becoming more efficient by -0.2917 x 10-4 units.
- 171 -
Assessment of Productive Efficiency of Airports
Table 7.24 Censored Tobit regression model estimation results
Variables Constant % International Passengers Non-delayed flights/Land area Non-delayed flights/Runway area Cargo/ Runway area Delay/Passenger Y2001 Y2002 Y2003 Number of observations = 211 Log Likelihood function = -65.8723
Proposed model 2 R ANOVA = 1.3106 Coefficient 0.6167** (6.950) 0.0104** (3.646) -0.1451 x 10-2** (-4.612) -0.5986 x 10-4** (-2.156) -0.7212 x 10-4** (-5.411) 0.1033 x 10-4** (4.896) -0.1724** (-2.730) -0.1200* (-1.893) -0.1171* (-1.857) Marginal 0.3006** (6.276) 0.0051** (3.796) -0.0707 x 10-2** (-4.672) -0.2917 x 10-4** (-2.178) -0.3515 x 10-4** (-6.250) 0.0503 x 10-4** (5.105) -0.0840** (-2.703) -0.0585* (-1.890) -0.0571* (-1.857)
Note: Dependent variable = Efficiency score Y2001 = 1 if compute performance score in year 2001, otherwise = 0 Y2002 = 1 if compute performance score in year 2002, otherwise = 0 Y2003 = 1 if compute performance score in year 2003, otherwise = 0 ** Estimated coefficient is significant at the 0.05 level (one-tailed) * Estimated coefficient is significant at the 0.10 level (one-tailed)
Percentage of international passengers (coefficient = 0.0104) is positively associated with the efficiency score. The higher proportion of international passengers leads to lower efficiency. This may be well explained by the longer service time of this target market in comparison to domestic passengers. In general an airport uses more
- 172 -
Assessment of Productive Efficiency of Airports
resources to service an international passenger than it does to service a domestic passenger. The marginal effect suggests that for every additional percent of international passenger, an airport would be less efficient by 0.0051 units. The average delay per passenger (coefficient = 0.1033 x 10-4) is also positively associated with the efficiency score as expected. It is understandable that the higher delay leads to lower efficiency but this rarely has been quantified in the past. In this model, the estimation of the marginal effect suggests that for every additional minute of average delay per passenger, an airport would become less efficient by 0.0503 x 10-4 units. In addition to the above mentioned variables, there may be some effects from anomalies in 2001, 2002 and 2003 since these dummy variables are also statistically significant. The negative signs for these years indicate that airports become efficient slightly more easily in comparison to year 2000. In summary, the case study has assessed the airport productivity of 56 U.S. airports where joint production of desirable and undesirable outputs is taken into consideration. It also compares results with the case that undesirable outputs are ignored as this is the case in previous studies (See Table 2.4 for the list and description). In the last part of the chapter a productivity prediction model was developed using the Censored Tobit Regression. It is found that the increment of factors such as non-delayed flights per land area, non-delayed flights per runway area, and cargo throughput per runway area contribute to the enhancement of productivity. Meanwhile, the higher proportion of international passengers and average delay per passenger tend to reduce the productivity of airport operations. The model captures anomaly effects in the years 2001, 2002 and
- 173 -
Assessment of Productive Efficiency of Airports
2003 by indicating that airports could become efficient slightly more easily than the year 2000, ceteris paribus. In the next chapter, the important findings and insights will be summarized. Potential future research extensions are also suggested.
- 174 -
Assessment of Productive Efficiency of Airports
CHAPTER 8 CONCLUSIONS AND FUTURE RESEARCH
8.1 Conclusions from assessing productivity of airports in MASs
Arguing that it may be more useful to analyze productivity of airports operating in a similar market structure, the case study focused on airports in multiple airports systems (MASs). The data set consisted of 72 airports from 25 MASs in North America, South America, Europe and Asia. Data Envelopment Analysis (DEA) technique was used to assess the relative efficiencies of these airports. It was assumed that land area, number of runways, and runway area were the proxies of operational inputs whereas number of annual aircraft movements and passengers were two main target outputs from the operations. The analysis period was 2000 – 2002.
8.1.1 Productivity of airports in MASs
The assessment indicates that there are two groups of efficient or highly productive airports, coined by the busy and the compact. The busy group consists of market leaders in large MASs such as O’Hare International (ORD), Los Angeles International (LAX) and Heathrow/London (LHR). Air traffic statistics (ACI, 2002 – 2004) confirm that they are among the busiest airports in the world. Airports in the compact group are alternative airports with relatively small land area and only have one or two runways. Clearly airports in both groups are classified as efficient airports because of their relatively higher runway utilization. In this respect, larger airport size does not
- 175 -
Assessment of Productive Efficiency of Airports
guarantee high efficiency. An implication from this result is that an airport must be very busy; otherwise it would not be regarded as an efficient airport. This may make good sense as long as such high utilization does not create undesirable congestion and delays.
8.1.2 Underutilized airports
It is very difficult for all airports in an MAS to be highly utilized because total air travel demand must somehow be distributed among airports (Caves and Gosling, 1999; de Neufville, 1995; de Neufville and Odoni, 2003; Pathomsiri and Haghani, 2005; Pathomsiri, Mahmassani and Haghani, 2004). The effort to manage them by either coordinating or regulating air traffic has not been successful in most cases (Caves and Gosling, 1999; Charles River, 2001; de Neufville, 1995; de Neufville and Odoni, 2003). Given that the capital investment in airport business is very lumpy, it is extremely difficult to keep all runways in an MAS busy (New York/New Jersey region may be an exception). Consequently, functional failure is followed. This seems to be the case in this case study. It is found that some airports such as Montreal-Mirabel (YMX), Glasgow Prestwick International (GLA), Schoenefeld/Berlin (SXF), and Tempelhof/Berlin (THF) are underutilized. In fact, Montreal-Mirabel (YMX) is a case study of an unsuccessful airport in textbooks (Caves and Gosling, 1999; de Neufville, 1995; de Neufville and Odoni, 2003). Schoenefeld/Berlin (SXF), and Tempelhof/Berlin (THF) and Tegel (TXL, another airport in the Berlin MAS) are planned to be consolidated in 2011. Construction is underway (Berlin Brandenburg International, 2005). In this sense, the proposed models in this case study are useful in pointing out over-investment.
- 176 -
Assessment of Productive Efficiency of Airports
8.1.3
Factors affecting productive efficiency of airports in MASs
The case study estimated a Censored Tobit regression model for explaining variations in efficiency scores of airport operations. Five groups of exploratory variables were investigated, i.e., airport characteristics, airport services, level of utilization, market characteristics, and ownership/management characteristics. It was found that factors such as utilization of land area and runway area, passengers per aircraft movement, market dominance and privately-operated management style contribute to the enhancement of productivity. Meanwhile higher proportion of international passengers tends to reduce the productivity. The model also captured anomaly effects in year 2002 (it was observed that an airport could become efficient slightly more easily with the same utilization rate, possibly due to a global drop in air traffic after the September 11 terrorist attacks). Given some planned measures, the model can be used to predict future productivity of an airport and should be very useful as a tool for planning airport business in a competitive market.
8.2 Conclusions from assessing productivity of U.S. airports
The traditional measurement of productive efficiency of airport operations typically focuses on marketable outputs such as throughput of passengers, aircraft movements and cargo. As confirmed by case study 1, the typical results indicate that efficient airports are very busy airports and frequently they are congested. Reduction in so-called “undesirable outputs” such as delays has never been given credit in the assessment although it is a major concern in airport management. Case study 2 aimed at re-assessing productive efficiency of airport operations by considering joint production of desirable and undesirable outputs.
- 177 -
Assessment of Productive Efficiency of Airports
To estimate relative productivity, airports are viewed as similar production units taking three representatives of capital inputs, i.e., land area, number of runways and runway area; then producing three main desirable outputs, i.e., aircraft movements, passengers and cargo throughput. By the nature of airport operation, there are two byproduct outputs, though undesirable, i.e., delayed flights and time delays. The efficient airports are the ones that both achieve relatively high levels of desirable outputs while keeping the undesirable outputs at relatively low levels. Mathematically speaking, the model ought to simultaneously maximize desirable outputs and minimize undesirable outputs. Data Envelopment Analysis (DEA) seems inappropriate since it seeks to maximize all outputs simultaneously. This dissertation proposed to use the nonparametric direction output distance function. The model is a linear programming problem. Solving it can identify a set of airports that form a linear piecewise efficient production frontier. For inefficient airports, it quantifies the levels of inefficiency. In addition, the maximum possible production can also be estimated to understand how much the potential outputs are. Results are beneficial in many management regards such as performance measurement, benchmarking, ranking and policy development. The approach was applied in case study 2 to assess the productivity of 56 major commercial U.S. airports. A recent panel data from 2000 – 2003 were used. In order to analyze the impact of the inclusion of the undesirable outputs, a model without accounting for delays was also estimated. There are several important findings and
- 178 -
Assessment of Productive Efficiency of Airports
insightful implications as discussed in Chapter 7. This chapter concludes with those discussions.
8.2.1 Productivity of U.S. commercial airports
Among 56 airports, approximately half of them are identified as efficient during 2000 – 2003. These airports include busy ones such as Hartsfield-Jackson Atlanta (ATL), Los Angeles International (LAX), LaGuardia (LGA), Memphis (MEM), Phoenix Sky Harbor International (PHX), San Diego International (SAN), and John Wayne airport (SNA). Other well-known busy airports such as O'Hare International (ORD), Midway International (MDW), Newark Liberty International (EWR), John F. Kennedy International (JFK), Anchorage International (ANC), Miami International (MIA), Seattle Tacoma International (SEA) and Lambert-St. Louis International (STL) though not classified as efficient show very low inefficiency levels. In addition, the model also identified several other less busy airports as efficient They include Birmingham International (BHM), Baltimore/Washington International (BWI), Port Columbus International (CMH), and Oakland International (OAK). These airports are credited because they have relatively low delays. The results indicate that there may be a balance between quantity and quality of outputs in the achievement of efficient outcomes; i.e., airports can trade-off utilization levels for reduced flight and time delays. For certain stakeholders, this option may be an optimal strategy. Passengers and shippers receive service with fewer flight delays. The FAA, as the regulator, has less concern over congestion and safety. Meanwhile, airport managers are able to balance
- 179 -
Assessment of Productive Efficiency of Airports
traffic volume with customer satisfaction. By all accounts, the inclusion of undesirable outputs in the analysis appears to provide a fairer assessment of airport efficiency. In 2003, the overall system (56 airports) has potential to accommodate increases of about 30% (~1,550 million passengers), 26% (~ 26 million movements) and 43% of total passengers, aircraft movements and cargo throughput, respectively. In numbers, these amounts are equivalent to totally 1,550 million passengers, 26 million movements and 34 million tons of cargo. This would make the system operate at the maximum possible production level. The estimated potential outputs vary across airports. For airport planning, the figure provides a good estimation of excess capacity. An airport manager may use this information in planning an airport improvement program. Finally, it is observed that when there is a major new investment in an airport, its productivity decreases during early years after the construction. This is the case for Detroit Metropolitan Wayne County (DTW) and George Bush Intercontinental (IAH). DTW opened its sixth runway in 2001 whereas IAH finished constructing runway expansion and extension in 2002.
8.2.2 Productivity growth of U.S. commercial airports
During the period 2000 – 2003, the changes in productivity are rather modest in the narrow range of -1.3% to +1.8%. The airport system on the average had productivity gains in two periods during 2000/2001 and 2001/2002 and productivity loss in 2002/2003. Between 2000 and 2001, the overall average rise in efficiency was 1.8 percent; and continued to increase by 1.2 percent in the next period before falling down
- 180 -
Assessment of Productive Efficiency of Airports
slightly 1.3 percent during 2002 – 2003. The slow growth and regress may be associated with the September 11 terrorist attacks which shook aviation industry worldwide and still had effects during the analysis period. The netted 1.8% productivity gains between 2000 and 2001 are mainly from efficiency change (airports become +8.8% more efficient) which compensates 7.0% productivity loss from technical change (frontier-shift effect). The situation is opposite in 2001/2002 period when overall 1.2% productivity gains mainly resulted from frontier shift (4.7%). Between 2002 and 2003 both efficiency loss (1.2%) and frontier shift (0.1%) collectively contribute to the overall 1.3% productivity loss.
8.2.3 Impact of delays on airport productivity
It is found out that by ignoring undesirable outputs, i.e., delayed flights and time delays, the results are drastically different in many important aspects. First of all, only a handful of airports (i.e., 6 to 7 depends on the year) are classified as efficient. Exclusively, they are very busy airports that include Hartsfield-Jackson Atlanta (ATL), Los Angeles International (LAX), LaGuardia (LGA), Memphis (MEM), Phoenix Sky Harbor International (PHX), San Diego International (SAN), and John Wayne airport (SNA). All other airports are classified as inefficient with different degrees. Unless traffic is not exceptionally dense, the airport will never be identified as efficient. This is because delays are out of the assessment. Second, the level of inefficiency as reflected through the efficient score is generally much higher. This is proven by several statistical tests. Consequently, airport
- 181 -
Assessment of Productive Efficiency of Airports
performance looks very poorly although it may not be the case from stakeholders’ perception. For instance, the results suggest that Boston Logan International (BOS) handled less than 50% of the level that it should have been able to handle. More precisely, it should have handled about 50 million passengers in 2003 (Table 5.8), rather than just 22.79 million (Table 4.1). The level of inefficiency may be overestimated when delays are not taken into the assessment. Third, the relative ranking of airports may also be distorted. In particular, smaller and less busy airports that are deemed to be inefficient may appear on the efficient frontier when delays are added as undesirable outputs. Fourth, the estimated maximum possible production may not be reasonable and practical. The results indicate potential increases of traffic from current levels at around 133%, 91%, and 364% as compared to around 23%, 20% and 35% when delays are considered. The discrepancy may be interpreted as amount of output loss due to cleaning up delays or keeping them at relatively low levels. It can also represent the tradeoff that an airport has to bear in exchange for higher quality of service. Fifth, the computed productivity indexes are statistically different when delays are accounted for. In many cases, the indexes provided opposite inference regarding the productivity growth. This is actually not surprising since the computation of indexes takes different sets of efficient scores; the resulting indexes are not necessarily similar. The point is that it is crucial to use the right efficiency scores so that the indexes will be meaningful. These are deemed to be the ones with account for undesirable output.
- 182 -
Assessment of Productive Efficiency of Airports
In conclusion, the assessment which does not consider joint production of desirable and undesirable byproduct such as delays will give biased measurements of airport productivity. The interpretation of results can be misleading. Any computation afterwards based on unreasonable efficiency scores including productivity indexes can be confusing. It is strongly recommended to take undesirable outputs into consideration since the results seem to be more reasonable and practical.
8.2.4 Selection of input and output measures
Scenario analyses (i.e., with and without delays, with and without land area) provide insights regarding the effects of chosen measures on the sensitivity of productivity. It is true in general that as the number of input and output measures increase, there will be more airports that are deemed efficient. Note however that the increase in number of airports on the efficient frontier was more dramatic when delays were added than when an additional input (land area) was added. This suggests that consideration of undesirable outputs is at least as important as the consideration of additional input in determining productive efficiency of airports. The failure to include undesirable outputs in the assessment of airport productivity could lead to misleading results. It is concluded here that selection of input, desirable and undesirable outputs should be carefully considered in tandem in order to provide meaningful, yet practical results. Ignoring undesirable outputs (such as delays) could lead to unwise policy choices for managing airports. For example, unless funding agencies or regulators give credit or rewards to airports for keeping delays at low levels, there will be little motivation to
- 183 -
Assessment of Productive Efficiency of Airports
improve quality of service. Instead, airports may prefer to focus on accommodating increasing levels of traffic without considering downside of these services.
8.2.5 Factors affecting productive efficiency of U.S. airports
A Censored Tobit regression model was estimated for explaining the variations in efficiency scores of airport operations. Five groups of exploratory variables were investigated, i.e., airport characteristics, airport services, level of utilization, market characteristics, and service characteristics. It was found that factors such as utilization of land area and runway area contribute to the enhancement of productivity. Meanwhile higher proportion of international passengers and average delay per passenger tend to reduce the productivity. The model also captured anomaly effects in the years 2001, 2002 and 2003 (it was observed that an airport could become efficient slightly more easily with the same utilization rate). Given some planned measures, the model can be used to predict future productivity of an airport and should be very useful as a tool for planning airport business in the U.S. and other geographical regions.
8.3 Suggested future research
This research pioneers the work on the assessment of productivity of airports operating in MASs and when joint production of desirable and undesirable outputs is taken into consideration. It opens up new opportunities for aviation researchers and practitioners to better understand the relation between inputs and outputs of airport operations. There are several potential extensions to this research that could be conducted in the future. Some of them are suggested here.
- 184 -
Assessment of Productive Efficiency of Airports
1) Consideration of comprehensive input and output measures An attempt may be made to collect other input and output measures and take them into consideration for assessing the productivity of U.S. and international airports. The input measures in this dissertation may be rather limited to the airside operation. In fact, one may want to see how other capital inputs such as number of gates, terminal area, and apron area could impact the productivity of airports. Financial inputs are also important for airport operations. Environmental factors (e.g., population density, accessibility, and market condition) also have significant impact on traffic volume which in turn affects productivity of airport operations. On the undesirable outputs, although this dissertation considers perhaps the most conceivable undesirable outputs i.e., delays but there are other undesirable outputs that airport stakeholders are also concerned with such as the number of mishandled baggage and accidents. Even delays could be expanded to encompass a wider number of delay causes. Externality such as noise is perhaps the most frequently-cited undesirable byproduct during the airport planning process. With the current technology, there is no way to get rid of them. However, no study has ever taken them into consideration while assessing productive efficiency of airport operation. It will be very interesting to see how externalities could affect the productivity of airports. Future research may include them into the model. One might argue that unless these inputs and output measures are not accounted for, the performance measures may be misleading. In this line of research extension, a lot of effort and resources are needed to collect the data since there does not seem to be a
- 185 -
Assessment of Productive Efficiency of Airports
consolidated database. As this dissertation showed, different sets of inputs and outputs can lead to very different results. A reasonable question is how to choose a set of input/output measures that yields robust results, yet is meaningful for airport management. 2) Application in the international context It will be very useful if one can extend the study framework to assess productivity of airports in the global context so that the valuable lessons may be learned from truly efficient airports, rather than benchmark among U.S. airports only. However, comparison of airports across nations is not an easy task. There will be several other factors involved that affect the efficiency of airport operations. For example, differences in organizational structure may provide different levels of control to airport managers. The definition and measurement of inputs and outputs are also an issue since different countries may adopt different approaches. Again, data availability will be a major hurdle. 3) Better understanding of factors affecting productive efficiency Many studies have focused on assessing productivity, but relatively few paid attention to the development of prediction models. More research effort may be put forth toward the development of casual models for explaining variation in airport productivity. Such models will enable the managers and policy makers to better understand factors that can enhance operational efficiency. In this area, one may want to investigate effects of other variables beyond those considered in this dissertation. For instance, it is interesting to study the effects of the common ownership of airports in an MAS on their
- 186 -
Assessment of Productive Efficiency of Airports
productivity. One might expect a higher efficiency due to strong coordination, but this has never been studied before. There is a lot more room for research in this direction. 4) Application to other transportation modes It is not an exaggeration to say that transportation activities create undesirable byproducts, regardless of modes. Bus, rail, and water-transport systems all create air pollution. Accidents occur every day on highways. Delays are incurred in all transportation modes. The proposed methodology is certainly well-suited for assessing productivity of other transportation modes. Policy makers may want to know performance of transportation services if these undesirable byproducts are considered. Recently, some researchers (mostly from the economics discipline) started looking at productivity of bus transit by considering joint production of desirable measures and pollutants (e.g., NO2 and CO2) (Noh and McMullen, 2006). Productivity and efficiency of trucking industry accounting for traffic fatalities is studied by Weber and Weber (2004). 5) Theoretical development As for DEA, there is much room for further developing the directional output distance function approach to treat certain cases. For example, it may be adapted to deal with categorical input or output measures such as operating conditions (e.g., snow-belt or not, hub or non-hub, existence of noise abatement program). With this model, it is possible to make an analysis closer to a like comparison which in turn provides fairer and more meaningful results for airport management. Furthermore, the model can be
- 187 -
Assessment of Productive Efficiency of Airports
developed to allow for non-radial expansion and contraction for use in the case that policy makers can reveal preference toward individual input and output measures.
- 188 -
Assessment of Productive Efficiency of Airports
REFERENCES
Abbott M., Wu S. (2002) Total factor productivity and efficiency of Australian airports, The Australian Economic Review, Volume 35, Number 3, pp. 244 – 260. Adler N., Berechman J. (2001) Measuring airport quality from the airlines’ viewpoint: an application of Data Envelopment Analysis. Transport Policy, Volume 8, Issue 3, pp. 171 – 181. Air Transport Research Society (2002) Airport benchmarking report: the global standard for airport excellence, ATRS, Vancouver. Air Transport Research Society (2003) Airport benchmarking report: the global standard for airport excellence, ATRS, Vancouver. Airports Council International (2002) Worldwide Airport Traffic Report – 2001. Airports Council International. (2003) Worldwide Airport Traffic Report – 2002. Airports Council International (2004) Worldwide Airport Traffic Report – 2003. Amemiya T. (1984) Tobit models: A survey, Journal of Econometrics, Volume 24, Number 1 – 2, pp. 3 – 61. Anderson P, Peterson N.C. (1993), A procedure for ranking efficient units in Data Envelopment Analysis, Management Science, Volume 39, Number 10, pp. 1261 – 1264. Ashford N. (1999) Airport finance, Wiley, London.
- 189 -
Assessment of Productive Efficiency of Airports
Banker R.D. (1984) Estimating most productive scale size using data envelopment analysis, European Journal of Operational Research, Volume 17, Issue 1, pp. 35 – 44. Banker R.D., Charnes A., Cooper W.W. (1984) Some models for estimating technical and scale efficiencies in Data Envelopment Analysis, Management Science, Volume 30, Number 9, pp. 1078 – 92. Banker R.D., Charnes A., Cooper W. W., Swarts J., Thomas and D.A. (1989) An introduction to Data Envelopment Analysis with some of its model and their uses, Research in governmental and Nonprofit Accounting, JAI Press, Connecticut 5, pp. 125 – 163. Banker R.D., Morey R.C. (1986) The use of categorical variables in Data Envelopment Analysis, Management Science, Volume 32, Number 12, pp. 1613 – 1627. Banker R.D., Thrall R.M. (1992) Estimation of returns to scale using Data Envelopment Analysis, European Journal of Operational Research, Volume 62, Issue 1, pp. 74 – 84. Bazargan M., Vasigh B. (2003) Size versus efficiency: a case study of US commercial airports, Journal of Air Transport Management, Volume 9, Issue 3, pp. 187 – 193. Berlin Brandenburg International (2005) Schedule,http://www.berlinairport.de/bbi/rubEnglish/rubProjekt/rubZeitplan/index.html, accessed on July 30, 2005.
- 190 -
Assessment of Productive Efficiency of Airports
Boame K.A. (2004) The technical efficiency of Canadian urban transit system, Transportation Research Part E, Volume 40, Issue 5, pp. 401 – 416. Boame K.A., Obeng K. (2005) Sources of productivity change: a Malmquist total factor productivity approach, Transport Reviews, Volume 25, Number 1, pp. 103 – 116. Boussofiane A, Dyson R. G., Thanassoulis E (1991) Applied Data Envelopment Analysis European Journal of Operational Research, Volume 52, Number 1, pp. 1 – 15. Brockett P L., Golony B. (1996) Using rank statistics for determining programmatic efficiency differences in Data Envelopment Analysis. Management Science, Volume 42, Number 3, pp. 466 – 472. Bureau of Transportation Statistics (2004) Transportation Statistics Annual Report. Bureau of Transportation Statistics (2005) Transportation Statistics Annual Report. Bureau of Transportation Statistics (2006) TranStats,http://www.transtats.bts.gov/, Accessed on March 31, 2006. Caves D. W, Christensen L. R., Diewert W. E. (1982) The economic theory of index numbers and the measurement of input, output and productivity, Econometrica 50, pp. 303 – 313. Caves R.E., Gosling G.D. (1999) Strategic Airport Planning, Elsevier Science Ltd., The Netherlands.
- 191 -
Assessment of Productive Efficiency of Airports
Chambers R. G. (1996) A new look at exact input, output, productivity and technical change measurement, Maryland Agricultural Experiment Station, Department of Agricultural and Resource Economics, University of Maryland, College Park. Chambers R. G., Färe R., Grosskopf S. (1996) Productivity growth in APEC countries, Pacific Economic Review, Volume 1, Issue 3, pp. 181 – 190. Charles River Associates Incorporated (2001) Reductions in flight operations as an alternative to runway reconfiguration at San Francisco International Airport, Final Report, CRA No. D02730-00. Charnes A., Cooper W. W., Lewin A. Y., Seiford L. M. (editors) (1994), Data Envelopment Analysis: theory, methodology and applications, Kluwer Acadeic Publishers, USA. Charnes A., Cooper W. W., Rhodes E. (1978) Measuring the efficiency of decision making units, European Journal of Operational Research, Volume 2, Number 6, pp. 429 - 444. Chung Y. H., Färe R., Grosskopf S. (1997) Productivity and undesirable outputs: a directional output distance function approach, Journal of Environmental Management, Volume 51, Issue 3, pp. 229 – 240. Coelli T., Perelman S. (1999) A comparison of parametric and non-parametric distance functions: with application to European railways, European Journal of Operational Research, Volume 117, Issue 2, pp. 326 – 339.
- 192 -
Assessment of Productive Efficiency of Airports
Cook W. D., Zhu J. (2005) Modeling performance measurement: applications and implementation issues in DEA, Springer, USA. Cooper W.W., Seiford L.M., Thanassoulis E., Zanakis S.H. (editors) (2004) DEA and its uses in different countries, European Journal of Operational Research, Volume 154, Issue 2, pp. 337 – 344. Cooper W.W., Seiford L. M., Tone K. (2000) Data Envelopment Analysis a comprehensive text with models, applications, references and DEA-solver software, Kluwer Academic Publishers. Cooper W. W., Seiford L. M., Zhu J. (2004) Data Envelopment Analysis: history, models and interpretations, Handbook on Data Envelopment Analysis, Kluwer Academic Publishers, USA, pp. 1 – 39. de Neufville R. (1995) Management of Multi-Airport Systems A Development Strategy. Journal of Air Transport Management. Volume 2, No. 2, pp. 99 – 110. de Neufville R., Odoni A. (2003) Airport Systems Planning, Design, and Management, McGraw-Hill Inc. Doganis R. (1992) The airport business, Routledge, UK. Doyle J., Green R. (1994) Efficiency and cross-efficiency in DEA: derivations, meanings and uses. Journal of the Operational Research Society, Volume 45, Number 5, pp. 567 – 578.
- 193 -
Assessment of Productive Efficiency of Airports
Emrouznejad A. (2006) DEA homepage,http://www.deazone.com/, accessed June 1, 2006. Färe R., Grosskopf S. (2004a) Modeling undesirable factors in efficiency evaluation: comment, European Journal of Operational Research 157, 242 - 245. Färe R., Grosskopf S. (2004b) New Directions: Efficiency and Productivity, Kluwer Academic Publishers. Färe R., Grosskopf S., Lindgren B., Roos P. (1994a) Productivity Developments in Swedish Hospitals: A Malmquist Output Index Approach, Data Envelopment Analysis theory, methodology and applications, Charnes A., Cooper W., Lewin A. Y., Seiford L. M. (editors), Kluwer Academic Publishers, 253 - 272. Färe R., Grosskopf S., Norris M., Zhang Z. (1994b) Productivity growth, technical progress and efficiency change in industrialized countries. The American Economic Review, Volume 84, Number 1, pp. 66 – 83. Färe R., Grosskopf S., Roos P. (1998) Malmquist productivity indexes: a survey of theory and practice, Index numbers: essays in honour of Sten Malmquist, Färe R., Grosskopf S., Russell R. R. (editors), Kluwer Academic Publishers, 127 - 190. Färe R., Grosskopf S., Whittaker G. (2004) Distance Functions with Applications to DEA, Handbook on Data Envelopment Analysis Directions: Efficiency and Productivity, Cooper W.W., Seiford L.M, Zhu J. (editors), Kluwer Academic Publishers, 139 - 152.
- 194 -
Assessment of Productive Efficiency of Airports
Färe R., Primond D. (1995) Multi-output production and duality: theory and applications, Kluwer Academic Publishers, Boston. Farrell M.J. (1957) The measurement of productive efficiency, Journal of the Royal Statistical Society, Series A, Volume 120, pp. 253 - 351. Federal Aviation Administration (2004a), the Terminal Area Forecast System,http://www.apo.data.faa.gov/faatafall.HTM, Accessed March 1, 2004. Federal Aviation Administration (2004b) Airport Data (5010),http://www.faa.gov/data_statistics/aviation_data_statistics/, Accessed March 1, 2004. Federal Aviation Administration (2005) FAA Operations & Performance Data: Operational Network (OPSNET),http://www.apo.data.faa.gov/, Accessed on August 29, 2005. Federal Aviation Administration (2006) FAA Operations & Performance Data: Airline Service Quality Performance (ASQP),http://www.apo.data.faa.gov/, Accessed on March 31, 2006. Fernandes E., Pacheco R. R (2001) Total airport productivity in Brazil, a paper presented at XIIth World Productivity Congress, Hong Kong, November 2001. Fernandes E., Pacheco R.R. (2002) Efficient use of airport capacity, Transportation Research Part A, Volume 36, Issue 3, pp. 225 – 238.
- 195 -
Assessment of Productive Efficiency of Airports
Fernandes E., Pacheco R.R. (2005) Impact of changes in management style on the performance of Brazilian airports, Proceedings of the Air Transport Research Society (ATRS) conference, Rio de Janeiro, Brazil, July 3 – 7, 2005. Francis G., Humphreys I. (2001) Airport regulation: reflecting on the lessons from BAA plc, Public Money & Management, January – March, pp. 49 – 52. Francis G., Humphreys I., Fry J. (2002) The benchmarking of airport performance, Journal of Air Transport Management 8, pp 239 – 247. Ganley J. A., Cubbin J. S. (1992) Public sector efficiency measurement: applications of Data Envelopment Analysis, Elsevier. Gillen D., Lall A. (1997) Developing measures of airport productivity and performance: an application of Data Envelopment Analysis. Transportation Research Part E, Volume 33, Number 4, 261 – 273. Gillen D., Lall A (1998) Non-parametric measures of efficiency of U.S airports, Proceedings of the Air Transport Research Society (ATRS) conference, Dublin, Ireland, July 20 - 21, 1998. Graham A. (2003) Managing airports: an international perspective, 2nd edition, Elsevier, UK. Greene W.H. (2002) LIMDEP Version 8.0 Econometric Modeling Guide, Volume 2, Econometric Software Inc., USA. Greene W.H. (2003) Econometric analysis, 5th Edition, Prentice Hall, USA.
- 196 -
Assessment of Productive Efficiency of Airports
Hansen M., Weidner T. (1995) Multiple Airport Systems in the United States: Current Status and Future Prospects, Transportation Research Record, Number 1506, pp. 8 – 17. Haynes K.E., Stough R.R., Shroff H.F.E. (1990), New methodology in context: Data Envelopment Analysis, Computers, Environment and Urban Systems, Volume 14, Number 2, pp. 85 – 175. Hooper P.G., Hensher D.A. (1997) Measuring total factor productivity of airports – an index number approach, Transportation Research Part E, Volume 33, Number 4, 249 – 259. Humphreys I. (1999) Privatisation and comercialisation: changes in UL airport ownership patterns, Journal of Transport Geography 7 (2), pp. 121 – 134. Humphreys I., Francis G. (2002) Performance measurement: a review of airports, International Journal of Transport Management 1, pp. 79 – 85. Itoh H. (2002) Efficiency changes at major container ports in Japan: a window application of Data Envelopment Analysis, The Applied Regional Science Conference (ARSC), Volume 14, Number 2, pp. 133 – 152. Jarach D. (2005) Airport marketing: strategies to cope with the new millennium environment, Ashgate Publishing Limited. Kerstens K. (1996) Technical efficiency measurement and explanation of French urban transit companies, Transportation Research Part A, Volume 30, Number 6, pp. 431 – 452.
- 197 -
Assessment of Productive Efficiency of Airports
London City Airport Consultative Committee (2005) Constructing the Airport,http://www.lcacc.org/history/construction.html, accessed on July 31, 2005. Maddala G.S. (1983) Limited-dependent and Qualitative Variables in Econometrics, Cambridge University Press, USA. Malmquist S. (1953) Index numbers and indifference surfaces, Trabajos de Estatistica, Volume 4, pp. 209 – 242. Martin J. C., Roman C. (2001) An application of DEA to measure the efficiency of Spanish airports prior to privatization, Journal of Air Transport Management, Volume 7, Issue 3, 149 – 157. Martin-Cejas R.R. (2002) An approximation to the productive efficiency of the Spanish airports network through a deterministic cost frontier, Journal of Air Transport Management, Volume 8, Issue 4, pp. 233 – 238. Martinez-Budria E., Diaz-Armas R., Navarro-Ibanez M., Ravelo-Mesa T. (1999) A study of the efficiency of Spanish port authorities using Data Envelopment Analysis, International Journal of Transport Economics, Volume XXVI, Number 2, pp. 237 – 253. Murillo-Melchor C. (1999) An analysis of technical efficiency and productivity changes in Spanish airports using the Malmquist index, International Journal of Transprot Economics, Volume XXVI, Number 2, pp. 271 – 292.
- 198 -
Assessment of Productive Efficiency of Airports
Noh D.W., McCullen B. S. (2006), Accounting for environmental standards in the measurement of bus transit efficiency: a directional output distance function approach, Manuscript, submitted. Norlander N. O., Roos P. (1998) Implementing productivity index: the case of the national corporation of Swedish pharmacies, Index numbers: essays in honour of Sten Malmquist, Färe R., Grosskopf S., Russell R. R. (editors), Kluwer Academic Publishers, 191 – 216. Nyshadham E. A., Rao V. K. (2000) Assessing efficiency of European airports: a total factor productivity approach, Public Works Management & Policy, Volume 5, Number 2, pp 106 – 114. Oum T.H., Yu C. (2004) Measuring airports’ operating efficiency: a summary of the 2003 ATRS global airport benchmarking report. Transportation Research Part E 40, 515 – 532. Oum T. H., Yu C., Fu X. (2003) A comparative analysis of productivity performance of the world’s major airports: summary report of the ATRS global airport benchmarking research report – 2002. Journal of Air Transport Management, Volume 9, Issue 5, 285 – 297. Pacheco R.R., Fernandes E. (2003) Managing efficiency of Brazilian airports, Transportation Research Part A, Volume 37, Issue 8, 667 – 680. Parker D. (1999) The performance of BAA before and after privatization, Journal of Transport Economics and Policy, Volume 33, Part 2, pp. 133 – 146.
- 199 -
Assessment of Productive Efficiency of Airports
Pathomsiri S., Haghani A. (2004) Benchmarking efficiency of airports in the competitive multiple-airport systems: the international perspective, Proceedings of the 19th Transport Chicago Conference (CD-ROM), Chicago, IL. Pathomsiri S., Haghani A. (2005) Taste Variations in Airport Choice Models, Journal of Transportation Research Record 1915, Transportation Research Board, National Research Council, Washington DC, pp. 27 – 35. Pathomsiri S., Haghani A., Schonfeld P. S. (2005) Operational efficiency of airports in multiple-airport systems, Transportation Research Board 84th Annual Meeting Compendium of Papers (CD-ROM), Washington DC. Pathomsiri S., Haghani A., Dresner M., Windle R. J. (2006a), Measurement and determination of airport productivity in competitive markets, Transportation Research Board 85th Annual Meeting Compendium of Papers (CD-ROM), Washington DC. Pathomsiri S., Haghani A., Dresner M., Windle R. J. (2006b) Productive and Operational Efficiency of US Airports with Joint Consideration of both Desirable and Undesirable Outputs, the 47th Annual Transportation Research Forum Proceedings (CD-ROM), New York, March 23 – 25, 2006. Pathomsiri S., Haghani A., Dresner M., Windle R. J. (2006c) Assessment of Airport Performance with Joint Consideration of Desirable and Undesirable Outputs, the 10th Annual World Conference Air Transport Research Society (ATRS) Proceedings (CD-ROM), Nagoya, Japan, May 26 – 28, 2006.
- 200 -
Assessment of Productive Efficiency of Airports
Pathomsiri S., Haghani A., Windle R.J., Dresner M. (2006) Role of Freight Transportation Services and Reduction of Delayed Flights on Productivity of US Airports, National Urban Freight Conference 2006 Proceedings (CD-ROM), Long Beach, CA, February 1 – 3, 2006. Pathomsiri S., Mahmassani H.S., Haghani A. (2004) Airport Choice Models for Baltimore-Washington Region, Transportation Research Board 83rd Annual Meeting Compendium of Papers (CD-ROM), Washington DC, January 2004. Pels, E., Nijkamp P., Rietveld P. (2001) Relative efficiency of European airports, Transport Policy, Volume 8, 183 – 192. Pels E., Nijkamp P., Rietveld P. (2003) Inefficiencies and scale economies of European airport operations, Transportation Research Part E, Volume 39, Issue 5, 341 – 361. Pina V., Torres L. (2001) Analysis of the efficiency of local government services delivery: an application to urban public transport, Transportation Research Part A, Volume 35, Issue 10, pp. 929 – 944. Pires Capobianco H.M., Fernandes E. (2004) Capital structure in the world airline industry, Transportation Research Part A, Volume 38, Issue 6, pp. 421 – 434. QMS (2005) Quantitative Micro Software,http://www.eviews.com, accessed on July 30, 2005. Ray S. C. (2004) Data Envelopment Analysis theory and techniques for economics and operations research, Cambridge University Press.
- 201 -
Assessment of Productive Efficiency of Airports
Ross A.D., Droge C. (2004) An analysis of operations efficiency in large-scale distribution systems, Journal of Operations Management, Volume 21, Issue 6, pp. 673 – 688. Rousseau J. J, Semple J. H. (1995) Radii of classification preservation in Data Envelopment Analysis: a case study of ‘program follow-through’. Journal of the Operational Research Society, Volume 46, Number 8, pp 943 – 957. Salazar De La Cruz F. (1999) A DEA approach to the airport productivity function, International Journal of Transport Economics, Volume XXVI, Number 2, pp. 255 – 270. Sarkis J. (2000) An analysis of the operational efficiency of major airports in the United States. Journal of Operations Management, Volume 18, Issue 3, 335 – 351. Sarkis J., Talluri S. (2004) Performance based clustering for benchmarking of U.S. airports. Transportation Research Part A, Volume 38, Issue 5, 329 -346. Schefczyk M. (1993) Operational performance of airlines: an extension of traditional measurement paradigms, Strategic Management Journal, Volume 14, pp. 301 – 317. Scheraga C.A. (2004) Operational efficiency versus financial mobility in the global airline industry: a data envelopment and Tobit analysis, Transportation Research Part A, Volume 38, Issue 5, pp. 383 - 404. Seiford L.M. (1996) Data Envelopment Analysis: the evolution of the state of the art (1978 – 1995), Journal of Productivity Analysis, Volume 7, pp. 99 – 137.
- 202 -
Assessment of Productive Efficiency of Airports
Seiford L.M., Zhu J. (2002) Modeling undesirable factors in efficiency evaluation, European Journal of Operational Research 142, 16 – 20. Seiford L.M., Zhu J. (2005) A response to comments on modeling undesirable factors in efficiency evaluation, European Journal of Operational Research 161, 579 – 581. Shephard R. W. (1970) Theory of cost and production functions, Princeton University Press, New Jersey. StataCorp (2005), Stata Statistical Software for Professionals,http://www.stata.com, accessed on July 30, 2005. Stewart International Airport (2005) Stewart History,http://www.stewartintlairport.com/home.html, accessed on Jul 31, 2005. Tobin J. (1958) Estimation of Relationships for Limited Dependent Variables, Econometrica, Volume 26, Number 1, pp. 24 – 36. Tretheway M. (1995) Book review (The Economic Performance of European Airports R.S. Doganis, A. Graham and A. Lobbenberg), Journal of Air Transport Management, Volume 2, Issues 3-4, p. 207. Tone K, Malmquist productivity index: efficiency changes over time, Handbook on Data Envelopment Analysis directions: efficiency and productivity, Cooper W.W., Seiford L.M, Zhu J. (editors), Kluwer Academic Publishers, 203 – 228. Tongzon J.L. (1995) Determinants of port performance and efficiency, Transportation Research Part A, Volume 29, Number 3, pp. 245 – 252.
- 203 -
Assessment of Productive Efficiency of Airports
Tongzon J. (2001) Efficiency measurement of selected Australian and other international ports using data envelopment analysis, Transportation Research Part A, Volume 35, Issue 2, pp. 107 – 122. Turner H., Windle R., Dresner M. (2004) North American containerport productivity: 1984 – 1997, Transportation Research Part E, Volume 40, Issue 4, pp. 339 – 356. Weber M.M, Weber W.L. (2004) Productivity and efficiency in the trucking industry accounting for traffic fatalities, International Journal of Physical Distribution & Logistics Management, Volume 34, Number 1, pp. 39 – 61. Windle R.J., Dresner M.E. (1992) Partial productivity measures and total factor productivity in the air transport industry: limitations and uses, Transportation Research Part A, Volume 26, Number 6, pp. 435 – 445. Yoshida Y. (2004) Endogenous-weight TFP measurement: methodology and its application to Japanese-airport benchmarking, Transportation Research Part E 40, 151 – 182. Yoshida Y., Fujimoto H. (2004) Japanese-airport benchmarking with the DEA and endogenous- weight TFP methods: testing the criticism of overinvestment in Japanese regional airports, Transportation Research Part E 40, 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, 295 – 303.
- 204 -
Assessment of Productive Efficiency of Airports
Zaim O. (2005) A framework for incorporating environmental indicators to the measurement of human well-being, Research paper No. 2005/05, World Institute for Development Economics Research. Zhu J. (2003) Quantitative models for performance evaluation and benchmarking Data Envelopment Analysis with spreadsheets and DEA Excel Solver, Kluwer Academic Publishers.
- 205 -
Assessment of Productive Efficiency of Airports
Bibliography
Mr. Somchai Pathomsiri was born and raised in Bangkok, the capital city of the Kingdom of Thailand. He earned three engineering degrees, i.e., B.Eng. (Honor) in Civil Engineering from King Mongkut Institute of Technology Thonburi (now King Mongkut's University of Technology Thonburi); M.Eng. in Traffic and Transportation Systems Engineering from the Department of Civil Engineering, Chulalongkorn University, Thailand; and PhD in Transportation Systems Engineering and Planning from the Department of Civil and Environmental Engineering, University of Maryland, USA. All three degrees were fully supported by merit-based scholarship. In addition, he also received an MBA (Executive program) from Thammasat University, Thailand. He began his career in the industry working in various roles including civil engineer, traffic engineer, transport engineer, project engineer, project manager and engineering manager. He had worked extensively in construction and transportation planning projects for both private and public sectors. His experience in transportation spans into all modes (i.e., land, water and air). For example, he worked on several feasibility studies and environmental impact assessment for highway, expressway, port and airports projects. He has been a registered professional engineer (P.E.) in Thailand since 1994. He had been a faculty member at the Department of Civil Engineering, Faculty of Engineering Mahidol University since 1997. He is taking care construction management and transportation program. His research interests include Logistics, transportation and
- 206 -
Assessment of Productive Efficiency of Airports
traffic systems analysis, travel behavior & demand modeling, transportation planning, GIS, aviation systems planning, applications of information technology in traffic management, transportation safety and security, transportation economics and infrastructure management. Presently, he lives happily with his amazing wife, Laddawan and three wonderful children, Nawanont (FeiFei), Nontthida (FinFin) and Phuwanont (FarnFarn) in Bangkok, Thailand.
- 207 -
doc_223500467.pdf
An airport is a location where aircraft such as fixed-wing aircraft, helicopters, and blimps take off and land. Aircraft may be stored or maintained at an airport. An airport consists of at least one surface such as a runway for a plane to take off and land, a helipad, or water for takeoffs and landings, and often includes buildings such as control towers, hangars and terminal buildings
ABSTRACT
Title: ASSESSMENT OF PRODUCTIVE EFFICIENCY OF AIRPORTS Somchai Pathomsiri Doctoral of Philosophy, 2006 Directed By: Professor Ali Haghani Department of Civil and Environmental Engineering
The move towards commercialization and privatization has pressured airports to become more productive and competitive. The need to devise an overall (total) productivity measure is increasingly important in airport business. The dissertation made three major research contributions. First, it assessed the productivity of airports operating in multiple airport systems (MASs). Second, it developed a more complete total factor productivity measure by considering joint production of desirable and undesirable outputs. Third, it developed models for explaining variations in productive efficiency. These are accomplished in two case studies. In case study 1, the Data Envelopment Analysis (DEA) is used to assess relative total productivity of 72 airports operating in 25 MASs during 2000 – 2002. The results indicate that highly utilized airports such as O’Hare International, Los Angeles International, Heathrow/London and LaGuardia are classified as efficient. The Censored Tobit regression model suggests that runway utilization market dominance, proportion of international passengers and ownership can be used to explain variations in productive efficiency.
In case study 2, the directional output distance function is applied to assess the productivity of 56 U.S. commercial airports during 2000 – 2003. Delays are considered as undesirable outputs. There are several important findings and insightful implications. First, about half of U.S. airports are actually operated efficiently. These airports include busy airports such as Hartsfield-Jackson Atlanta, LaGuardia, and Memphis together with less busy airports with relatively low delays such as Baltimore/Washington International and Oakland International. Second, the overall system has potential to accommodate about 1,550 million passengers, 26 million movements and 34 million tons of cargo. Third, during 2000 - 2003, annual growth of productivity is modest in the range of -1.3% to +1.8%. Fourth, by ignoring delays the assessment provides drastically different results in terms of number of efficient airports, level of inefficiency, ranking, and estimated maximum possible outputs. Fifth, the consideration of undesirable output is as important as the consideration of additional inputs and desirable outputs. The Censored Tobit regression model suggests that runway utilization, proportion of international passengers and average delay per passenger can be used to explain variations in productive efficiency.
ASSESSMENT OF PRODUCTIVE EFFICIENCY OF AIRPORTS
By Somchai Pathomsiri
Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park in partial fulfillment of the requirements for the degree of Doctoral of Philosophy 2006
Advisory Committee: Professor Ali Haghani, Chair Assistant Professor Kelly J. Clifton Professor Martin E. Dresner Professor Hani S. Mahmassani Professor Paul M. Schonfeld Professor Robert J. Windle
© Copyright by Somchai Pathomsiri 2006
Assessment of Productive Efficiency of Airports
Preface
As a transportation professional, I am rather fortunate to be able to work in all modes of transportation, i.e., land, water and air. This dissertation adds another chapter in air transportation experience to my career. I was interested in aviation since I was an undergraduate student studying pavement design for airfield. But it was not until 1990s that I realized the fascination of aviation systems planning when I worked as a transport engineer in two projects, i.e., 1) Airport Systems Master Plan Study in Thailand and 2) Feasibility Study and Master Plan Development for Joint Military-Civilian Used UTaphao International Airport. It was so memorable time to work with several aviation professionals including an old-hand project manager and a good friend of my family, Mr. Clifford R. King (then with Louis Berger International, Inc., currently a senior project manager at Bechtel Corporation). Since then, I have had aviation in my heart. This dissertation was accidentally started while I was completing a term paper on airport choice modeling in Baltimore-Washington multiple airport system for a class ENCE688Y (Advances in Transportation Demand Analysis) in the Fall 2002 taught by Professor Hani Mahmassani. The literature review led me to learn further that aviation community is interested in measuring performance of airport for a variety of reasons, including benchmarking and investment appraisal. I was so surprised to know that research to assess overall airport performance had just started in late 1990s. I then started working on the topic seriously and published our first paper “Benchmarking efficiency of airports in the competitive multiple-airport systems: the international perspective” at the 19th Annual Transport Chicago Conference in June 2004 (with Professor Ali Haghani). Subsequently, Professors Paul Schonfeld, Martin Dresner and Robert Windle kindly ii
Assessment of Productive Efficiency of Airports
accepted the invitation to jointly work and improve the quality of papers. We co-authored another two papers for the 84th and 85th Transportation Research Board Annual Meeting, Washington DC, 2005 - 2006. I am particularly grateful to many valuable comments from my co-authors and several anonymous reviewers. Their comments enabled us to extend and expand the scope of our research, essentially to answer practical aviation issues. One of the frequent comments is about my overemphasis on quantity of outputs and ignoring their quality, although measures such as delays are a major concern of the airport management. Such comments bring the dissertation to the last phase, i.e., the assessment of airport productivity with joint consideration of desirable and undesirable outputs. I stumbled upon a relatively new theory in production economics, i.e., the directional output distance function, while I was searching for a method to deal with undesirable outputs in the productivity assessment. I received useful guidance from Professors Rolf Färe and Shawna Grosskopf of Oregon State University who devised the theory and eminently populated applications in recent years. We could start a new research on airport productivity by jointly considering delays as major output measures, though undesirable, along with other traditional desirable outputs (e.g., number of passengers, aircraft movement, and freight throughput). We have published the new findings in three papers so far, i.e., National Urban Freight Conference, Long Beach, CA (February 2006), the 47th Annual Transportation Research Forum, New York (March 2006) and the 10th Annual Air Transport Research Society (ATRS) World Conference, Nagoya, Japan (May 2006). This dissertation is partly based
iii
Assessment of Productive Efficiency of Airports
on the above-mentioned six publications and another paper which is under review by Transportation Research Part E: Logistics and Transportation Review.
iv
Assessment of Productive Efficiency of Airports
Dedication
To my mother, U-SA Kow whose vision is always clear and correct. You enabled so many things in my life and enlighten me in many ways. I can be successful today because I have you, mom. I really can’t thank you enough.
v
Assessment of Productive Efficiency of Airports
Acknowledgements
Always for this level of accomplishment, there are many people directly or indirectly involved. I apologize if I unintentionally omit one of you. First of all, I wish to express my gratitude to my academic advisor, Professor Ali Haghani for continuous support throughout my doctoral study. Working under his supervision has been particularly rewarding and most gratifying, especially since I have been able to carry out research in various fields (e.g., logistics, travel behavior and demand modeling, GIS, operations research, and aviation system planning) with his full support and confidence. I also want to thank Professors Kelly Clifton, Martin Dresner, Hani Mahmassani, Paul Schonfeld, and Robert Windle for serving in my dissertation committee. Their comments and suggestions were of particularly great value to the quality of the dissertation. In particular, I highly appreciate friendly advice from Professors Paul Schonfeld, Robert Windle and Martin Dresner while we were co-authoring several papers. I learned endless lessons from them and really enjoyed working with them. Hopefully, we could continue the productive collaboration in the future endeavors. Special thanks also go to Professors Rolf Färe and Shawna Grosskopf of Oregon State University for their guidance on the theory and rich applications of the directional output distance function. I greatly benefited from a string of communications with them. In addition, comments and discussion with Professors Steven Burks of the University of Minnesota and B. Starr McMullen of the Oregon State University greatly improve the quality of the dissertation. There are also many airport staff members and managers in the U.S. and around the world who willingly shared their airport information and had fruitful discussions about airport business. I highly appreciate their contribution.
vi
Assessment of Productive Efficiency of Airports
Finally, I want to share my success and pleasure with my wonderful family members who have stood beside me with love and encouragement for years. I am forever indebted to their immeasurable, constant and continuous and tireless support during my study, especially, my wife, Laddawan whose sacrifice has made nothing impossible. I am so proud to have you and our lovely three children, Nawanont, Nontthida and Phuwanont in my life. I also want to express the gratitude to my wife’s family members who always strongly encourage and support us. It is the highest honor of my life to have such a great family support.
vii
Assessment of Productive Efficiency of Airports
TABLE OF CONTENTS
Page Preface ……………………………………………………………………………………ii Dedication ………………………………………………………………………………. v Acknowledgements ……………………………………………………………………...vi TABLE OF CONTENTS ……………………………………………………………...viii LIST OF TABLES ……………………………………………………………………...xii LIST OF FIGURES …………………………………………………………………... xv CHAPTER 1 1.1 1.2 1.3 1.4 1.5 1.6 INTRODUCTION ……………………………………………………. 1
Airport business ………………………………………………………………….. 1 Importance of airport productivity study ………………………………………… 2 Motivation of the dissertation research …………………………………………... 5 Research objectives and scope …………………………………………………... 5 Research contributions …………………………………………………………… 6 Organization of the dissertation ………………………………………………….. 6 LITERATURE REVIEW ……………………………………………. 9
CHAPTER 2 2.1
Productivity Measures …………………………………………………………… 9 2.1.1 2.1.2 Partial Factor Productivity (PFP) measure ………………………………. 9 Total Factor Productivity (TFP) measure ………………………………. 12
2.2
Methodology for computing TFP measure ……………………………………... 13 2.2.1 2.2.2 Parametric approach ……………………………………………………..13 Non-parametric approach ………………………………………………..17
viii
Assessment of Productive Efficiency of Airports
TABLE OF CONTENTS (Continue)
Page CHAPTER 2 2.3 LITERATURE REVIEW
Data Envelopment Analysis (DEA) …………………………………………….. 35 2.3.1 2.3.2 Background ……………………………………………………………... 35 Model development …………………………………………………….. 37
2.4
Discussion ………………………………………………………………………. 43 RESEARCH METHODOLOGY …………………………………... 48
CHAPTER 3 3.1 3.2 3.3 3.4
Characterization of production possibility set ……………………………………. 48 Output distance function ………………………………………………………….. 49 Directional output distance function ……………………………………………… 51 Malmquist productivity index with the presence of undesirable outputs ………… 58 CASE STUDY 1
CHAPTER 4
PRODUCTIVITY OF AIRPORTS IN MULTIPLE AIRPORT SYSTEMS ……… 68 4.1 4.2 4.3 4.4 Definition of multiple airport system ……………………………………………...69 Modeling airport operations ……………………………………………………….74 Input and output measures of airport operations …………………………………..74 Data collection ……………………………………………………………………. 76 CASE STUDY 1: RESULTS AND DISCUSSION………………… 81
CHAPTER 5 5.1 5.2 5.3 5.4
Selection of a DEA model ………………………………………………………... 81 Efficiency scores ………………………………………………………………….. 82 Determination of airport productivity …………………………………………….. 86 Factors affecting productive efficiency of airports in MASs ……………………...88
ix
Assessment of Productive Efficiency of Airports
TABLE OF CONTENTS (Continue)
Page CHAPTER 6 CASE STUDY 2
PRODUCTIVITY OF U.S. AIRPORTS ……………………………………………... 98 6.1 6.2 Modeling airport operations ……………………………………………………….98 Input and output measures of airport operations …………………………………..99
6.3 Sample characteristics …………………………………………………………… 102 6.3.1 Size of sample ……………………………………………………………... 102 6.3.2 Analysis period ……………………………………………………………. 104 6.3.3 Data source and definition ………………………………………………… 105 CHAPTER 7 7.1 CASE STUDY 2: RESULTS AND DISCUSSION ………………. 115
Impact of the inclusion of undesirable outputs …………………………………...115 7.1.1 Classification of efficient airports ……………………………………….. 117 7.1.2 The number of efficient airports ………………………………………… 120
7.1.3 Difference in efficiency scores …………………………………………...121 7.1.4 7.1.5 Ranking …………………………………………………………………..121 Maximum possible production outputs ………………………………….. 126
7.2 Lumpiness of airport investment ………………………………………………... 136 7.3 Changes in productivity over time ………………………………………………. 137
7.4 Scenario analysis ………………………………………………………………… 148 7.5 Determination of airport productivity …………………………………………… 167
7.6 Factors affecting productive efficiency of U.S. airports ………………………… 169
x
Assessment of Productive Efficiency of Airports
TABLE OF CONTENTS (Continue)
Page CHAPTER 8 8.1 CONCLUSIONS AND FUTURE RESEARCH ………………….. 175
Conclusions from assessing productivity of airports in MASs …………………..175 8.1.1 Productivity of airports in MASs ………………………………………... 175 8.1.2 Underutilized airports …………………………………………………….176
8.1.3 Factors affecting productive efficiency of airports in MASs …………….177 8.2 Conclusions from assessing productivity of U.S. airports ………………………. 177 8.2.1 Productivity of U.S. commercial airports ……………………………….. 179 8.2.2 Productivity growth of U.S. commercial airports ……………………….. 180 8.2.3 8.2.4 Impact of delays on airport productivity ………………………………… 181 Selection of input and output measures …………………………………. 183
8.2.5 Factors affecting productive efficiency of U.S. airports ………………… 183 8.3 Suggested future research ……………………………………………………….. 184
REFERENCES ………………………………………………………………………..188 Bibliography ………………………………………………………………………….. 206
xi
Assessment of Productive Efficiency of Airports
LIST OF TABLES Page Table 2.1 Examples of partial factor productivity measures in aviation sector ……….. 11 Table 2.2 List of publications on airport productivity studies by parametric Approach ……………………………………………………………………. 16 Table 2.3 List of publications on airport productivity studies using index number method ……………………………………………………………... 18 Table 2.4 List of publications on airport productivity studies using DEA ……………. 23 Table 2.5 Summary of DEA models …………………………………………………... 44 Table 4.1 List of multiple airport systems and corresponding Herfindahl Concentration Indices, 2002 ………………………………………………... 72 Table 4.2 Descriptive statistics of 72 airports in MASs, 2000 – 2002 ………………... 80 Table 5.1 Efficiency scores, 2000 – 2002 …………………………………………….. 84 Table 5.2 Comparisons of statistics between efficient and inefficient airports ……….. 92 Table 5.3 Censored Tobit regression: preliminary model estimation results …………. 93 Table 5.4 Censored Tobit regression: proposed model estimation results ……………. 94 Table 6.1 List of 56 US airports under consideration and their outputs in 2003 ……...109 Table 6.2 Descriptive statistics of samples 2000 – 2003 …………………………….. 114 Table 7.1 Efficiency scores for Case 1 and Case 2 …………………………………... 118 Table 7.2 Comparisons of efficiency scores between Cases 1 and 2 by paired sample t-test ………………………………………………………... 122 Table 7.3 Comparisons of efficiency scores between Cases 1 and 2 by Nonparametric Paired tests ………………………………………………... 123 Table 7.4 Ranking of airport productivity …………………………………………… 124
xii
Assessment of Productive Efficiency of Airports
LIST OF TABLES (Continue) Page Table 7.5 Maximum possible passengers, aircraft movements and cargo throughput in 2000 ………………………………………………………… 128 Table 7.6 Maximum possible passengers, aircraft movements and cargo throughput in 2001 ………………………………………………………… 130 Table 7.7 Maximum possible passengers, aircraft movements and cargo throughput in 2002 ………………………………………………………… 132 Table 7.8 Maximum possible passengers, aircraft movements and cargo throughput in 2003 ………………………………………………………… 134 Table 7.9 Luenberger productivity indexes, Case 1 …………………………………..138 Table 7.10 Luenberger productivity indexes, Case 2 ………………………………...143 Table 7.11 Comparisons of Luenberger productivity indexes between Cases 1 and 2 ……………………………………………………………. 145 Table 7.12 Comparisons of Luenberger productivity indexes by paired sample t-test ……………………………………………………………... 147 Table 7.13 Comparisons of Luenberger productivity indexes by nonparametric tests ……………………………………………………… 147 Table 7.14 Efficiency scores for Cases 1 and 3 ……………………………………... 151 Table 7.15 Comparisons of efficiency scores between Cases 1 and 3 by paired sample t-test ……………………………………………………… 153 Table 7.16 Comparisons of efficiency scores between Cases 1 and 3 by nonparametric paired tests ………………………………………………. 154
xiii
Assessment of Productive Efficiency of Airports
LIST OF TABLES (Continue) Page Table 7.17 Maximum possible passengers, aircraft movements and cargo throughput in 2000, Cases 1 and 3 ………………………………………. 156 Table 7.18 Maximum possible passengers, aircraft movements and cargo throughput in 2001, Cases 1 and 3 ………………………………………. 158 Table 7.19 Maximum possible passengers, aircraft movements and cargo throughput in 2002, Cases 1 and 3 ………………………………………. 160 Table 7.20 Maximum possible passengers, aircraft movements and cargo throughput in 2003, Cases 1 and 3 ………………………………………. 162 Table 7.21 Luenberger productivity indexes, Cases 1 and 3 …………………………164 Table 7.22 Comparisons of Luenberger productivity indexes by paired sample t-test Cases 1and 3 ………………………………………………. 166 Table 7.23 Comparisons of Luenberger productivity indexes by nonparametric paired tests ………………………………………………. 167 Table 7.24 Censored Tobit regression model estimation results ……………………. 172
xiv
Assessment of Productive Efficiency of Airports
LIST OF FIGURES Page
Figure 2.1 Difference between DEA and regression concept ………………………... 37 Figure 2.2 Determination of efficient production frontier …………………………… 39 Figure 3.1 Output possibility set and distance functions …………………………….. 50 Figure 3.2 Graphical illustration of directional output distance function concept …… 55 Figure 3.3 The Luenberger productivity indicator …………………………………… 64 Figure 6.1 Locations of 56 airports …………………………………………………. 104 Figure 6.2 Scatter plot between number of delayed flights and number of passengers, 2003 ………………………………………………………… 112 Figure 6.3 Scatter plot between delay/passenger and density of movements, 2000 – 2003 ………………………………………………………………112
xv
Assessment of Productive Efficiency of Airports
CHAPTER 1 INTRODUCTION
1.1 Airport business The aviation industry has experienced swift ups and downs after the deregulation of air transportation in the late 1970s. The competition in the air transportation industry is now fierce. In recent years, it has become more common for airports to advertise and promote their services to lure customers just like other businesses do. Airport business models have changed dramatically from being perceived as a fundamental public service in the same way as roads and public transport, to a commercial activity. Over the past twenty years, it has become obvious that airports can actually be run as highly successful and profitable businesses (Doganis, 1992). On one hand, airports are an essential part of the air transportation system. They provide all the infrastructure needed to enable passengers and freight to transfer from surface to air mode of transport and to allow airlines to take off and land. On the other hand, airports also offer a wide variety of commercial facilities ranging from shops, restaurants, hotels, and conference services (Graham, 2003; Jarach, 2005). There is no doubt that an airport can be a big business. The public floatation of the British Airports Authority in the summer of 1987 is valued at £1.3 billion (Doganis, 1992). It is expected that rapid growth of air traffic would require enormous amount of funding to support airport improvement programs. This places increasing pressures on public finance. Such pressures have led governments all over the world to consider privatization and commercialization to relieve them from the financial burden of airport ownership -1-
Assessment of Productive Efficiency of Airports
(Ashford, 1999; Francis and Humphreys, 2001). Airport managers have to adapt themselves in response to such pressures as well. They are now acting more like corporate business managers. They think strategically, identify markets, set objectives and goals, cast competitive strategies, implement, monitor, evaluate outcomes, and respond to the dynamics of market competition. Their job is much more complicated than before. 1.2 Importance of airport productivity study To study airport productivity is to study the relationship between inputs and outputs of airport operation. With such a relation, airport managers can easily determine the probable traffic level that airports should accommodate, given any level of inputs. This is very useful in monitoring, managing and planning airports. In addition, it allows managers to benchmark their operational performance with peers and set appropriate output targets for improving their business. Until the 1980s, the systematic monitoring and comparing of airport performance was not a widely practiced activity within the airport industry. This can largely be attributed to insufficient commercial and business pressures for airports and the general lack of experience of benchmarking techniques within the public sector. With airport privatization and commercialization has come a marked interest in performance comparisons and benchmarking. As airports become more commercially-oriented, they have been keen to identify the strong performers in the industry and adopt what are seen as best practices (Graham, 2003). Hooper and Hensher (1997) commend that the growing importance of airport performance measurement is accompanied by the trend toward corporatized or even privatized airports. In recent survey from the world’s top 200 busiest
-2-
Assessment of Productive Efficiency of Airports
passenger airports (Francis, Humphreys and Fry, 2002; Humphreys and Francis, 2002), the results reveal that airport managers are now using several performance measures to monitor their businesses. Several authors comment on the importance of studying airport productivity. For instance Sarkis (2002) argues that evaluating airport operational efficiency is important for a number of reasons including communities’ reliance on airports for economic wellbeing; air carriers’ ability to choose among competing airports due to deregulation, and the fact that federal funding for airport improvements is based on performance measures. Performance evaluation and improvement studies of airport operations have important implications for a number of airport stakeholders. They assist air carriers in identifying and selecting more efficient airports on which to base their operations. Likewise municipalities would benefit from efficient airports in terms of attracting business and passengers. They also assist federal government in making effective decisions on optimal allocation of resources to airport improvement programs, and in evaluating the efficacy of such programs on the bottom line efficiency of airports. Finally, benchmarking their own airports against comparable airports is one way for operations managers to ensure competitiveness (Sarkis and Talluri, 2004). The need to develop appropriate service and productivity indicators for airport operation has been recognized and there is a small, but growing literature on the subject. Though there have been appeals to measure “overall productivity”, there is little evidence that the tools of productivity measurement that have been applied in other parts of the transport sector have had serious application in the case of airports (Tretheway, 1995). As the literature review in Chapter 2 will reveal, it only began in the late 1990s. Until
-3-
Assessment of Productive Efficiency of Airports
recently, a good number of studies have been conducted comparing productivity and operational efficiency of airports around the world, including Australia (Abbott and Wu, 2001; Hooper and Hensher, 1997), U.K. (Parker, 1999), U.S. (Gillen and Lall, 1997, 1998; Bazargan and Vasigh, 2003; Sarkis, 2000; Sarkis and Talluri, 2004), Spain (Martin and Roman, 2001), Brazil (Fernandes and Pacheco, 2001, 2002, 2005; Pacheco and Fernandes, 2003), Japan (Yoshida, 2004; Yoshida and Fujimoto, 2004). Occasionally the scope was expanded beyond a country to one continent such as Europe (Pels, Nijkamp and Rietveld., 2001, 2003) and international level (Adler and Berechman, 2001; Oum and Yu, 2004; Oum, Yu and Fu, 2003). Surprisingly, none has ever studied the productivity of airports operating in a specific market such as multiple airport systems (MAS), although they involve much more capital investment. At best, MAS airports are treated in the mixed samples with airports from single airport systems. The exception are only Pathomsiri and Haghani (2004); Pathomsiri, Haghani and Schonfeld (2005); Pathomsiri, Haghani, Dresner and Windle, (2006a) which will be summarized within this dissertation. Moreover, airport productivity research is rather restricted in the sense that productive efficiency is solely based on consideration of marketed outputs. Nonmarketed outputs or so-called “undesirable outputs” such as delays have been largely ignored, though they are also a major concern to airport stakeholders. This may be due to the lack of analysis technique. In principle where there is joint production of desirable and undesirable outputs, accounting for both of them intuitively should provide a more complete measure of airport productivity.
-4-
Assessment of Productive Efficiency of Airports
1.3
Motivation of the dissertation research This dissertation is motivated by several factors. First, it is clear that the aviation
society still has insufficient understanding regarding the productivity of airports operating in specific markets such as multiple airport systems. Second, the aviation society lacks understanding of the relationship between airport’s inputs and outputs, especially when undesirable byproducts are taken into consideration. Development of an applicable model is eminently necessary. The results should give a more complete measurement of airport productivity. Third, it is believed that the results have substantial implications which are very useful for managing airports. Last but not least, as the literature review in chapter 2 will reveal, the dissertation is a pioneering work. It likely creates an impact and entices researchers to re-think the way they assess productivity of airports. It is expected that further development of the applicable models for fairer assessment will follow. 1.4 Research objectives and scope This dissertation attempts to address the shortcomings of the previous airport productivity studies. In particular, it aims to accomplish the following four main objectives. 1) Assess the productivity of airports operating specifically in multiple airport systems as well as develop a model for predicting their relative efficiency. 2) Assess the productivity of U.S. commercial airports by accounting for joint production of desirable and undesirable outputs as well as develop a model for predicting their relative efficiency.
-5-
Assessment of Productive Efficiency of Airports
3) Estimate changes of airport productivity and sources of productivity growth during the study period, i.e., 2000 - 2003. 4) Analyze the impact of the inclusion of undesirable outputs on the productivity measurement and productivity growth. Since the research aims to provide timely information useful for managing airport in the modern era, the study period will span over recent years, i.e., 2000 – 2002 for research objective 1) and 2000 – 2003 for research objectives 2) to 4). Most data are expected to be from consolidated databases such as Federal Aviation Administration (FAA), Airports Council International (ACI), and Air Transport Research Society (ATRS). Supplement data may be collected directly from primary sources. 1.5 Research contributions The dissertation makes three major research contributions. First, rather than using mixed sample of airports, it assesses the productivity of airports operating in a similar market structure, i.e., multiple airport systems (MASs). Second, unlike previous airport productivity studies, this dissertation makes the first attempt to develop a more complete total factor productivity measure by also taking into account undesirable byproducts from airport operations, i.e., delays. Third, this dissertation also develops causal models for explaining variations in productive efficiency. The three contributions are accomplished by using recent panel data in two case studies. 1.6 Organization of the dissertation The dissertation is organized into eight chapters. The first chapter discusses the revolution of airport business, importance of airport productivity study, motivation of the
-6-
Assessment of Productive Efficiency of Airports
dissertation research as well as objectives and scope of the research. Chapter 2 reviews literature related to productivity study with emphasis in airport sector. Classification of productivity measures and a methodology to compute them are described. Since Data Envelopment Analysis is the most widely-used method for measuring airport productivity, its concept is briefly explained in this chapter to provide basic understanding of model development and its weaknesses. The DEA model will be used in one of the two case studies in this dissertation. Chapter 3 explains in details the proposed research methodology for assessing productivity of airports where joint production of desirable and undesirable outputs is taken into account. This chapter starts with the characterization of production possibility set, and illustration of output distance function and its modification, i.e., the directional output distance function which is the adopted model for analysis in the case study. The chapter also illustrates the computation of Malmquist and Luenberger productivity indexes and their components that are useful for explaining changes of productivity over time. Chapter 4 describes the first case study. The study is to assess productive efficiency of airports operating in MASs by using DEA as well as develop causal models for explaining variations in efficiency level. Chapter 5 presents and discusses the results from case study 1. Chapter 6 describes the second case study of 56 U.S. commercial airports. The contents cover modeling of airport operation, selection of inputs and output measures and characteristics of samples. The directional output distance function is applied to access the productivity of these airports. Chapter 7 presents and discusses the results. It provides contrast comparisons between with and without consideration of
-7-
Assessment of Productive Efficiency of Airports
undesirable outputs (delays). Other substantial results include productivity growth during 2000 – 2003, statistical analysis and scenario analysis. Important findings and insightful information are pointed out. Lastly, chapter 8 concludes the dissertation and suggests some potential areas for future research.
-8-
Assessment of Productive Efficiency of Airports
CHAPTER 2 LITERATURE REVIEW
This chapter rigorously reviews previous work directly related to airport productivity. The focus is on the classification of productivity measurement, the applied methodologies for measuring productivity, discussion of their advantageous and disadvantageous as well as the consideration for use. Summary of major findings and implications are also discussed. 2.1 Productivity measures In economics, productivity is defined as the amount of output per unit of input. In other words, the productivity measure is the ratio between output(s) and input(s). The definition, though very concise, is quite problematic to be applied in assessing productivity of airports. This is essentially due to the nature of airport operation which takes multiple inputs (such as labor and capital) for producing multiple outputs (such as movement of aircrafts, number of passengers and cargo throughput). Given various possible inputs and outputs, there are really many different ways of computing the productivity measure. Nevertheless, productivity measures can be categorized broadly into two groups of either partial factor or total (overall) factor productivity measures. 2.1.1 Partial Factor Productivity (PFP) measure Partial factor productivity (PFP) measures generally relate an airport’s output to a single input (factor). Labor productivity measures such as passengers per employee, aircraft movements per employee and ton landed per employee, are good examples. Table
-9-
Assessment of Productive Efficiency of Airports
2.1 summarizes more examples of PFP measures that have been used in airport business. A recent survey (Humphreys and Francis, 2002) revealed that the move towards privatization and commercialization has led to new performance measures being introduced to reflect the changing management goals. New measures fall into three categories, i.e., 1) financial measures to monitor commercial performance, 2) measures to meet the requirements of government regulators and 3) environmental measures. PFP measures have the advantage of being easy to compute, requiring only limited data and are easy to understand. As a result, many airport managers around the world usually adopt PFP measures to benchmark their performance (Francis, Humphreys and Fry, 2002; Humphreys and Francis, 2002). It is common to see such measures appear routinely in aviation trade publications (ACI 2002-2004; ATRS 2002 – 2003). Nevertheless, the measures can often be misleading when looking at the overall picture of the airport operation. For instance, it is possible to raise productivity in terms of one input, at the expense of reducing the productivity of other inputs. In the case of airports, which are fairly capital intensive, a partial productivity measure of labor productivity does not give a very clear picture of whether the performance of the institution is being improved (Abbott and Wu, 2002). Moreover, there are many possible PFP ratios, given multiple inputs and outputs of airport operation. There is usually a tradeoff among those measures. Airport may look better on one measure but can be worse on the others. As far as the overall assessment is concerned, it is preferable to use some form of overall (total) productivity measures that better shows the relation between all outputs and inputs.
- 10 -
Assessment of Productive Efficiency of Airports
Table 2.1 Examples of partial factor productivity measures in aviation sector
Scope of measure
Global performance of airport
Category
Profitability
Examples of performance measures
income per passenger rate of return on capital revenue to expenditure ratio profit per workload unit (WLU) cost per WLU (excluding depreciation and interest) operating cost per WLU capital cost per WLU labor cost per WLU aeronautical cost per WLU total revenue per WLU aeronautical revenue as a share of total aeronautical revenue per WLU non-aeronautical revenue per WLU concession revenue per area value added per unit of capital costs WLU per unit of net asset value total revenue per unit of net asset value WLU per employee revenue per employee value added per employee passengers/employee aircraft movements per runway aircraft movements per length of runway aircraft movements per hourly capacity passenger per aircraft movement service time for check-in time to reclaim baggage gate utilization rates passengers per terminal area baggage handled per unit of time baggage service reliability over time distances to reach departure gates crowding (passenger density) variability in service times passenger service ratings average time required to deliver freight at cargo terminal prior to aircraft departure theft and breakage rates index of aeronautical charges index of non-aeronautical charges aircraft turn-around times
Cost-efficiency
Cost-effectiveness (revenue earning)
Partial productivity measures
Capital productivity Labor productivity
Performance of particular processes
Runways
Passenger processing
Baggage handling Customer service Passengers
Cargo Airlines
Note: A “workload unit (WLU)” is equal to one passenger or 100 kilogram of cargo. Source: Hooper and Hensher (1997); Francis, Humphreys and Fry (2002); Humphreys and Francis (2002); Oum, Yu and Fu (2004).
- 11 -
Assessment of Productive Efficiency of Airports
2.1.2
Total Factor Productivity (TFP) measure In early 1990s, the literature on performance measurement for airports was
focused on the use of partial measures that yield an incomplete representation of the important relationship between multiple inputs and outputs. The lack of published research on overall measures of performance places a limit on our understanding of productive processes in the airport sector (Hooper and Hensher, 1997). As partial factor productivity measure indicates, performance has many dimensions. The growing literature on measuring the performance of airports is addressing the limitations of PFP measures in capturing all of those dimensions. A common way to deal with the problem of too many PFP measures is to derive an aggregate measure that takes into account all significant inputs and outputs simultaneously. Such measure is often called “Total Factor Productivity (TFP)” measure. Such overall TFP measure is useful for managers who are assessing the global productivity of an airport. It considers that different airports face different economic conditions and therefore may use input factors in varying proportions. For example, an airport that exhibits low labor productivity may not necessarily be inefficient from an overall perspective; it may merely be substituting capital with labor to take advantage of a wage rate (Nyshadham and Rao, 2000). TFP based measures have recently received increased attention in air transportation research and become a preferred measure. See for example Gillen and Lall (1997, 1998); Hooper and Hensher (1997); Oum and Yu (2004); Pathomsiri, Haghani, Dresner and Windle (2006a); Pels, Nijkamp and Rietveld (2001, 2003); Windle and Dresner (1992); and Yoshida and Fujimoto (2004). Since the TFP
- 12 -
Assessment of Productive Efficiency of Airports
measure is more suitable than PFP measures for assessing the productivity of airports, the subsequent review will focus on the methodology to derive the TFP measure. 2.2 Methodology for computing TFP measure There are several methods for deriving the TFP measure. The methods generally fall into two broad categories i.e., parametric and non-parametric approaches. Each approach has advantages and disadvantages. Their applicability usually depends on the availability of data. In some cases, both approaches are used to obtain complementary results (Pels, Nijkamp and Rietveld, 2001, 2003) or confirm the conclusions (Yoshida and Fujimoto, 2004). 2.2.1 Parametric approach Conceptually speaking, the parametric approach works in three major steps, i.e., 1) Transforming inputs into a common unit by assigning appropriate weights to individual inputs 2) Transforming outputs into a common unit by assigning appropriate weights to individual outputs so that an aggregate output can be computed and 3) Given a priori production function which represents logical relationship between the composite output in 2) and various transformed inputs in 1), estimate a set of parameters associated with individual transformed inputs. The results will give an estimated production function of airport operation explaining the transformation of inputs into outputs. With this function, it is possible to estimate the probable output level for a given set of inputs. Whenever the actual output is below the probable level, an airport is not being operated efficiently. In addition, by - 13 -
Assessment of Productive Efficiency of Airports
assuming that airports in the sample are similar; the productivity of airports can be benchmarked by comparing the difference between the actual output and the probable level. The further from the probable level means the less efficient operation. There are two major issues involved when one decides to use the parametric approach. First, what are the appropriate weights for transforming inputs and outputs? Second, what is the suitable functional form? Regarding the first question, Hooper and Hensher (1997) argue that the appropriate input weights should be the cost shares which represent the contributions of each input to costs. They also suggested that the output weights be the cost elasticities as long as they are readily available from prior research. However, in the most of empirical studies the absence of such elasticities has led to the use of revenue shares as proxies. Nyshadham and Rao (2000) have also adopted cost and revenue share respectively as input and output weights for their productivity assessment of 25 European airports. Hooper and Hensher (1997) commented that the use of prices as output proxies implicitly presumes that the airport is pricing efficiently but, since monopoly pricing is an issue of concern, it is problematic to derive an output measure from income. Indeed better measures for output quantity would have been landings for aeronautical output and passenger plus meeter-greeter throughput and the volume of cargo handled for nonaeronautical output. As for the second issue, the choice of a priori production function is rather subjective; and its suitability is usually based on the goodness-of-fit. Martin-Cejas (2002) estimates a deterministic cost frontier using translog function to assess the productive efficiency of 31 Spanish airports during 1996 – 1997. Pels, Nijkamp and Rietveld (2001)
- 14 -
Assessment of Productive Efficiency of Airports
estimate two stochastic production frontiers in their productivity study of 34 European airports during 1995 – 1997. The first function has number of passengers as the dependent (output) variable. The second function aims to explain the number of aircraft movements. Both of them are translog function. Based on the same dataset, their subsequent publication (Pels, Nijkamp and Rietveld, 2003) also estimate two stochastic production frontiers with the same two dependent (output) variables, but with different set of explanatory variables. The literature review indicates that translog is the most widely -used function in airport productivity studies. Although there are issues on weights and selection of production form, the parametric does have some advantages over the non-parametric approach. First of all, it can both measure and explain inefficiency simultaneously. Second, the parametric method allows for statistical testing of the presence of a deviation from the efficient frontier and returns to scale. Table 2.2 summarizes previous studies that used a parametric approach. It can be seen that there are very few studies. Availability of cost and revenue sharing data seem to be a big hurdle that limits the applicability of this approach. Many researchers therefore have resorted to an alternative approach, i.e., nonparametric.
- 15 -
Assessment of Productive Efficiency of Airports
Table 2.2 List of publications on airport productivity studies by parametric approach Author(s) Pels, Nijkamp and Rietveld (2001) Productivity model Year: 1995-1997 Stochastic (pooled cross-section production time series) frontier (SPF) Size: 34 European airports Sample Functional form
translog Air transport movements (ATM) = f{constant, number of runways, number of aircraft parking positions at the terminal, number of remote aircraft parking positions} Number of passengers (PAX) = f{constant, number of baggage claim units, number of aircraft parking positions at the terminal, number of remote aircraft parking positions} translog Total cost (TC) = f{unit of traffic transported, labor price, capital price}
Remark - Compute the most productive scale size (mpss) which represents the maximum productivity for any given input-output combination (Banker, 1984).
Martin-Cejas (2002)
Pels, Nijkamp and Rietveld (2003)
Year: 1996 – 1997 (pooled cross-section time series) Size: 31 Spanish airports Year: 1995-1997 (pooled cross-section time series) Size: 34 European airports
Deterministic cost frontier
- Unit of traffic transported (UT) = number of passengers + (kilograms of freight/100) - All variables (except dummies) are standardized around mean. - Treat number of runways as a fixed factor - Estimate also the DEA model (see Table 2.4 for the same authors)
Stochastic production frontier (SPF) and inefficiency model
translog Air transport movements (ATM) = f{constant, year dummy, airport area, number of runways, number of aircraft parking positions at the terminal, number of remote aircraft parking positions} ATM Inefficiency = f{slot coordination dummy, time restriction dummy} Air passenger movements (APM) = f{constant, , year dummy, predicted ATM, number of check-in desks, number of baggage claim units } APM Inefficiency = f{ constant, time restriction dummy, average airlines’ load factor}
- 16 -
Assessment of Productive Efficiency of Airports
2.2.2
Non-parametric approach The key characteristic of non-parametric approach is that it does not need to
specify a priori production function. No parameter needs to be estimated. Among other methods, index number and Data Envelopment Analysis (DEA) are the most popular in previous airport productivity studies. The index number method works similarly to the first two steps of parametric approach. Each input (output) needs to be assigned an appropriate weight so that individual inputs (outputs) are transformed into the same unit of measurement. Thus a weighted aggregate input (output) can be computed. The resulting aggregate input/output are called input/output indexes. By definition the total productivity index is the ratio of the weighted aggregate output index to a weighted aggregate input index. The higher value of TFP indicates higher efficiency. Thus the TFP measure can be used to rank performance of airports. Since the method involves weights, the discussion of weight issue in parametric approach is applicable here. In their productivity study of four Australian airports during 1989 – 1992, Hooper and Hensher (1997) use cost and revenue shares respectively as associated weights to inputs and outputs and obtain aggregate input and output indexes. Similarly, Nyshadham and Rao (2000) also use cost and revenue shares in their productivity study of 25 European airports. Other studies that adopted index number approach to compute TFP measure include Oum, Yu and Fu (2003), Oum and Yu 2004Yoshida (2004), Yoshida (2004), as well as Yoshida and Fujimoto (2004). Table 2.3 summarizes publications on airport productivity studies using index number method.
- 17 -
Assessment of Productive Efficiency of Airports
Table 2.3 List of publications on airport productivity studies using index number method Author(s) Hooper and Hensher (1997) Sample Year Year: 1988- 9, 1991-2 Size: 6 Australian airports Model Multilateral translog index number of TFP Input 1. capital stock 2. labor expenditures 3. other costs Output 1. deflated aeronautical revenues 2. deflated nonaeronautical revenues Remark - Use share of revenues as weights to compute output index - Estimate two regression models for estimating output-adjusted TFP. One model regressed TFP with output index (composite of cargo tonnages, movements, passengers, employers and labor costs). The other adds airport dummy variable. - A work load unit is defined as either one passenger or 100 kilograms of cargo. - Use percentage share of the revenue as weights to compute output index - Use percentage share of cost as weights to compute input index - Compute Spearman rank correlation between TFP and several PFP measures; then estimate a regression model for explaining TFP by PFP measures.
Nyshadham and Rao (2000)
Year: 1995 Size: 25 European airports
Multilateral translog index number of TFP
1. operating cost per work load unit 2. capital cost per work load unit 3. other costs per work load unit
1. aeronautical revenue per work load unit (WLU) 2. non-aeronautical revenue per work load unit (WLU)
- 18 -
Assessment of Productive Efficiency of Airports
Table 2.3 List of publications on airport productivity studies using index number method (Continued) Author(s) Oum, Yu and Fu (2003) Sample Year Year: 1999 Size: 50 major airports in AsiaPacific, Europe and North America Model Endogenous Weighted (EW) TFP input and output index numbers Input 1. number of fulltime equivalent employees who work directly for an airport operator 2. number of runways 3. number of gates 4. other costs l 1. runway length 2. terminal area Output 1. aircraft movements 2. number of passengers 3. cargo throughput 4. non-aeronautical revenues 1. aircraft movements 2. number of passengers 3. cargo throughput 1. aircraft movements 2. number of passengers 3. cargo throughputs - Use DEA efficiency score as truncated dependent variable and estimate a Tobit regression model to check inefficiency of regional airports and airports operated in 1990s. - Use EW-TFP index number as a dependent variable and estimate a regression model to check inefficiency of regional airports and airports operated in 1990s. Remark - Estimate a regression model for explaining the variation in TFP: TFP = f{constant, Asia-pacific dummy, airport size, % international passengers, % of aeronautical revenues}
Yoshida (2004)
Year: 2000 Size: 30 Japanese airports
Yoshida and Year: 2000 Fujimoto Size: 67 Japanese (2004) airports
Endogenous weighted (EW) TFP input and output index numbers DEA-InputCRS DEA-InputVRS Endogenous weighted (EW) TFP index number
1. runway length 2. terminal area 3. number of employees in the terminal 4. average access cost
- 19 -
Assessment of Productive Efficiency of Airports
Table 2.3 List of publications on airport productivity studies using index number method (Continued) Author(s) Oum and Yu (2004) Sample Year Year: 2000 – 2001 Size: 76 airports in Asia Pacific, Europe and North America Model Variable Factor Productivity (VFP – input index number) Input 1. number of fulltime equivalent employees who works directly for an airport operator 2. other costs Output 1. aircraft movements 2. number of passengers 3. cargo throughput 4. non-aeronautical revenues Remark - Estimate a regression model for explaining the variation in VFP: VFP = f{constant, airport size, % international passengers, % cargo traffic, capacity constraints, % of non-aeronautical revenues, outsourcing dummy}
- 20 -
Assessment of Productive Efficiency of Airports
One of the difficulties in using this method is that it requires a complete set of prices and quantity data. In many cases, these data are not available. Due to data limitations Hooper and Hensher (1997) were restricted to a few years during the early 1990s for four airports, so the study essentially presented a very limited indication of the performance of the Australian airport systems. Another weakness in this approach is the use of total revenue of the airports as an indicator of output. It is justifiable as long as prices, and therefore revenue, are not a reflection of the degree of market power of the institution considered. In the case of airports this might be the case and so it is preferable to use a total factor productivity valuation approach that does not depend upon prices that might be distorted by market imperfections (Abbott and Wu, 2002). Martin and Roman (2001) argued that some financial measures can be misleading indicators, as a consequence of the relative market power that might exist. Monopolistic airports might be able to make substantial profits even if they were inefficient. More importantly, prices are applicable for marketed outputs only, but it is difficult to calculate for non-marketed outputs, such as delays, noise and other externalities. During the past decade, aviation researchers have resorted to use an alternative non-parametric method which gets away from weight issue i.e., Data Envelopment Analysis (DEA). DEA is perhaps the most widely used method for assessing productivity of airport, regardless of approaches. DEA may be a true non-parametric method. It does not require any weights. It does not need to assume a production function. Instead, it builds an empirical piecewise linear production function from sample data. The only required data are the quantity of inputs and outputs. This is perfectly applicable in airport context where the breakdown between revenue and average prices for freight cargo and passenger
- 21 -
Assessment of Productive Efficiency of Airports
traffic are not made available, but output and input volume figures are. Therefore, DEA is the ideal method for estimating TFP measure. During the past decade, DEA seems to be the prevailing method used in assessing airport productivity. Table 2.4 lists publications that adopted DEA as an analytical method. Since DEA is the prevailing method in airport productivity study, the next section will be devoted to the review of DEA. It should be noted that the review is by no means exhaustive, but is focused on model development and some important features. The publications in Table 2.4 will also be referred to more. For more theoretical insights and applications about DEA, a good number of textbooks can be consulted. See for examples in Charnes, Cooper, Lewin and Seiford (1994); Cooper, Seiford and Tone (2000); Zhu (2003); Cooper, Seiford and Zhu (2004); Ray (2004); Cook and Zhu (2005).
- 22 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA Author(s) Gillen and Lall (1997) Sample Year Year: 1989 – 1993 Size: 23 of the top U.S. airports Model DEAOutput-CRS Input I. Terminal services 1. number of runways 2. number of gates 3. terminal area 4. number of employees 5. number of baggage collection belts 6. number of public parking spaces II. Movements 1. airport area 2. number of runways 3. runway area 4. number of employees Output I. Terminal services 1. number of passengers 2. pounds of cargo Remark - Estimate two Tobit regression models for explaining terminal and movements efficiency
DEAOutput-VRS
II. Movements 1. air carrier movements 2. commuter movements
- 23 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Gillen and Lall (1998) Sample Year Year: 1989 – 1993 Size: 22 of the top U.S. airports Model DEAOutput-CRS Input I. Terminal services 1. number of runways 2. number of gates 3. terminal area 4. number of employees 5. number of baggage collection belts 6. number of public parking spaces II. Movements 1. airport area 2. number of runways 3. runway area 4. number of employees Output I. Terminal services 1. number of passengers 2. pounds of cargo Remark - Compute Malmquist TFP by component
DEAOutput-VRS
II. Movements 1. air carrier movements 2. commuter movements
- 24 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) MurilloMelchor (1999) Sample Year Year: 1992 – 1994 Size: 33 Spanish civil airports under management of AENA (Spanish Airports and Air Transport) Model DEAInput-CRS DEAInput-VRS Input 1. number of workers 2. accumulated capital stock approximated by the amortization estimated in constant value 3. intermediate expenses 1. employment 2. capital stock 3. non labor cost 4. capital cost 5. changes in gross domestic product (GDP) Output 1. number of passengers Remark - Compute Malmquist index for individual pair of years
Parker (1999)
Salazar De la Cruz (1999)
Year: financial years (as of March 31) from 1988/89 – 1996/97 Size: 22 UK airports, including all of British Airports Authority (BAA)’s major airports Year: 1993 – 1995 Size: 16 main Spanish airports serviced mixed domestic and international passenger traffic; range 1 – 20 million passengers
DEAInput-VRS
1. number of passenger 2. cargo and mail
- Compute mean efficiency rating over 88/89 – 96/97 and use it to rank 22 airports before and after privatization.
DEAOutput-CRS
1. total economic cost e.g., cost for annual operations, the current costs and the internal interest on the net assets
1. annual passengers 2. total returns 3. returns on infrastructure services 4. operative returns 5. final returns
- Empirically, observe the extent to which input and output contribute to the change in efficiency by visualizing from graph
- 25 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Sarkis (2000) Sample Year Year: 1990 – 1994 Size: 44 major U.S airports Model DEAInput-CRS DEAInput-VRS Input 1. operating costs 2. number of airport employees 3. number of gates 4. number of runways Output 1. operational revenue 2. number of passengers 3. aircraft movements 4. general aviation movements 5. amount of cargo shipped Remark - Include the following variants 1. Simple cross-efficiency (SXEF) (Doyle and Green, 1994) 2. Aggressive cross-efficiency (AXEF) (Doyle and Green, 1994) 3. Ranked efficiency (RCCR) (Anderson and Peterson, 1993) 4. Radii of classification ranking (GTR) (Rousseau and Semple, 1995) - Perform nonparametric MannWhitney U-test to test the differences of efficiency scores between hub/non-hub, MAS/SAS, and snowbelt/non-snowbelt
- 26 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Adler and Berechman (2001) Sample Year Year: 1996 Size: 26 airports in Western Europe, North America and the Far East Model DEAInput-VRS (dual formulation) Input 1. peak short and medium haul charges 2. inversed number of passenger terminal 3. inversed number of runways 4. distance to the city center 5. minimum connecting time international – international 6. average delay per aircraft movement in minutes Output Three principal components derived from the following five measures of service quality from airlines’ perspective 1. suitability 2. operational reliability and convenience 3. cost of using airport 4. overall satisfaction and airport quality 5. factual questions with respect to the wave system and demand Remark - Survey airport quality of service from airlines rating 14 questions on Likert scale; and due to excessive number of total variables (inputs + outputs), the authors apply Principal Component Analysis (PCA) statistical method to reduce the total number inputs/outputs - Apply super-efficient DEA model (Anderson and Peterson, 1993) to fully rank the airports and report unbound results (infeasibility in primal) for some airports.
- 27 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Sample Year Fernandes Year: 1998 and Pacheco Size: 35 Brazilian (2001) domestic airports Model Input DEA-Non1. mean number of oriented-CRS employees 2. payroll expenditure, including direct and indirect benefits 3. operating expenditures 4. apron area 5. departure lounge area 6. number of check-in counters 7. length of curb frontage 8. number of vehicle parking spaces 9. baggage claim area DEA1. labor expense Output-VRS 2. capital expense, DEAincluding Output-CRS amortization of fixed assets 3. material expense Output 1. number of passengers, 2. cargo plus mail, 3. operating revenues 4. commercial revenues 5. other revenues Remark
Martin and Roman (2001)
Year: 1997 Size: 37 Spanish airports
1. air traffic movements 2. number of passengers 3. tonnage of cargo
- Compute technical efficiency by using reciprocal of efficiency score obtained from solving DEA - Compute scale efficiency - Interpret target output and input slack
- 28 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Pels, Nijkamp and Rietveld (2001) Sample Year Year: 1995 – 1997 (pooled crosssection time series) Size: 34 European airports Model Air transport movements (ATM) model DEAInput-CRS DEAInput-VRS Input 1. airport area 2. runway length 3. number of aircraft parking positions at the terminal 4. number of remote aircraft parking positions Output 1. Air transport movements (ATM) Remark - Estimate also the stochastic production frontier (see Table 2.2 for the same authors)
1. terminal area Air passenger 2. number of movements aircraft parking (APM) model positions at the terminal 3. number of remote aircraft parking positions 4. number of check-in desks 5. number of baggage claim units
1. Air passenger movements (APM)
- 29 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Abbott and Wu (2002) Sample Year Year: 1989/1990 to 1999/2000 Size: 12 main Australian airports Year: 1998/99 Size: 12 main Australian and 13 other international airports Fernandes Year: 1998 and Pacheco Size: 33 Brazilian (2002) major domestic airports Model DEAInput-CRS Input 1. number of staffs 2. capital stock 3. runway length 1. number of staffs 2. runway length 3. land area 4. number of aircraft standing areas 1. area of apron 2. area of departure lounge 3. number of check-in counters 4. length of frontage curb 5. number of parking spaces 6. baggage claim area Output 1. number of passengers 2. freight cargo in tons Remark - Compute Malmquist total factor productivity (TFP) index, - Estimate Tobit regression for explaining variation in Malmquist TFP
DEAInput-CRS
DEAOutput-VRS
1. number of passengers 2. freight cargo in tons 1. domestic passengers
- Analyze inefficiency level, slacks, potential number of domestic passengers in comparison to demand forecast
- 30 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Bazargan and Vasigh (2003) Sample Year Year: 1996 – 2000 Size: Top 45 U.S. airports, 15 each from large, medium and small hubs (by FAA’s definition) during the study period Model DEAInput-CRS Input 1. operating expenses 2. non-operating expenses 3. number of runways 4. number of gates including gates with jet ways and other non jet- way gates 1. average number of employees 2. payroll, including direct and indirect benefits 3. operating expenses 1. land area 2. number of runways 3. area of runways Output 1. number of passengers 2. air carrier operations 3. number of commuters, GA and military 4. aeronautical revenues 5. non-aeronautical revenues 6. percentage of ontime operations 1. domestic passengers 2. cargo plus mail 3. operating revenues 4. commercial revenues 5. other revenues, 1. aircraft movements 2. number of passengers Remark - Achieve a full ranking of all airports by introducing a virtual super efficient airport with existing airports so that there will be only one efficient airport. Its inputs and outputs are as follows: - Test the difference among three hub types by non-parametric Kruskal-Wallis test.
Pacheco and Year: 1998 Fernandes Size: 35 Brazilian (2003) domestic airports
DEAInput-VRS
- Use efficient scores from Fernandes and Pacheco (2002) as physical efficiency score and management efficiency score from this study to create Boston Consultancy Group (BCG) matrix - Perform paired-sample t-test to see if there is significant difference in efficiency scores before and after September-11. - Compute target inputs and outputs for inefficient airports
Pathomsiri and Haghani (2004)
Year: 2000, 2002 Size: 63 airports in multiple airport system worldwide
DEAOutput-VRS
- 31 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Pels, Nijkamp and Rietveld (2003) Sample Year Year: 1995-1997 (pooled crosssection time series) Size: 34 European airports Model Air transport movements (ATM) model DEAInput-CRS DEAInput-VRS Input 1. airport area 2. number of runways 3. number of aircraft parking positions at the terminal 4. number of remote aircraft parking positions Output 1. Air transport movements (ATM) Remark - Estimate also the stochastic production frontier (see Table 2.2 for the same authors) - Number of runways is treated as a fixed factor and adopted Banker and Morey (1986) formulation.
Air passenger 1. ATM movements 2. number of (APM) model check-in desks 3. number of baggage claim units
1. Air passenger movements (APM)
- 32 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Sarkis and Talluri (2004) Sample Year Year: 1990 – 1994 Size: 44 major U.S. airports Model DEAInput-CRS Input 1. operational costs 2. number of airport employees 3. number of gates 4. number of runways Output 1. operational revenue 2. passengers 3. aircraft movements 4. number of general aviation movements 5. total cargo I. Financial performance 1. operating revenues 2. commercial revenues 3. other revenues II. Operating performance 1. passengers embarked plus disembarked 2. tonnage of cargo embarked plus disembarked Remark - Rank airports by mean crossefficiency scores (AXEF) (Doyle and Green, 1994) - Identify benchmarks by using the hierarchical clustering technique based on correlation coefficients of the columns in the crossefficiency matrix. The average linkage method is utilized to derive the clusters. Airports in each cluster have a benchmark.
Fernandes Year: 1998 and and Pacheco 2001 (2005) Size: 58 airports administered by the Brazilian Airport Infrastructure Enterprise, Infraero
DEAInput-VRS
1. payroll, including direct and indirect benefits 2. operating and other expenses 3. average number of employees
- 33 -
Assessment of Productive Efficiency of Airports
Table 2.4 List of publications on airport productivity studies using DEA (Continued) Author(s) Pathomsiri, Haghani and Schonfeld (2005) Pathomsiri, Haghani, Dresner and Windle (2006a) Sample Year Year: 2000 , 2002 Size: 72 airports in multiple airport system worldwide Year: 2000 - 2002 Size: 72 airports in multiple airport systems worldwide Model DEAOutput-VRS Input 1. land area 2. number of runways 3. area of runways 1. land area 2. number of runways 3. area of runways Output 1. aircraft movements 2. number of passengers 1. aircraft movements 2. number of passengers Remark - Use parametric and nonparametric statistical methods to test the difference of efficiency scores before and after September 11 - Estimate Tobit regression model to explain variation in airport productivity
DEAOutput-VRS
- 34 -
Assessment of Productive Efficiency of Airports
2.3 2.3.1
Data Envelopment Analysis (DEA) Background DEA is a relatively new “data oriented” approach for evaluating performance of a
set of peer entities called Decision Making Units (DMUs) which convert multiple inputs into multiple outputs. The story of DEA begins with Edwardo Rhodes’s Ph.D. dissertation research at Carnegie Mellon University. The research was to evaluate Program Follow Through – the educational program for disadvantaged students (mainly black and Hispanic) undertaken in U.S. public schools with support from the Federal Government. It was the challenge of estimating relative technical efficiency of the schools involving multiple outputs and inputs, without using the information on prices that resulted in the formulation of the CCR (Charnes, Cooper and Rhodes) ratio form of DEA and the first publication (Charnes, Cooper and Rhodes, 1978). The DEA models use the optimization method of mathematical programming to generalize the Farrell (1957) single-output/input technical efficiency measure to the multiple-output/multiple-input case. Thus DEA began as a new Management Science tool for technical-efficiency analyses of public sector DMUs (Charnes, Cooper, Lewin and Seiford, 1994). The definition of a DMU is generic and flexible (Cook and Zhu, 2005; Cooper, Seiford, Zhu, 2004). Since the introduction in 1978, researchers in a number of fields have quickly recognized its usefulness and applicability. In recent years, there have been a great variety of applications of DEA in evaluating the performances of many kinds of entities engaged in many different activities in many different contexts in many different countries (Cooper Seiford and Tone, 2000; Cook and Zhu, 2005; Cooper, Seiford and Zhu, 2004). Seiford (1996) provides a bibliography since its first publication in 1978 to - 35 -
Assessment of Productive Efficiency of Airports
1995. Some textbooks exclusively cover DEA (Cooper, Seiford and Tone, 2000; Zhu, 2003; Ray, 2004; Cooper, Seiford and Zhu, 2004; Cook and Zhu, 2005). From time to time journals publish special issues on DEA theory and applications (Haynes, Stough, and Shroff, 1990; Cooper, Seiford, Thanassoulis and Zanakis, 2004). Emrouznejad (2006) has maintained a website that describes a rich family of DEA models. DEA opened up possibilities for use in cases which have been resistant to other approaches because of the complex (often unknown) nature of the relationship between the multiple inputs and multiple outputs involving DMUs. DEA requires very few assumptions. It does not need a priori assumption on functional form. This has made DEA applications quickly pervasive. In transportation, DEA has been applied to assess productivity of several activities such as public transit (Kerstens, 1996; Pina and Torres, 2001; Boame, 2004; Boame and Obeng, 2005), railway (Coelli and Perelman, 1999), large-scale distribution systems (Ross and Droge, 2004), ports (Tongzon, 1995; Budria, Diaz-Armas, Navarro-Ibanez and Ravelo-Mesa, 1999; Tongzon, 2001; Itoh, 2002; Turner, Windle and Dresner, 2004), and airlines (Schefczyk, 1993; Scheraga, 2004; Pires Capobianco and Fernandes, 2004). DEA has become a useful analytical tool for productivity study and performance analysis during the past two decades. In airport sector, researchers started using DEA in the late 1990s. The early works include Gillen and Lall (1997, 1998); Murillo-Melchor (1999); and de la Cruz (1999). Recently, there are a good number of publications using DEA to assess productivity of airports in different regions. Table 2.4 summarizes DEA publications in airport sector. For each study, the Table describes author(s), sample characteristics, analysis period, type of applicable DEA model, as well as set of inputs and outputs. The remark in the last
- 36 -
Assessment of Productive Efficiency of Airports
column notes major extra work beyond application of DEA in those studies. For example, after solving DEA models, Gillen and Lall (1997); Pathomsiri, Haghani, Dresner and Windle (2006a) estimated Tobit regression models for explaining variation in output efficiency scores. Meanwhile, Gillen and Lall (1998) and Abbott and Wu (2001) compute Malmquist index to explain changes of total factor productivity over time. 2.3.2 Model development
Output Y
B
CB = efficient production frontier
A4 A8 B’
C’B’ = average production function
A2
Positive residual
A5
Negative residual
A3
A6 A7
C C’ O A1
Input X
Figure 2.1 Difference between DEA and regression concept DEA is a methodology directed to frontiers rather than central tendency. Suppose
that there is a set of hypothetical airports whose airside operation take only single input X (e.g., runway) and produces single output Y (e.g., aircraft movements). Their input and output measures are scatter plotted in Figure 2.1. If one were to use a regression model to
- 37 -
Assessment of Productive Efficiency of Airports
estimate the production function of this operation, the fitted line would be C’B’ which passes through the “cloud” of data points. This regression line basically explains the average production. For airports A1, A6, and A7 this production function gives negative residuals or overestimated production. Meanwhile the fitted line will give positive residuals or underestimated production for airports A2, A4 and A8. The residuals are the portion of production that results from other factors beyond this single output. Instead of fitting a line to the data, DEA tries to learn from airports that lie above the line (airports with positive residuals). These airports are outliers that provide a good benchmark meaning that for a given input X, there is no other airport producing more Y. An alternative line CB is therefore drawn to represent the maximum possible production function or efficient production frontier that encompasses all airports. Any airports on this frontier are regarded as efficient whereas other airports within the frontier are inefficient. The further an airport is from the frontier, the more inefficient it is. DEA determines the efficient production frontier by estimating the distance for individual airports. Figure 2.2 explains the mechanism. DEA checks each airport to find out whether it lies on the frontier. Consider airport A3 which is below the frontier CB. Denote a scalar multiplier ? to current output y2 for boosting the production to the maximum level at A9 on the frontier. A9 may be viewed as a virtual airport whose input and output levels are the linear combination of airports A2 and A8, i.e., (?1x1+ ?2x2, ?1y1, ?2y2). Intuitively, for efficient airports, i.e., A2 and A8, their multipliers equal to one because they do not need to boost the production any further. All other inefficient airports will have some value depending on how inefficient they are. In real application, the production consists of multiple inputs and
- 38 -
Assessment of Productive Efficiency of Airports
outputs, rather than single input/output as shown in Figure 2.2; therefore it would be impossible to visualize. In this case the multiplier ? can be estimated by solving the following LP problem:
Output Y
B
A8 (x2, y2)
A9 (?1x1+ ?2x2, ?1y1, ?2y2)
A4
A11 A2 (x1, y1) yo A10
?
A5 A3 (xo, yo) A6 A7
C O
A1 xo Figure 2.2 Determination of efficient production frontier
Input X
max ? k s.t.
k?K
?? ??
k
+ y km ? s m = ?y km , m = 1,..........., M , ? x kn + s n = x kn , n = 1,.............., N ,
(2.1)
k?K
k
? k ? 0, k = 1,.........., K
Where k , m and n represent index of airports (k = 1,2......, K ) , index of outputs
(m = 1,2......, M ) and index of inputs (n = 1,2......, N ) respectively. ?k is an intensity vector associated with each airport and has k elements. x kn , y km are quantity of input n
- 39 -
Assessment of Productive Efficiency of Airports
+ ? and output m of airport k respectively. s m and s n are output and input slacks
respectively. ? or the output efficiency score is a scalar by which the current output level has to be multiplied in order to reach the frontier. If an airport is on the frontier, solving this LP will result in an optimal objective function ? * k = 1 . In other words, it is
sufficiently productive and does not need to increase output. ? is bound by [1, ?). The efficiency score can be used as a TFP measure. The LP needs to be solved k times, each time for an individual airport. The LP formulation in (2.1) is known as “Output-Oriented with Constant Returnto-Scale Characterization DEA model”, or in short DEA-Output-CRS hereinafter. As the name implies, the formulation seeks to determine if an airport is on the frontier in the output direction, for a given level of inputs. The analysis provides an assessment of how efficiently the inputs are being utilized. The DEA-Output-CRS has been used by several researchers including Gillen and Lall (1997, 1998); Fernandes and Pacheco (2002); de la Cruz (1999); Martin and Roman (2001); Pathomsiri and Haghani (2004); Pathomsiri, Haghani and Schonfeld (2005); Pathomsiri, Haghani, Dresner and Windle (2006a). In fact, the inefficiency can be determined in other directions as well. Figure 2.2 shows another two possible directions. The first is in the direction of input, i.e., projecting A3 to the frontier at A10. If this is the case, the corresponding LP formulation is given in (2.2).
- 40 -
Assessment of Productive Efficiency of Airports
min ? k s.t.
k?K
?? ??
k
+ y km ? s m = y km , m = 1,..........., M , ? x kn + s n = ?x kn , n = 1,.............., N ,
(2.2)
k?K
k
? k ? 0, k = 1,.........., K
All notations are the same as defined above. ? k or the input efficiency score is a scalar by which the current input level has to be multiplied in order to reach the frontier. The LP formulation in (2.2) is known as “Input-Oriented with Constant Return-to-Scale Characterization DEA model”, or in short DEA-Input-CRS hereinafter. The model determines whether there is inefficiency in input, for a given level of output. If an airport is on the frontier, solving this LP will result in an optimal objective function ? k* = 1 . In other words, the current level of input is probable and does not need to be reduced. ? is bound by (0, 1]. The input efficiency score can be used as a TFP measure. The LP needs to be solved k times, each time for an individual airport. The DEA-Iutput-CRS has been used by several researchers including Abbott and Wu (2002); Adler and Berechman (2001); Bazargan and Vasigh (2003); Fernandes and Pacheco (2005); Murillo-Melchor (1999); Pacheco and Fernandes (2003); Parker (1999); Pels, Nijkamp and Rietveld (2001, 2003); Sarkis (2000); Sarkis and Talluri (2004); and Yoshida and Fujimoto (2004). In Figure 2.2 there is another direction which is the shortest possible distance, i.e., projecting A3 to the frontier at A11. In this case DEA does not care about direction. The LP formulation simultaneously expands the output and contracts inputs. The efficiency score indicates inefficiency level in both input and output. The model is called “Nonoriented with Constant Return-to-Scale Characterization DEA model”, or in short DEA-
- 41 -
Assessment of Productive Efficiency of Airports
Non-oriented-CRS. The model is rarely used in airport productivity studies since it is not practical to freely adjust inputs/outputs mix. An airport manager is unlikely to choose the combination of capital inputs (e.g., runway, taxiway, and terminal building) and passenger throughput. Either input or output may be not controllable. It is found that only Fernandes and Pacheco (2001) adopted the DEA-Non-Oriented-CRS to assess the productivity of 35 Brazilian airports. The formulation is not given here, but can be found in several textbooks including Zhu (2003), Ray (2004) and Cooper, Seiford and Zhu (2004). Regardless of the chosen orientation, there is no effect on the classification of efficient airports because the resulting efficient frontiers are identical. However, it does affect results regarding inefficient airports. Researchers have to justify the choice of orientation. Regarding the use of input orientation, Abbott and Wu (2002) justify by reasoning that “airports have fewer controls over outputs than they do over inputs. The volume of airline traffic is somewhat exogenous to the control of airports’ managers depending as it does mainly on the general level of economic activity, both in the host city and the Australian and international economies more generally.” Meanwhile, Pacheco and Fernandes (2003) justify that they were dealing with Brazilian airports of various sizes. Martin and Roman (2001) justify the use of output orientation in their assessment of Spanish airports by reasoning that “We think that once an airport has invested in the building of new runways or new terminals, it is difficult for managers to disinvest to save costs, therefore invalidating the input-orientation.” Meanwhile, Fernandes and Pacheco
- 42 -
Assessment of Productive Efficiency of Airports
(2002) argue that the main issue of their analysis is the potential output from organizations with various sizes. In addition to the CRS frontier type, DEA can be carried out under the assumptions of variable returns to scale by introducing a scale constraint into the model. In VRS frontier type DMUs are not penalized for operating at a non-optimal scale (Banker, 1984; Banker and Thrall, 1992). Ganley and Cubbin (1992) consider the CRS frontier type as the long-term view as opposed to short-term view for VRS frontier. Martin and Roman (2001) argue that due to the existence of different scale airports in Spain, a VRS frontier should be used. Nonetheless they estimate also the CRS model. Parker (1999) argues that given the variation in the size of the airports in his dataset, VRS is the more realistic assumption than CRS. Murillo-Melchor (1999) however, argues that scale efficiency requires that the production size corresponds to the long-run. For this reason, this efficiency is assessed with respect to the technology of a long-run model i.e., constant returns to scale. Table 2.5 summarizes some important DEA models that have been used in previous airport productivity studies. The efficient targets in the last row compute the probable levels of input and outputs for those inefficient airports.
2.4 Discussion
The data availability on prices tends to limit the applicability of parametric approach. Literature review clearly indicates that non-parametric approach such as index number and DEA are more widely used by researchers. During the past decade, many researchers have adopted DEA to assess productivity of airports in different regions around the world.
- 43 -
Assessment of Productive Efficiency of Airports
Table 2.5 Summary of DEA models
Frontier type CRS
min ? s.t.
k?K
Input-oriented
Output-oriented
max ? s.t.
+ y km ? s m = y km , m = 1,..........., M , k?K
?? ??
k
?? ??
k
+ = ?y km , m = 1,..........., M , y km ? s m ? = x kn , n = 1,.............., N , x kn + s n
k?K
k
x kn + s = ?x kn , n = 1,.............., N ,
? n
k?K
k
? k ? 0, k = 1,.........., K
VRS Add
? k ? 0, k = 1,.........., K
k?K
?? ?? ??
k
=1
NIRS
Add
k?K
k
?1
NDRS
Add
k?K
k
?1
Efficient target
? ˆ kn = ?x kn ? s n x , n = 1,........., N + ˆ km = y km + s m y , m = 1,........., M
? ˆ kn = x kn ? s n x , n = 1,........., N + ˆ km = ?y km + s m y , m = 1,.........., M
Based on the review, it can be observed that previous studies assess productivity by only looking at desirable outputs such as passengers, aircraft movements, cargo and revenues. Inherently in the nature of airport operations, there are always undesirable byproducts being produced such as delays, mishandled baggage and accidents. In addition, airport operations also create externalities, notably noise and pollution. These byproducts may also be considered to be airport outputs, although undesirable, and they
- 44 -
Assessment of Productive Efficiency of Airports
are major concerns of the aviation industry. All airport stakeholders wish to minimize these undesirable outputs, or at least keep them at acceptable levels. Accounting for undesirable outputs in decision making is therefore, a goal of managers in the aviation industry. However, none of them considers joint production of desirable and undesirable outputs in the assessment, except Yu (2004). In that study, the author considered aircraft noise (in 1000 New Taiwan dollars) as the lone undesirable output. There are several limitations in this work. It is not clear how noise is measured and transformed into monetary unit. The sample size of 14 Taiwanese airports is too small when compare to the number of inputs (5) and outputs (3) measures. In DEA framework, the sample size should be much greater than number of inputs times outputs (Cooper, Seiford and Tone, 2000: page 252). Otherwise the discriminatory power will be deteriorated. That is the reason why Yu (2004) reports many efficient airports. Furthermore, other major undesirable outputs are excluded. In the US, delays are a major concern of air services. BTS (2006) routinely records on-time performance of flights and delays. In Europe, the situation about air traffic control is getting worse. In 2000, around 30% of flights experienced delays more than 15 minutes and air traffic control was the most important causes of delays (Martin and Roman, 2001). In fact some researchers have discussed about undesirable outputs but did not address them. In their ad-hoc Tobit regression models, Gillen and Lall (1997) noted that greater noise restriction tend to lower movement performance. To clean up noise, airports need to trade their movements low. Some researchers have pointed out the association between efficient airports and delays. Salazar de la Cruz (1999) observed that those airports that define the frontier show very high level of utilization, confirmed by further
- 45 -
Assessment of Productive Efficiency of Airports
congestion problems and expansion works. Furthermore, he suggests that it will be more prudent to consider the inefficiencies associated with level of usage, climate conditions, economy of design, construction or quality level including delays, etc. Especially, the necessity to consider the impact of capacity and delays jointly requires the introduction of specific behavioral models for each airport, information for which is not easily available. Based on previous work (Adler and Berechman, 2001; Bazargan and Vasigh, 2003; Fernandes and Pacheco, 2002; Gillen and Lall, 1997, 1998; Martin and Roman, 2001; Pacheco and Fernandes 2003; Pathomsiri and Haghani, 2004; Pathomsiri, Haghani and Schonfeld, 2005; Pathomsiri, Haghani, Dresner and Windle, 2006a; Pel, Nijkamp and Rietveld, 2001, 2003; Sarkis, 2000; Sarkis and Tulluri, 2004), DEA results tend to identify busy airports as efficient. Frequently, these efficient airports are also congested. It may be that one airport creates greater numbers of delayed flights than another, but produces the same level of desirable outputs per unit of input. Unless delays are taken into account, both airports would show the same productivity level. Consideration of undesirable outputs is not as straightforward as desirable outputs, but quite problematic. In DEA literature, there is a general guideline for distinguishing between input and output variables. If the lower level of measure is better, it should be classified as an input; but if the higher quantity is desirable, that variable is classified as an output. This is not true in airport operation where the higher quantity of undesirable outputs such as noise, pollution, delays, and accident are not desirable. Moreover, these outputs are not inputs in airport operation either. Even so, Adler and Berechman (2001) consider delay as an input.
- 46 -
Assessment of Productive Efficiency of Airports
Estimation is also an issue. The limitation lies in the mathematical mechanism for determining if an airport is on the efficient frontier. Using DEA-Output orientation would seek to maximize the expansion of all outputs, rather than maximize only the desirable outputs and minimize the undesirable. In reality, an airport manager never wishes to expand both number of passengers and delays simultaneously. The ad-hoc DEA is not applicable either (Färe and Grosskopf, 2004a, 2004b; Seiford and Zhu, 2002, 2005). The issue will need a special mathematical formulation. As a result, it is a very challenging task to analyze airport productivity where there is joint production of desirable and undesirable outputs. Accounting for both types of outputs should provide a more complete measure of airport productivity. Furthermore, consideration of undesirable outputs such as reduction in delays may lead to different evaluation which in turn results in different management policy. For example, it could affect the time when expansions and new facilities must be operated. This pioneer research will address this problem and attempt to point out the effects of joint consideration of desirable and undesirable outputs. It should be noted that all studies that are categorized as TFP are termed as such because they consider more than one input and output. It is virtually impossible to consider all of the factors in a productivity study. Since this dissertation also considers multiple important inputs and outputs, it can also be considered as a TFP study.
- 47 -
Assessment of Productive Efficiency of Airports
CHAPTER 3 RESEARCH METHODOLOGY
This chapter presents the methodology that will be used to assess the productive efficiency of airport operation with joint consideration of both desirable and undesirable outputs. The methodology is based on the production theory from economics discipline. The chapter starts off with the characterization of production technology in order to represent the relationship between input and output measures. The traditional axiom of production theory i.e., the distance function is then introduced. Taking this as a building block an optimization model, called the directional output distance function, is developed. It is of a non-parametric type applicable for modeling production system with multiple inputs and outputs and provides measures of performance without appealing to prices. Finally, the productivity index number is devised for use in analyzing productivity changes over time.
3.1 Characterization of production possibility set
In environmental economics one often wishes to distinguish between desirable
+ + ) outputs. In the production context the former output ( y ? RM ) and undesirable (b ? R J
is typically a marketable goods and the latter is often not marketed, but rather a byproduct which may have deleterious effects on the environment or human health, and therefore its disposal is often subject to regulation. As a result, it should be useful to explicitly model the effects of producing both types of outputs, taking into account their characteristics and their interactions (Färe and Grosskopf, 2004b).
- 48 -
Assessment of Productive Efficiency of Airports
Let’s consider a production process that desirable and undesirable outputs may be jointly produced, i.e., b is a byproduct of the production of y. Here, the application is an airport operation that processes throughputs of passengers, aircraft movements and cargo by using its infrastructure such as land, runway, and terminal. In this case, the desirable or marketable outputs are number of passengers, movements and amount of cargo transported. There is also undesirable byproduct, i.e., delays (others may include noise and pollution). The basic problem is that given technology, producing these throughputs means simultaneously producing delays even though their production is undesirable. The production technology T describes the possible transformations of inputs
+ + ( x ? RN ) into ( y ? RM ) and undesirable (b ? R J+ ) outputs. The production possibility
set is defined as a set of desirable and undesirable outputs that can be produced from a given level of inputs. This set is represented by:
P ( x) = {( y, b) : x can produce ( y, b)}
3.2 Output distance function
(3.1)
For the sake of illustration, assume that airport operation produces only two desirable outputs y1 and y 2 (which may be passengers and aircraft movements) from a given input vector. Figure 3.1 shows a hypothetical output possibility set. Note that the true shape of the set is unknown. The frontier of the set is defined as the output vector that cannot be increased by a scalar multiple without leaving the set. In the Figure, the frontier represents efficient combinations of outputs y1 and y 2 . However, not all airport operations are efficient; therefore there must be an inefficient airport that lies below the efficient frontier. The basic idea in distinguishing efficient airports from inefficient ones
- 49 -
Assessment of Productive Efficiency of Airports
is to determine how far the current operations are from the frontier. In Figure 3.1, airports A and B are right on the frontier; hence obviously are efficient airports. Meanwhile, airport C is away from the frontier by the distance BC or AC depending on the direction of measurement. As a result, airport C is not efficient.
y2 B A
E g
#
C
#
D
P(x)
0
O
Figure 3.1 Output possibility set and distance functions
y1
Shephard’s output distance function (Shephard, 1970) can be used to determine how far an airport is from the frontier. It is defined as the ratio of actual output to maximum potential output and equals to the reciprocal of Farrell’s output technical efficiency measure (Farrell, 1957). For any airport, the Shephard’s output distance function is: v Do ( x , y )
= inf{? : ( x, y ) ? P( x)} ? (3.2)
- 50 -
Assessment of Productive Efficiency of Airports
?1
y ? ? = ?sup{? : ( x, ) ? P( x)}? ? ? ?
(3.3)
r OC . This is a measure of how far the For airport C in Figure 3.1, Dc ( x, y ) = OA operation of airport C is from the frontier. The function is equal to one for all efficient airports and less than one for inefficient ones. The higher value of distance function indicates higher operational efficiency. The reciprocal of the output distance function ( OA ) or the Farrell measure gives the maximum proportional expansion in all outputs OC
that is feasible given inputs. The distance function completely characterizes the v production technology T , because as long as y ? P( x) ? Do ( x, y ) ? 1 (Färe, Grosskopf, Norris and Zhang (1994b); Färe and Primont, 1995). However, the generalization of the output distance function in (3.2) to include
v y b undesirable output by simply redefining it as Do ( x, y, b) = inf{? : ( x, , ) ? P( x)} ? ?
would not be meaningful since it would mean proportionate expansion of undesirable and desirable outputs as much as possible, without crediting the reduction of undesirables. In assessing productive efficiency of airports where there is joint production of desirable and undesirable outputs, this is not well-applicable. A rational airport manager should aim at maximizing only desirable, but minimizing undesirable outputs.
3.3 Directional output distance function
Due to the existence of both desirable and undesirable outputs in the output possibility set (3.1), the Shephard’s output distance function needs to be modified so that the efficiency measure will be able to credit for expansion of desirable and reduction of
- 51 -
Assessment of Productive Efficiency of Airports
+ undesirable outputs. First necessary notations are formally defined. Let y ? RM denote a + vector of desirable outputs, b ? R J+ denote a vector of undesirable outputs, and x ? R N
denote a vector of inputs. In airport context, K airports with ( x k , y k , bk ) are examined. The output possibility set P ( x) = {( y, b) : x can produce ( y, b)} in (3.1) satisfies certain axioms laid out by Shephard (1970), namely:
Property 1 Property 2 Property 3 Property 4
P (0) = {0,0}
+ P( x) is convex and compact for each x ? R N
( y, b) ? P( x) and ( y ' , b) ? ( y, b) imply ( y ' , b) ? P( x)
P( x) ? P( x' ) implies x ? x'
Property 1 states that zero inputs essentially yield zero outputs and any nonnegative input yields at least zero output. Sometimes this property is called a condition of no free lunch. Property 2 requires that only finite output should be produced given finite inputs. Property 3 imposes strong or free disposability of desirable outputs which means that it allows any desirable outputs to be disposed costlessly and still remain in P ( x). In other words, the disposal of any output can be achieved without incurring any costs in term of reducing the production of other outputs. Property 4 imposes strong or free disposability of inputs. The inputs are also allowed to be disposed costlessly. It also implies that an increase in any one input does not reduce the size of P ( x). Although in production theory it is common to assume that outputs are strongly disposable, it may not be appropriate for production technologies such as present airport operation in which undesirable outputs such as delays and noise cannot be costlessly - 52 -
Assessment of Productive Efficiency of Airports
disposed. Under regulated environment, an airport is forced to clean up its undesirable outputs or to reduce its levels. Desirable and undesirable outputs should be treated asymmetrically in terms of their disposability characteristics (Zaim, 2005). Even in the absence of regulations, increased environmental consciousness from stakeholders still require careful treatment of undesirable outputs as weakly disposable. To model the idea that there is a cost to reducing undesirable outputs, the next property is assumed.
Property 5
Weak disposability between desirable and undesirable outputs: If
( y, b) ? P ( x) and 0 ? ? ? 1, then (?y,?b) ? P( x). Property 5 implies that if undesirables are to be decreased, then the desirable outputs must also be decreased, holding inputs x constant. In other words, both desirable and undesirable outputs may be proportionally contracted, but undesirable outputs cannot, in general, be freely disposed. It models the idea that there is a cost to ‘cleaning up’ undesirable outputs. In the airport operation context, it implies that fewer delays can be achieved by letting an airport to service fewer aircraft movements. Finally, to recognize the nature of joint production of desirable and undesirable outputs, the following property is assumed:
Property 6
Null-jointness, if ( y, b) ? P( x), and b = 0, then y = 0.
This property states that if an output vector ( y, b) is feasible and there are no undesirable outputs produced, then under the null-jointness only zero desirable output can be produced. Equivalently, if some positive amount of the desirable output is produced then undesirable output must also be produced. In our airport operation context, nulljointness implies that where there are aircraft movements, there must be some delayed
- 53 -
Assessment of Productive Efficiency of Airports
flights which could be occurred by any cause (e.g., air carrier, extreme weather, nonextreme weather conditions, airport operations, late-arrival aircraft, security, human error, and accident). There are several ways of integrating the above six properties into the representation of output possibility set, including parametric and nonparametric approaches. The focus here is on the nonparametric model using DEA. The representation of output set is in the form of piecewise linear. Based on the six properties, the production technology for an individual airport k or P ( x k ) may be represented by the following output set: P ( x k ) = {( y, b) : (3.4)
k?K
?? ??
k
y km ? y km , m = 1,..........., M ,
kj
k?K
?? b
k k
= bkj , j = 1,............, J ,
k?K
x kn ? x kn , n = 1,.............., N ,
?k ? 0, k = 1,.........., K }
The constraints for the undesirable outputs b j , j = 1,....., J are equality
constraints, which under the constant returns to scale models the idea that these outputs are not freely disposable. Meanwhile free disposability of desirable outputs y m , m = 1,..., M and inputs x n , n = 1,...., N are allowed by using the inequalities in their respective constraints. ? k is an intensity vector. Figure 3.2 represents the construct of P( x) from four hypothetical airports i.e., A, B, C, and D. These airports are assumed to use the same amount of inputs, x , but produce
- 54 -
Assessment of Productive Efficiency of Airports
different amounts of desirable output, y and undesirable output, b. Since the linear programming of DEA approach is being used to estimate the output distance function, P( x) is drawn as piecewise linear rather than smooth curve as in Figure 3.1. The output possibility set, P( x), is bounded by 0ABCD0. Airports A, B, and C form an efficient frontier. y = desirable output G B I J H C
F (y, b)
A g = (gy, -gb)
P(x)
D 0 E b = undesirable output
Figure 3.2 Graphical illustration of directional output distance function concept
This figure illustrates how the assumptions are used in the construct. The origin (0,0) is included in P( x) because of the null-jointness assumption. The assumption of weak disposability implies that for any point on or inside P( x) , a proportional contraction in both ( y, b) is feasible. The vertical line segment CD occurs because of strong disposability between desirable outputs. The negative slope portion BC is possible because sometimes traffic may be blocked due to a long queue of delayed flights; hence
- 55 -
Assessment of Productive Efficiency of Airports
reducing throughput. Note that if undesirable outputs are ignored, P( x) will be the area bounded by 0GBCD0. Next, the interest is to assess the level of inefficiency for all airports which will tell how far each airport is from the efficient frontier. For airport F, the distance should be measured along the diagonal line FJ or in the direction of vector g = ( g y ,? g b ) . This measurement is justified on the premise that it seeks to maximize the expansion of desirable outputs and contraction of undesirable outputs simultaneously. The directional output distance function is then formulated as follows:
D 0 ( x, y, b; g y ,? g b ) = sup{? : ( y + ? g y , b ? ? g b ) ? P( x)}
?
(3.5)
To assess the level of inefficiency for an individual airport, the following linear programming problem is solved:
max ? s.t.
k?K
?? ??
k
y km ? y km + ? g y , m = 1,..........., M ,
kj
k?K
?? b
k k
= bkj ? ? g b , j = 1,............, J ,
(3.6)
k?K
x kn ? x kn , n = 1,.............., N ,
? k ? 0, k = 1,.........., K
The selection of a directional vector g = ( g y ,? g b ) is rather flexible. For example, using g = (0, b) implies that the level of inefficiency is measured along the horizontal line FH or projecting airport F to the frontier at H. Meanwhile, using g = ( y,0) yields the projection on the frontier at I. Using g = (1,?1) gives the same weight to both desirable and undesirable outputs. In this study, the vector g = ( y,?b) will be used, which means
- 56 -
Assessment of Productive Efficiency of Airports
that the projected direction depends on individual airport’s outputs. The linear programming in (3.6) is then rewritten as (3.7). max ? s.t.
k?K
?? ??
k
y km ? (1 + ? ) y km , m = 1,..........., M ,
kj
k?K
?? b
k k
= (1 ? ? )bkj , j = 1,............, J ,
(3.7)
k?K
x kn ? x kn , n = 1,.............., N ,
? k ? 0, k = 1,.........., K
The directional output distance function D 0 ( x, y, b; g y ,? g b ) or an optimal ? takes the minimum value of zero when it is not possible to expand the desirable outputs and contract undesirable outputs. This means that the airport is efficiently producing at the maximum possible outputs. To assess the productivity of K airports, the linear programming in (3.7) is solved K times, once for each individual airport. Thereafter, the optimal ? k will be called a efficiency score. A higher value of ? k indicates a lower level of efficiency. As a result, it can also be used to rank the performance of airports. The terms (1 + ? ) y km plus the corresponding output slacks and (1 ? ? )bkj in (3.7) give the projection of desirable and undesirable outputs onto the frontier. For an efficient airport with ? = 0 , the terms are simply ( y km , bkj ) or the current level of outputs. For inefficient airports, these terms represent the maximum possible production outputs or highest potential outputs that an airport could have produced. The results may provide benchmarks for airports to improve operational efficiency. However, as is shown in
?
- 57 -
Assessment of Productive Efficiency of Airports
Chapter 5, the selection of an appropriate set of outputs is crucial to the reasonableness of the benchmark. In order to relate the Shephard’s output distance function and the directional output distance function, let g = ( y, b) , then through (3.5), it becomes:
?
D 0 ( x, y, b; y, b)
= sup{? : ( Do ( x, ( y, b) + ? ( y, b)) ? 1} = sup{? : (1 + ? ) Do ( x, y, b) ? 1}
= sup{? : ? ?
1 ? 1} Do ( x , y , b ) (3.8)
=
1 ?1 Do ( x, y, b)
The expression in (3.8) shows that Shephard’s output distance function is a special case of the directional output distance function. The relation between the two can be written as: D 0 ( x, y, b; y,?b) =
?
1 ?1 Do ( x, y, b)
(3.9)
or equivalently,
Do ( x, y,?b)
=
1 v 1 + Do ( x, y, b; y,?b)
(3.10)
3.4
Malmquist productivity index with the presence of undesirable outputs
The concept of Malmquist productivity index was first introduced by Malmquist (1953) to compare the input of a production unit at two different points in time in terms of the maximum factor by which the input in one period could be decreased such that the
- 58 -
Assessment of Productive Efficiency of Airports
production unit could still produce the same output level of the other time period. The idea leads to the Malmquist input index. It has further been studied and developed in the non-parametric framework by several authors. See for example, among others, Caves, Christensen and Diewert (1982), Färe, Grosskopf, Lindgren and Roos (1994), Färe R., Grosskopf S., Norris M., Zhang Z. (1994) and Färe, Grosskopf and Roos (1998). It is an index representing the Total Factor Productivity (TFP) growth of a decision making unit (DMU), in that it reflects progress or regress in efficiency along with progress of the frontier technology over time under multiple inputs and multiple outputs framework. Given panel data, the Malmquist index evaluates the productivity change of an airport between two time periods. It is defined as the product of “Catch-up” and “Frontier-shift” terms. The catch-up (or recovery) term relates to the degree that an airport attains for improving its efficiency, while the frontier-shift (or innovation) term reflects the change in the efficient frontier surrounding the airport between the two time periods. Suppose that undesirable outputs are ignored. To analyze change of productivity over time, ( x, y ) is superscripted with corresponding time period. Then ( x t , y t ) and ( x t +1 , y t +1 ) are measures of inputs and outputs in period t and t + 1 respectively. From time t to t + 1 operational efficiency of airport k may change or (and) the frontier may shift. The output-oriented Malmquist productivity index using period t as the base period is the following (Caves, Christensen and Diewert, 1982).
t ? Do ( x t +1 , y t +1 ) ? Mo = ? ? t t t ? Do ( x , y ) ?
(3.11)
- 59 -
Assessment of Productive Efficiency of Airports
Alternatively, by using period t + 1 as the base period, the output-oriented Malmquist productivity index can be written as follows
? D t +1 ( x t +1 , y t +1 ) ? M o = ? o t +1 t t ? ? Do ( x , y ) ?
(3.12)
In order to avoid choosing an arbitrary frontier as reference, Färe, Grosskopf, Lindgren, and Roos (1994) suggest using the geometric average of the two indexes above. The resulting index is
t t +1 ? Do ( x t +1 , y t +1 ) Do ( x t +1 , y t +1 ) ? =? ? t t t t +1 t t ? Do ( x , y ) Do ( x , y ) ? 1/ 2
M
t +1 t
(3.13)
An alternative way of writing the Malmquist total factor productivity index is:
t t D t +1 ( x t +1 , y t +1 ) ? Do ( x t +1 , y t +1 ) Do (xt , yt ) ? = o t t t ? ? t +1 t +1 t +1 t +1 Do ( x , y ) ? Do ( x , y ) Do (xt , yt ) ? 1/ 2
M
t +1 t
(3.14)
Where the ratio outside the bracket measures the change in relative efficiency (i.e., the change in how far observed production is from maximum potential production) between period t and t + 1 . The geometric mean of the two ratios inside the bracket captures the shift in technology between the two periods evaluated at x t and x t +1 , that is
t +1 Do ( x t +1 , y t +1 ) t Do (xt , yt )
Efficiency change (EFFCH) =
(3.15)
t ? D t ( x t +1 , y t +1 ) Do (xt , yt ) ? Technical change (TECHCH) = ? t o +1 t +1 t +1 t +1 t t ? ? Do ( x , y ) D o ( x , y ) ?
1/ 2
(3.16)
Although the Malmquist index can in principal deal with undesirable outputs since it does not require knowledge on prices, the distance functions on which it is based
- 60 -
Assessment of Productive Efficiency of Airports
do not credit an airport for reducing level of undesirable outputs. Chung, Färe, and Grosskopf (1997) defined an output-oriented Malmquist-Luenberger (ML) productivity index that is comparable to the Malmquist index. Take undesirable outputs into consideration, ( x t , y t , b t ) and ( x t +1 , y t +1 , b t +1 ) are measures of inputs, desirable and undesirable outputs in t and t + 1 respectively. If the directional vector g t = ( y t ,?b t ) and g t +1 = ( y t +1 ,?b t +1 ) are chosen in corresponding periods, the output-oriented Malmquist-Luenberger productivity index is:
vt t t t t v t +1 t t t t 1/ 2 ? ? (1 + Do ( x , y , b ; y ,?b t )) (1 + Do ( x , y , b ; y ,?b t )) v t t +1 t +1 t +1 t +1 t +1 v t +1 t +1 t +1 t +1 t +1 t +1 ? =? ? (1 + Do ( x , y , b ; y ,?b )) (1 + Do ( x , y , b ; y ,?b )) ?
ML
t +1 t
(3.17)
v t +1 Here Do means that the reference technology is constructed based on data from
period t + 1 and the data being evaluated is included in the parentheses with its associated time period; for example ( x t , y t , b t ) would mean that the data to be evaluated are from period t. The directional vector g is time dependent. The definition is such that when the directional vector g is ( y, b) rather than ( y,?b) , the Malmquist-Luenberger index coincides with the Malmquist index. The Malmquist-Luenberger measure indicates productivity improvements if its value is greater than one. The value of less than one indicates decreases in productivity. In other words, it means that with the same amount of inputs as in period t + 1 , the greater quantity of outputs is produced as in period t (Murillo-Melchor, 1999). The productivity remains unchanged if M o is unity. Similar to the case of Malmquist, the Malmquist-Luenberger index can also be decomposed into two components, namely. - 61 -
Assessment of Productive Efficiency of Airports
MLEFFCH
t +1 t
vt t t t t 1 + Do ( x , y , b ; y , ?b t ) v t +1 t +1 t +1 t +1 t +1 t +1 = 1 + Do ( x , y , b ; y ,?b )
(3.18)
MLTECHCH tt +1 =
v t +1 t t t t v t +1 t +1 t +1 t +1 t +1 t +1 1 / 2 ? (1 + Do ( x , y , b ; y ,?b t )) (1 + Do ( x , y , b ; y ,?b )) ? vt t t t t v t t +1 t +1 t +1 t +1 t +1 ? ? t ? (1 + Do ( x , y , b ; y ,?b )) (1 + Do ( x , y , b ; y ,?b )) ?
(3.19) The product of (3.17) and (3.18) equals to MLtt+1 . The decomposition makes it possible to measure the change of technical efficiency and the movement of the frontier for a specific airport. Equation (3.18) measures the magnitude of technical efficiency change between periods t and t + 1 . The value of less than 1 indicates regress in technical efficiency. In other words, given a level of inputs, the same average output of all samples would have lead to produce more efficiently in period t + 1 than in period t (MurilloMelchor, 1999). Meanwhile the value greater than 1 indicates improvements. The technical efficiency remains unchanged if the value is unity. The second term measures the shift of frontier between periods t and t + 1 . Alternatively, Färe, and Grosskopf (2004b) construct another productivity index that has an additive structure, i.e., in terms of differences rather than ratios of MalmquistLuenberger ratio indexes in (3.17) – (3.19). The index is an output-oriented version of The Luenberger Productivity index introduced by Chambers (1996). Specifically, the index is: Ltt+1 =
v t +1 t +1 t +1 t +1 t +1 t +1 1 v t +1 t t t t Do ( x , y , b ; y ,?b t ) ? Do ( x , y , b ; y ,?b ) 2
[
- 62 -
Assessment of Productive Efficiency of Airports
vt t t t t v t t +1 t +1 t +1 t +1 t +1 + Do ( x , y , b ; y ,?b t ) ? Do ( x , y , b ; y , ?b )
]
(3.20)
Following the idea of Chambers, Färe, and Grosskopf (1996) the Luenberger productivity index can be additively decomposed into an efficiency change and a technical change component,
vt t t t t v t +1 t +1 t +1 t +1 t +1 t +1 LEFFCH tt +1 = Do ( x , y , b ; y ,?b t ) ? Do ( x , y , b ; y , ?b )
(3.21)
and LTECHCH tt +1 =
v t t +1 t +1 t +1 t +1 t +1 1 v t +1 t +1 t +1 t +1 t +1 t +1 D o ( x , y , b ; y , ?b ) ? D o ( x , y , b ; y , ?b ) 2
[
v t +1 t t t t vt t t t t + Do ( x , y , b ; y ,?b t ) ? Do ( x , y , b ; y ,?b t )
]
(3.22)
respectively. The sum of these two components equals the Luenberger productivity index. The index and its components signal improvements with values greater than zero, and declines in productivity with values less than zero. As usual, selection of the directional vector is flexible. If the vector g = ( y,?b) is chosen, i.e., choosing the observed desirable and (negative) undesirable output vector to determine the direction, then each airport may be evaluated in a different direction, i.e., in its own direction. This is just typically the case for Shephard type distance functions.
- 63 -
Assessment of Productive Efficiency of Airports
y = desirable output
M
#
Pt+1(x)
D’ N
L H A
#
D
g = (1, -1)
Pt(x)
0
b = undesirable output
Figure 3.3 The Luenberger productivity indicator
Figure 3.3 illustrates how the Luenberger productivity indicator is constructed. For simplicity, it is assumed that inputs are the same in period t and t + 1 and are represented by x = x t = x t +1 . Without loss of generality, a directional vector g = (1,1) is assumed and illustration is for the case of technological progress (the frontier shifts to the left at period t + 1 ). Given g = (1,1), the directional output distance function is an estimate of the simultaneous unit expansion in the desirable output and unit contraction in the undesirable output. An airport is observed to produce at point D in period t and at D' in period t + 1 . If the airport was to eliminate technical inefficiency it could operate at H in period t and at M in period t + 1 . The Luenberger efficiency change indicator is LEFFCH tt +1 = DH D' M ? og og and the Luenberger technical change indicator is
- 64 -
Assessment of Productive Efficiency of Airports
LTECHCH tt +1 =
1 ? DL DH D' M D' N ? 1 ? HL NM ? ? + ? = ? + . 2? og og og ? og ? ? og ? 2 ? og ? and
v t t +1 t +1 t +1 Do ( x , y , b ;1,1) < 0,
By
construction, indicates
v t +1 t t t Do ( x , y , b ;1,1) > 0
so
LTECHCH > 0
technical progress. To compute the Malmquist-Luenberger productivity index in (3.17) and Luenberger productivity index in (3.20), including their decomposed components in (3.18), (3.19), (3.21) and (3.22), four distance functions must be estimated,
t t +1 t i.e., Do ( x t , y t , b t ; y t ,?b t ), Do ( x t +1 , y t +1 , b t +1 ; y t +1 ,?b t +1 ), Do ( x t +1 , y t +1 , b t +1 ; y t +1 , b t +1 ) and
t +1 ( x t , y t , b t ; y t ,?b t ) . The latter two are mixed-period distance functions which are Do
obtained by evaluating performance of an airport from one period in another period. Färe, Grosskopf, Norris, and Zhang (1994) make use of the fact that the output distance function is reciprocal to the output-based Farrell measure of technical efficiency, and then modify the directional output distance function Do ( x, y, b; y,?b) in (3.7) to accommodate time period. The computation steps are summarized below.
t t t t t t t t 1) To estimate Do ( xk ' , y k ' , bk ' ; y k ' ,?bk ' ) for airport k ' , compare ( y k ' , bk ' ) to the
?
frontier at time t , and solve the following linear program:
t t t t t t Do ( xk ' , y k ' , bk ' ; y k ' , ?bk ' ) = max ?
s.t.
k?K
?? ??
t k t k
t t y km ? (1 + ? ) y k 'm , m = 1,..........., M , t kj
k?K
?? b
t k
= (1 ? ? )bkt ' j , j = 1,............, J ,
(3.23)
k?K
t t x kn ? xk 'n , n = 1,.............., N ,
?tk ? 0, k = 1,.........., K
- 65 -
Assessment of Productive Efficiency of Airports
t +1 t +1 t +1 t +1 t +1 t +1 t t 2) To estimate Do ( xk ' , y k ' , bk ' ; y k ' ,?bk ' ) for airport k ' , compare ( y k ' , bk ' ) to
the frontier at time t + 1, and solve the following linear program:
t +1 t +1 t +1 t +1 t +1 t +1 ( xk Do ' , y k ' , bk ' ; y k ' , ?bk ' ) = max ?
s.t.
k?K
?? ?? ??
t +1 k
t +1 t +1 ? (1 + ? ) y k y km 'm , m = 1,..........., M ,
k?K
t +1 t +1 k kj
b
1 = (1 ? ? )bkt + ' j , j = 1,............, J , t +1 ? xk 'n , n = 1,.............., N ,
(3.24)
k?K
t +1 t +1 k kn
x
?tk+1 ? 0, k = 1,.........., K
t t +1 t +1 t +1 t +1 t +1 t +1 t +1 3) To estimate Do ( xk ' , y k ' , bk ' ; y k ' ,?bk ' ) for airport k ' , compare ( y k ' , bk ' )
to the frontier at time t , and solve the following linear program:
t t +1 t +1 t +1 t +1 t +1 ( xk Do ' , y k ' , bk ' ; y k ' , ?bk ' ) = max ?
s.t.
k?K
?? ??
t k t k
t +1 t ? (1 + ? ) y k y km 'm , m = 1,..........., M , t kj 1 = (1 ? ? )bkt + ' j , j = 1,............, J ,
k?K
?? b
t k
(3.25)
k?K
t t +1 ? xk x kn 'n , n = 1,.............., N ,
?tk ? 0, k = 1,.........., K
t +1 t t t t t t t 4) To estimate Do ( xk ' , y k ' , bk ' ; y k ' , ?bk ' ) for airport k ' , compare ( y k ' , bk ' ) to the
frontier at time t + 1, and solve the following linear program:
- 66 -
Assessment of Productive Efficiency of Airports
t +1 t t t t t ( xk Do ' , y k ' , bk ' ; y k ' , ?bk ' ) = max ?
s.t.
k?K
?? ?? ??
t +1 k
t +1 t ? (1 + ? ) y k y km 'm , m = 1,..........., M ,
k?K
t +1 t +1 k kj
b
= (1 ? ? )bkt ' j , j = 1,............, J ,
t ? xk 'n , n = 1,.............., N ,
(3.26)
k?K
t +1 t +1 k kn
x
?tk+1 ? 0, k = 1,.........., K
Substituting (3.23) – (3.26) for the corresponding terms in (3.17) – (3.19), the Malmquist-Luenberger productivity index and the two components can be obtained. Similarly, substituting (3.23) – (3.26) for the corresponding terms in (3.20) – (3.22), the Luenberger productivity index and the two components can be obtained. In summary, the Malmquist-Luenberger and Luenberger productivity indexes together with their components provide more insightful information regarding sources of productivity change between two time periods. Chapter 4 will describe the first case study which is summarized from three publications (Pathomsiri and Haghani, 2004; Pathomsiri, Haghani and Schonfeld, 2005; Pathomsiri, Haghani, Dresner and Windle, 2006a). The study is the first attempt to assess productivity of airports operating in multiple airport systems using the DEA model. This case study provides primary understanding on typical results when undesirable outputs are not ignored. Then another case study of U.S. airports will address the shortcomings later.
- 67 -
Assessment of Productive Efficiency of Airports
CHAPTER 4 CASE STUDY 1 PRODUCTIVITY OF AIRPORTS IN MULTIPLE AIRPORT STSYEMS
Many metropolitan regions around the world are served by multiple commercial airports. These regions are called “multiple airport systems” or MASs among aviation community. There have been many stories about functional failures in planning and managing of MASs worldwide due to over-investment or underutilization (Caves and Gosling, 1999; de Neufville, 1995; de Neufville and Odoni, 2003). It is well-documented that several airports in MAS cannot achieve sufficient traffic to economically justify capital investment. In other words, the investment is actually not sufficiently productive. However, it is very surprising that there is no productivity study that focuses on airports in multiple airport systems, though MASs involve several times more capital investment. If one is about to assess two airports comparable in both size and market, an airport in single-airport system which is enjoying its monopolistic status perhaps performs no less efficiently than an airport that is struggling with competitors in an MAS. In a productivity study of mixed samples of airports operating in single-airport and multiple-airport systems, it was found that U.S. airports in MASs were not operating more efficiently than other U.S. airports (Sarkis, 2000). This case study aims to fill this gap by looking at the productivity of airports in MASs. The results may be perceived as like-a-like comparisons among airports operating - 68 -
Assessment of Productive Efficiency of Airports
in the same market structure. Due to the unavailability of data on undesirable outputs at international airports, the study is therefore restricted to the consideration of desirable outputs only. As a result, Data Envelopment Analysis model is applicable. This chapter describes the definition of MAS, modeling of airport operation, input and output measures of airport operations and data collection. Note that the content in this chapter is based on three publications, i.e., Pathomsiri and Haghani (2004), Pathomsiri, Haghani and Schonfeld (2005) and Pathomsiri, Haghani, Dresner and Windle (2006a).
4.1 Definition of multiple airport system
Multiple airport system (MAS) is explicitly defined in few publications (de Neufville, 1995; de Neufville and Odoni, 2003; Hansen and Weidner, 1995). In one textbook (de Neufville and Odoni, 2003), the authors defined an MAS as “the set of significant airports that serve commercial transport in a metropolitan region, without regard to ownership or political control of the individual airports.” This definition involves four important points. First, MAS focuses on airports serving commercial transport. Second, MAS refers to a metropolitan region rather than a city. The region can expand to cover several cities as in the case of New York/New Jersey. Third, MAS focuses on the market, not the ownership of the airports. Although five airports in London area are owned by three different organizations, they form the London MAS since they all serve the same market. Finally, MAS focuses on significant airports. The authors suggest a threshold of more than one million passengers per year for identifying significant airports. Another paper (Hansen and Weidner, 1995) defined an MAS as two or more airports operating with scheduled passengers enplanements in a contiguous metropolitan
- 69 -
Assessment of Productive Efficiency of Airports
area in such a way as to form an integrated airport system and satisfies both of the following criteria: Each airport in the system is included in the same community by the Federal
Aviation Administration (FAA) or within 50 km. (30 miles) of the primary airport of an FAA-designated “large-hub” community, or each airport is in the same Metropolitan Statistical Area (MSA) or Consolidated MSA (CMSA). The Herfindahl Concentration Index (HCI) for the airports is less than 0.95.
HCI is the sum of squared market shares of all airports in an MAS. For example, in 2003, distribution pattern of passenger traffic in Baltimore/Washington MAS was 20,094,756 (39.34%), 16,767,767 (32.83%) and 14,214,803 (27.83%) at
Baltimore/Washington International (BWI), Washington Dulles International (IAD) and Ronald Reagan Washington National (DCA) respectively. Therefore, HCI is equal to 0.39342 + 0.32832 + 0.27832 = 0.34. Similarly, in Houston MAS, in 2002 George Bush Intercontinental (IAH), William P. Hobby (HOU) and Ellington Field (EFD) accommodated 33,905,253 (80.69%), 8,035,727 (19.112%) and 76,035 (0.18%) passengers respectively. HCI is equal to 80.692 + 19.122 + 0.182 = 0.688. By the above criteria, the Houston MAS is not so concentrated but somehow competitive. Note that for a single airport system, HCI is 1.0. For an MAS where traffic is evenly divided among N airports, HCI is 1/N (Baltimore/Washington MAS may be a close example). The above two examples of MAS definitions indeed are very similar. The difference may be the significance of the airport. One uses a threshold of passenger traffic to identify the MAS whereas the other uses HCI, regardless of traffic volume. It is
- 70 -
Assessment of Productive Efficiency of Airports
harmless to think of the latter criterion as a measure of sufficient significance in a sense that only competitive MASs are included in the analysis. Since definitions are in good agreement and well-documented in the publications, 14 MASs in the U.S. from (Hansen and Weidner, 1995) are then adopted. For more comprehensive study, the scope is expanded to other MASs worldwide. In this study, the lists of MASs in publications (Caves and Gosling, 1999; de Neufville, 1995; de Neufville and Odoni, 2003) were collected and checked if they satisfy both definitions. Eventually, we identified another 11 non-US MASs. Totally, there are 25 MASs in this study, involving 75 airports in four continents, i.e., North America, South America, Europe and Asia. Table 4.1 provides the list of all 25 MASs together with the airports in the systems along with the International Civil Aviation Organization (ICAO) airport codes. The Table shows two computed HCIs, one based on passenger and the other based on aircraft movements. This means that airports in the same region may compete for passengers or aircraft movements or both. As a result, as long as either HCI is below 0.95, the region is an MAS.
- 71 -
Assessment of Productive Efficiency of Airports
Table 4.1 List of multiple airport systems and corresponding Herfindahl Concentration Indices, 2002
Region 1 2 Chicago, IL, USA New York City, NY, USA Airport Code ORD MDW CGX EWR JFK LGA ISP HPN SWF LAX SNA ONT BUR PSP LGB OXR PMD SFO SJC OAK STS CCR SBA SMX DFW DAL IAH HOU EFD BWI IAD DCA MIA FLL PBI PNS VPS Airport Name O'Hare International Midway International Merrill C. Meigs Newark Liberty International, NJ John F. Kennedy International, NY LaGuardia, NY Long Island MacArthur, NY Westchester County, NY Stewart International, NY Los Angeles International, CA John Wayne, CA Ontario International, CA Bob Hope, CA Palm Spring International, CA Long Beach, CA Oxnard, CA Palmdale Regional, CA San Francisco International, CA Mineta San Jose International, CA Oakland International, CA Sonoma County, CA Buchanan Field, CA Santa Barbara, CA Santa Maria Public, CA Dallas/Fort Worth International, TX Dallas Love Field, TX George Bush Intercontinental, TX William P. Hobby, TX Ellington Field, TX Baltimore/Washington International, MD Washington Dulles International, VA Ronald Reagan Washington National, DC Miami International, FL Fort Lauderdale - Hollywood International, FL Palm Beach International, FL Pensacola Regional, FL Okaloosa Regional, FL Aircraft Movements 922,817 304,304 31,972 405,562 287,606 362,439 223,063 167,776 123,642 645,424 368,627 149,292 162,211 85,243 350,603 88,027 33,352 351,453 207,510 371,988 114,854 142,329 159,835 76,426 765,109 245,564 456,831 246,230 102,016 304,921 372,636 215,691 446,235 280,737 166,908 130,826 118,423 HCI-Air 0.596 0.191 Total Passengers 66,565,952 17,371,036 86,483 29,202,654 29,943,084 21,986,679 1,890,580 930,097 362,017 56,223,843 7,903,066 6,517,050 4,620,683 1,108,695 1,453,412 45,306 226 31,456,422 11,115,778 13,005,642 3,598 0 728,307 58,104 52,828,573 5,622,754 33,905,253 8,035,727 76,035 19,012,529 17,075,965 12,871,885 30,060,241 17,037,261 5,483,662 1,345,970 631,592 HCI-PAX 0.670 0.315
3
Los Angeles, CA, USA
0.209
0.543
4
San Francisco, CA, USA
0.240
0.415
5 6 7 8 9 10
Santa Barbara, CA, USA Dallas/Fort Worth, TX, USA Houston, TX, USA Washington, DC, USA Miami, FL, USA Pensacola, FL, USA
0.562 0.632 0.432 0.349 0.383 0.501
0.863 0.826 0.688 0.342 0.443 0.565
- 72 -
Assessment of Productive Efficiency of Airports
Table 4.1 List of multiple airport systems and corresponding Herfindahl Concentration Indices, 2002 (Continued)
Region 11 12 13 14 15 16 17 18 Detroit, MI, USA Cleveland, OH, USA Norfolk, VA, USA Oshkosh/Appleton, WI, USA Montreal, Canada Rio de Janeiro, Brazil Sao Paulo, Brazil Buenos Aires, Argentina Airport Code DTW DET CLE CAK ORF PHF ATW OSH YUL YMX GIG SDU GRU CGH AEP EZE DOT SFD LHR LTN LGW STN LCY GLA EDI PIK CDG ORY TXL SXF THF LIN MXP SVO VKO DME HND NRT Airport Name Detroit Metropolitan Wayne County, MI Detroit City, MI Cleveland Hopkins International, OH Arkon-Canton, OH Norfolk International, VA Newport News/Williamsburg International, VA Outagamie County Regional, WI Wittman Regional, WI Montreal-Dorval International Montreal-Mirabel International Rio De Janeiro-Galeao International Santos Dumont Sao Paulo Guarulhos International Congonhas Aeroparque Jorge Newbery Ezeiza International International Don Torcuato San Fernando London Heathrow London Luton London Gatwick London Stansted London City Glasgow Edinburgh Glasgow Prestwick International Roissy-Charles-de Gaulle Orly Tegel Schoenefeld Tempelhof Linate Malpensa Sheremetyevo Vnukovo Domodedovo Tokyo International (Haneda) New Tokyo International (Narita) Aircraft Movements 490,885 69,066 251,758 119,958 125,622 228,504 57,755 115,288 192,225 32,977 83,731 117,144 160,451 266,231 91,350 50,755 23,392 34,819 466,554 80,921 242,380 170,774 56,102 105,197 118,419 43,346 510,098 211,080 127,470 37,389 48,026 110,494 214,886 124,630 65,759 84,102 282,674 164,270 HCI-Air 0.784 0.563 0.542 0.555 0.750 0.514 0.531 0.316 Total Passengers 32,477,694 0 10,455,204 894,798 3,464,246 515,056 491,744 3,912 7,816,052 990,937 5,810,868 5,626,328 12,804,091 12,562,319 4,519,424 4,054,473 23,148 11,676 63,338,641 6,496,258 29,628,423 16,049,288 1,604,773 7,807,060 6,932,106 1,487,113 48,350,172 23,169,725 9,879,888 1,688,028 612,867 7,815,316 17,441,250 10,895,225 3,120,210 6,683,268 61,079,478 28,883,606 HCI-PAX 1.000 0.855 0.775 0.984 0.800 0.500 0.500 0.497
19
London, United Kingdom
0.305
0.379
20 21 22 23 24 25
Glasgow, United Kingdom Paris, France Berlin, Germany Milan, Italy Moscow, Russia Tokyo, Japan
0.378 0.586 0.440 0.551 0.357 0.535
0.422 0.562 0.680 0.573 0.404 0.564
- 73 -
Assessment of Productive Efficiency of Airports
4.2
Modeling airport operations
Each airport in the MAS is viewed as a production unit. As a result, airport operation can be modeled as a production process that requires some inputs for running day-to-day operations in order to produce some target outputs. Necessary inputs include production factors such as capital and labor. Most airport managers set target to maximize movement of aircrafts, passenger throughput and quantity of cargo transported. These outputs are highly desirable and the primary reason for building an airport. Due to the unavailability of data on undesirable outputs (e.g., delays, noise) at international airports, the assessment has to be restricted to the consideration of desirable outputs only. As a result, Data Envelopment Analysis (DEA) is applicable as an analytical tool.
4.3 Input and output measures of airport operations
The formulation of any DEA model given in Table 2.5 requires data on quantity of inputs and desirable outputs, ( x kn , y km ) for individual airports. The selection of inputs and outputs is an important decision issue in the assessment of airport productivity. The general suggestion is to include all important measures that are in the interest of the management. Such measures should be common for all airports so that the performance would provide meaningful interpretation. In practice, the main problem is the availability of the data across all airports rather than model limitations. After all, three common physical inputs are considered in this analysis:
x1 = Land area (LAND), acre x 2 = Number of runways (RW)
x3 = Runway area (RWA), acre - 74 -
Assessment of Productive Efficiency of Airports
These inputs are necessary infrastructure for all airports. Land area (acre) represents a considerable share of capital investment that an airport should fully utilize. Number of runways counts all existing runways at the airport, regardless of their utilization level. Runway area is the summation of product between length and width of all runways. The consideration of runway area should explains variations in productivity better than using the number of runway alone, since it takes into account the effect of size and design configuration such as length, width, and separation. Other inputs such as terminal area, number of gates, number of employees and expense cannot be included due to the lack of complete data across samples. For the set of desirable outputs, it is assumed that airport managers aim at producing the following two outputs as much as possible:
y1 = Aircraft movements y 2 = Passengers
Number of aircraft movements includes all kinds of movements, i.e., commercial aircrafts, cargo aircrafts, general aviation, and others. The number of passengers counts both arriving and departing passengers for all type of commercial passengers, i.e., international, domestic and direct transit passengers. Other desirable outputs such as cargo throughput and revenues cannot be considered due to the lack of complete data across all samples, especially for small U.S. and non-U.S. airports. Inclusion of these outputs will reduce sample size drastically; hence, it is decided to maintain all samples. Note that these input and output measures may be rather limited to partial factors of airport operations, but they have been used in previous studies such as Gillen and Lall (1997, 1998); Pels, Nijkamp and Rietveld
- 75 -
Assessment of Productive Efficiency of Airports
(2001,2003) (see Table 2.4 for details). Given the above-selected three inputs and three desirable outputs, the assessment may be perceived as the measurement of productive efficiency of airside operation.
4.4 Data collection
The study period is during 2000 – 2002. On the output side, statistics on the number of passengers and aircraft movements are collected from Airports Council International publications (ACI, 2002 - 2004). The missing data are supplemented from several sources such as FAA website (FAA, 2004a), airports’ official websites, airport newsletters, reports, airport contacts and e-mail correspondences. Collecting input data caused more trouble since there is no single source available at hand. Airport Master Record (AMS) database (FAA, 2004b) contains the most recent data on characteristics of US airports. The best effort was made to verify this recent data with airport managers whether there are runway expansions or constructions during the period 2000 – 2002. Some airports had improved their runways. For example, George Bush Intercontinental (IAH) expanded and extended runway 15R/33L to 10000’ x 150’ in 2002. Detroit Metropolitan Wayne County (DTW) opened its 6th runway on December 11, 2001. The number of runways is edited accordingly. The number of runway and runway acreage are computed precisely by the time it is in service during the year, rounding down in month. In case of DTW, for example, it is concluded that it had 5 and 6 runways in 2001 and 2002, respectively. Input data of non-US airports are more difficult to collect since there is no database such as AMS (FAA, 2004b). Inevitably, one has to rely on information from airports’ official websites, airport newsletters, reports, airport contacts and e-mail correspondences. Also, it is - 76 -
Assessment of Productive Efficiency of Airports
verified with airport managers whether there was any change in runway configuration during 2000 – 2002. Similarly to the US airport case, it is found that Narita International (NRT) opened its new second parallel runway in April 2002. As a result, it used 1.667 (= 1 + 8/12) runways in 2002. Ultimately, the study had to drop Santos Dumont/Rio de Janeiro (SDU), International Don Torcuato/Buenos Aires (DOT) and Vnukovo/Moscow (VKO) airports from the sample due to unavailable land area data. The final dataset used in this study contains 72 airports with complete input and output data. The sample size is relatively larger than most previous studies (Abbott and Wu, 2002; Adler and Berechman, 2001; Bazargan and Vasigh, 2003; Fernandes and Pacheco, 2001, 2002, 2005; Gillen and Lall, 1997, 1998; Hooper and Hensher, 1997; Martin and Roman, 2001; Martin-Cejas (2002); Murillo-Melchor, 1999; Nyshadham and Rao, 2000; Oum, Yu and Fu, 2003; Pacheco and Fernandes 2003; Parker, 1999; Pels, Nijkamp and Rietveld, 2001, 2003; Salazar de la Cruz,1999; Sarkis, 2000; Sarkis and Tulluri, 2004; Yoshida, 2004; Yu, 2004). The number of samples was checked against several applicable rules of thumb to guarantee the sufficiency and meaningful interpretation. In DEA applications, one frequent problem is a lack of discriminatory power between DMUs as a result of an excessive number of measures with respect to the total number of DMUs. The larger the number of input and output measures for a given number of airports the less discriminatory the DEA model becomes. Given a certain set of samples, this means that the addition of measures will reduce the discriminatory power of the DEA model. Essentially, this is because it is possible that an airport may dominate all others on one measure, which in turn makes it look equally efficient compared to other efficient airports. This is a major issue encountered by Parker (1999),
- 77 -
Assessment of Productive Efficiency of Airports
Adler and Berechman (2001) and Yu (2004). To avoid this problem, the straightforward way is to guarantee that there will be a sufficient number of airports for comparison, regarding any of the measures. Boussofiane, Dyson and Thanassoulis (1991) recommend that the total number of DMUs be much greater than the number of inputs times the number of outputs. Compared to this analysis where three inputs and two outputs are selected for the assessment of airport productivity; the number of samples needs to be much more than 3 x 2 or 6 airports in order to reduce the chance that an airport is too dominant compared to the others on a particular measure. According to the recommendation, the sample size of 72 airports is deemed satisfactory. To avoid losing discriminatory power, Cooper, Seiford and Tone (2000: page 103) recommend that the desired number of DMUs exceed m + s several times. They suggest a more stringent rule of thumb in the following formula (Cooper, Seiford and Tone, 2000: page 252). n ? max{m x s, 3(m + s )} (4.1)
where n is the number of DMUs, m and s are the numbers of input and output measures respectively. Substituting m and s , yields the minimum number of samples: n ? max{3 x 2, 3(3 + 2)} = max{6,15} = 15 (4.2)
Again, the sample size of 72 satisfies this recommendation. After all, it can be concluded that the sample size is sufficient for the analysis. Table 4.2 summarizes the descriptive statistics of the samples. Input measures are rather stable over time; only slight - 78 -
Assessment of Productive Efficiency of Airports
changes in number of runways and runway acreage are recorded. The variation in terms of the range between maximum and minimum is wide, indicating that airports in MAS are much different in scale of operation. Similarly, airports’ outputs are widely variable indicating that airports are much different in scope of operation. Six partial productivity ratios are also shown to provide more information on the utilization of airport. In the next chapter, results from assessing productivity of 72 airports by DEA model will be presented. The assessment will be discussed with respect to operational efficiency of individual airports. In addition, Censored Tobit regression models are also estimated for explaining variations in total productivity level.
- 79 -
Assessment of Productive Efficiency of Airports
Table 4.2 Descriptive statistics of 72 airports in MASs, 2000 - 2002
Variables Land acreage area (LAND) Statistics Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. Min Max Mean S.D. 2000 70 18,076 2,865 3,419 1 7 2.69 1.26 9.78 295.54 79.56 52.84 30,479 908,989 234,800 188,593 0 72,144,244 14,063,577 18,120,258 0 37,316 6,561 7,678 2001 70 18,076 2,865 3,419 1 7 2.69 1.26 9.78 295.54 79.66 52.81 31,240 911,917 223,829 179,799 0 67,448,064 13,363,106 16,989,050 0 32,254 6,379 7,449 2002 70 18,076 2,865 3,419 1 7 2.72 1.28 9.78 298.60 80.98 54.41 31,972 922,817 216,093 172,890 0 66,565,952 13,240,016 16,786,913 0 32,333 6,436 7,662 2 736 157 156 16,489 184,314 76,211 40,772 299 14,983 3,021 2,052 0 21,112,880 4,502,157 5,049,647 0 572,148 142,365 141,131
Number of runways (RW)
Runway acreage area (RWA)
Annual aircraft operations (AIR)
Annual total passengers (PAX)
PAX/LAND
AIR/LAND
2 774 173 174
19,957 195,858 83,260 45,021 362 15,765 3,332 2,257 0 27,389,915 4,861,631 5,665,696 0 583,604 151,358 148,524
2 757 163 162
18,088 189,452 79,783 43,090 328 15,401 3,166 2,135 0 25,379,370 4,657,794 5,379,152 0 548,702 145,135 141,262
AIR/RW
AIR/RWA
PAX/RW
PAX/RWA
- 80 -
Assessment of Productive Efficiency of Airports
CHAPTER 5 CASE STUDY 1 RESULTS AND DISCUSSION
5.1 Selection of a DEA model
Similar to Martin and Roman (2001), it is assumed that once an airport has invested in the infrastructure, it is difficult for managers to disinvest to save costs. Consequently, airport managers are more interested to know the probable levels of outputs, given the existing infrastructure (Fernandes and Pacheco, 2002). From this viewpoint, the outputorientation DEA model is preferred. Since the analysis is focused on rather narrow study period during 2000 – 2002, the variable return to scale (VRS) frontier type is chosen to reflect the short-term view (Ganley and Cubbin, 1992). In VRS frontier type DMUs are not penalized for operating at a non-optimal scale (Banker, 1984; Banker and Thrall, 1992). After all, the applicable model is the DEA-Output-VRS. Its mathematical formulation is given below: max ? s.t.
k?K
?? ?? ??
k
+ y km ? s m = ?y km , m = 1,..........., M , ? x kn + s n = x kn , n = 1,.............., N ,
(5.1)
k?K
k
k?K
k
= 1, k = 1,.........., K
?k ? 0
For each year during 2000 – 2002, the DEA model in (5.1) is solved 72 times; i.e., one time for each airport, to determine the optimal efficiency scores ? * . ? * measures the - 81 -
Assessment of Productive Efficiency of Airports
level of inefficiency. An efficient airport will have ? * = 1 which means it does not need to increase its outputs. It is already on the efficient production frontier. In other words, the airport is more productive than others, given the same amount of inputs. All inefficient airports will have ? * > 1. The higher value of ? * shows greater inefficiencies.
5.2 Efficient scores
Table 5.1 presents the efficiency scores of 72 airports during 2000 – 2002. Bold typeface highlights airports on efficient production frontier. For example, in 2002, there are 12 efficient airports i.e. O’Hare International (ORD), Merrill C. Meigs (CGX) 1, LaGuardia (LGA), Los Angeles International (LAX), John Wayne (SNA), Oxnard (OXR), Palmdale (PMD), Congonhas/Sao Paulo (CGH), Heathrow/London (LHR), Stansted/London (STN), City/London (LCY), and Haneda/Tokyo (HND). These 12 airports form a piece-wise linear efficient production frontier under variable return-to-scale assumption. Seemingly, efficient airports can be classified into two groups i.e., the busy and the compact. The busy group is usually a primary or major airport in the region such as O’Hare International (ORD), Los Angeles International (LAX), Aeroparque Jorge Newbery/Buenos Aires (AEP), Heathrow/London (LHR), Haneda/Tokyo (HND) and Narita/Tokyo (NRT). Their land areas are relatively large, ranging from 2,000 – 8,000 acres. Annual passenger traffics are consistently among the top of the world. Another interesting observation is that they dominate the market by having more than 50% of passengers. Their high traffic flows enable airports to operate more efficiently at higher utilization level than others. The compact group includes Merrill C. Meigs (CGX), LaGuardia (LGA), John Wayne (SNA), Oxnard (OXR), Palmdale (PMD), Congonhas/Sao Paulo (CGH),
1
The airport was permanently closed in March 2003.
- 82 -
Assessment of Productive Efficiency of Airports
Stansted/London (STN) and City/London (LCY). These airports are relatively small in size (between 70 – 960 acres) with one or two runways. They are alternative airports, except Congonhas/Sao Paulo (CGH) which is a primary airport constrained in downtown area. Although traffic may not be so high, they still can be efficient airports because of their sufficiently high utilization rate. Airports with efficiency scores below two may be considered satisfactorily efficient in terms of input utilization. The airports with scores consistently higher than two should be monitored closely for improving efficiency. Some airports with consistently very high scores such as Montreal-Mirabel (YMX), Glasgow Prestwick International (GLA),
Schoenefeld/Berlin (SXF), and Tempelhof/Berlin (THF) are significantly under-utilized or over-invested. These airports tend to use comparable inputs to others but service far fewer aircraft movements and passengers.
- 83 -
Assessment of Productive Efficiency of Airports
Table 5.1 Efficiency scores, 2000 – 2002
No.
1
Multiple Airport System
Chicago, IL, USA
2
New York City, NY, USA
3
Los Angeles, CA, USA
4
San Francisco, CA, USA
5 6 7 8 9 10 11 12
Santa Barbara, CA, USA Dallas/Fort Worth, TX, USA Houston, TX, USA Washington, DC, USA Miami, FL, USA Pensacola, FL, USA Detroit, MI, USA Cleveland, OH, USA
Airport Code ORD MDW CGX EWR JFK LGA ISP HPN SWF LAX SNA ONT BUR PSP LGB OXR PMD SFO SJC OAK STS CCR SBA SMX DFW DAL IAH HOU EFD BWI IAD DCA MIA FLL PBI PNS VPS DTW DET CLE CAK - 84 -
2000
1.000 1.302 1.000 1.251 2.041 1.000 2.166 2.067 2.842 1.000 1.000 2.479 2.407 4.670 1.259 1.000 1.000 1.640 1.596 1.230 2.828 1.773 2.547 5.198 1.085 1.931 1.396 1.951 5.813 2.050 1.283 1.414 1.132 1.720 2.490 3.293 3.285 1.486 1.214 1.739 4.117
2001
1.000 1.293 1.000 1.251 2.119 1.000 2.058 2.049 3.330 1.000 1.000 2.399 2.354 4.510 1.285 1.000 1.000 1.786 1.691 1.335 2.730 2.404 2.578 4.982 1.164 1.956 1.352 1.916 6.240 1.902 1.407 1.637 1.180 1.615 2.460 3.252 3.203 1.525 1.846 1.888 3.953
2002
1.000 1.174 1.000 1.234 2.076 1.000 1.996 2.197 2.981 1.000 1.000 2.446 2.266 4.324 1.231 1.000 1.000 1.816 1.977 1.398 3.210 2.558 2.527 4.823 1.206 1.809 1.375 1.806 4.710 1.940 1.361 1.779 1.134 1.552 2.732 2.818 3.113 1.692 2.664 1.990 4.014
Assessment of Productive Efficiency of Airports
Table 5.1 Efficiency scores, 2000 – 2002 (Continued)
No. 13 14 15 16 17
18
Multiple Airport System Norfolk, VA, USA Oshkosh/Appleton, WI, USA Montreal, Canada Rio de Janeiro, Brazil Sao Paulo, Brazil
Buenos Aires, Argentina
19
London, United Kingdom
20 21 22 23 24
25
Glasgow, United Kingdom Paris, France Berlin, Germany Milan, Italy Moscow, Russia
Tokyo, Japan
Airport Code ORF PHF ATW OSH YUL YMX GIG GRU CGH AEP EZE SFD LHR LTN LGW STN LCY GLA EDI PIK CDG ORY TXL SXF THF LIN MXP SVO DME HND NRT
2000 3.107 1.844 5.871 4.478 2.792 9.679 4.592 2.016 1.082 1.000 5.630 2.810 1.000 1.191 1.113 1.000 1.000 2.899 3.756 8.583 1.395 2.192 2.798 8.022 7.784 3.260 1.480 2.887 8.025 1.070 1.000
2001 3.161 1.734 6.642 4.596 2.846 10.306 4.021 1.976 1.000 1.000 5.910 2.681 1.000 1.275 1.080 1.000 1.000 2.402 3.272 8.300 1.290 2.304 2.687 9.144 7.744 2.458 1.520 2.676 5.809 1.000 1.000
2002 2.927 1.613 6.383 3.894 2.638 11.137 4.249 2.170 1.000 1.260 6.858 2.455 1.000 1.490 1.199 1.000 1.000 2.270 3.063 8.457 1.221 2.286 2.744 9.759 7.676 2.367 1.615 2.750 4.142 1.000 1.101
Note: Bold typeface highlights efficient airports.
- 85 -
Assessment of Productive Efficiency of Airports
5.3
Determination of airport productivity
It may not be sufficient to just describe efficiency score based on historical data. For planning an airport, the understanding of factors affecting efficiency score is even more useful. In this case study, causal models for explaining the variations in efficiency score are estimated so that an airport manager can predict future productivity based on given information. In particular, this information is treated as usual exploratory variables. The information may include number of runways, land area, number of gates, noise strategies, proportion of General Aviation (GA) traffic, proportion of international passengers, type of ownership/management, etc. The dependent variable is the efficiency score that indicates the total productivity level of an airport. By the nature of the DEA-Output-VRS model, the value of efficiency scores can only be in the range of 1 to infinity. Because of this special type of limited dependent variable, simple regression is not an appropriate model. Its underlying assumptions are violated, causing inconsistency in estimated coefficients. The Censored Tobit regression model (Tobin, 1958; Maddala, 1983; Amemiya, 1984; Gillen and Lall, 1997; Greene, 2002; Greene, 2003) is more appropriate. In this case, efficiency score of airport y i is represented by Equation (5.2).
?? xi + ? i yi = ? ?1 if y i > 1 if y i ? 1
(5.2)
y i is an efficiency score that is observable for values greater than 1 and is censored for values less than or equal to 1. Efficiency scores of all efficient airports are
- 86 -
Assessment of Productive Efficiency of Airports
censored at 1, regardless of values of independent variables x i . ? and ? i are the coefficients and the error term of the Tobit model respectively. The coefficients ? can be estimated with the Maximum Likelihood (ML) method. ML estimation for the Tobit model involves dividing the observations into two sets. The first set contains uncensored observations. The second set contains censored observations. For y i > 1 , assuming y i ~ N ( ? , ? 2 ) , then the log-likelihood function is written as shown in Equation (5.3).
ln L =
Uncensored
?
ln
1 ? y i ? ? xi ? ? 1 ? ?xi ? ?? ? + ? ln ?? ? ? ? ? ? Censored ? ? ?
(5.3)
where ? and ? are the respective probability and cumulative density functions. Unlike simple regression models, the estimated coefficients cannot be interpreted as marginal effects. Equation (5.4) is used to compute marginal effect of variable k (Gillen and Lall, 1997).
?x ? 1 ?E ( y | x) = Prob(Uncensored | x) ? k = ? ( )? k ? ?x k
(5.4)
Goodness-of-fit may be measured by using R 2 ANOVA , computed by Equation (5.5) . This fit measure takes the variance of the estimated conditional mean divided by the variance of the observed variable (Greene, 2002).
1 n ? ' ?' ? ? ? yi ? y ? ? n i =1 ? ? ?
? 1 n ? ? y y ? ? ? ? i n i =1 ? ? 2
R2 ANOVA =
2
=
Var(Predicted conditional mean) Var(Dependent variable)
(5.5)
- 87 -
Assessment of Productive Efficiency of Airports
Some econometric packages can be used to estimate Censored Tobit Regression model such as EViews (QMS, 2005), LIMDEP (Greene, 2002), and STATA (StataCorp, 2005). In this case study, LIMDEP version 8.0 (Greene, 2002) is used.
5.4 Factors affecting productive efficiency of airports in MASs
The Tobit model has efficiency score ? as the dependent variable. Related literature suggests many possible exploratory variables qualify as independent variables (Gillen and Lall, 1997). In this case study, five groups of independent variables are investigated. The proxy of each group entering the model is essentially based on data availability. First, Airport characteristics are represented here by physical characteristics, basically input measures that are used in the DEA model, i.e., land area (LAND), number of runway (RW) and runway area (RWA). These inputs certainly play a major role in accommodating traffic. However, one should be aware that having more of these inputs does not necessarily mean more outputs. Second, Airport services are mainly represented by outputs of airport operations which consist of number of aircraft movements (AIR) and passengers (PAX). One would expect that more services contribute to higher efficiency. However, this is not necessarily true since efficiency takes into account both inputs and outputs. Accordingly, another group of variables is introduced, i.e., level of utilization. Third, Level of utilization may be a better determinant of operational efficiency since it takes into accounts both input and output measures. This case study considers
- 88 -
Assessment of Productive Efficiency of Airports
seven ratio variables, i.e. annual total passengers/land area (PAX/LAND), annual aircraft movements/land area (AIR/LAND), annual aircraft movements/runway (AIR/RW), annual aircraft movements/runway acreage area (AIR/RWA), annual total
passengers/number of runways (PAX/RW), annual total passengers/runway acreage area (PAX/RWA), and annual total passengers/annual aircraft movements (PAX/AIR). Intuitively, higher values of these ratios should result in more efficient operation. However the interpretation must be very careful since excessive utilization may imply undesirable congestion and delay. Whenever congested airports are classified as efficient in the results they should not be considered appropriate benchmarks. Instead, other less efficient ( ? near 1) may provide more practical benchmarks. Fourth, Market characteristics include target market (e.g., passengers, aircraft operation, cargo, general aviation and military service), market share, market focus (e.g., domestic, international, tourist, business passengers), and irregularity of time periods. Although such characteristics would be interesting to analyze, collecting them for complete cross-national sample is prohibitively expensive. For example, an attempt was made to collect the percentage of general aviation (GA) operations at airports since serving more GA operations tends to lower operational efficiency (Gillen and Lall, 1997). However, such data are not available for many airports. Similarly, the data are not available for other potential variables. To avoid discarding many airports from the analysis, the entering variables have to be limited to available data. Consequently, three variables, namely the percentage of international passengers (INTER), Y2001 and Y2002 are entered the estimation. It is unclear how this proportion affects airport efficiency.
- 89 -
Assessment of Productive Efficiency of Airports
Y2001 and Y2002 aim at capturing anomalies occurring during these two years, notably the 9-11 terrorist attacks. Fifth, Ownership/management characteristics may be another factor affecting airport efficiency. In particular, the study is interested in two contrasting types of ownership/management, namely publicly-owned and privately-owned. There are some good reasons to argue that the latter type yields higher efficiency. For example, a privately-owned entity faces higher risks. This is likely the case when there is little or no subsidy from public funds. This variable is coded as dummy variable equal to one when the airport is privately-owned or there is strong evidence that it is behaving as a commercialized profit-seeking entity. In the US, Stewart International (SWF) is one such example. It has been privatized to the National Express Group Plc. in 1998 (Steward International Airport, 2005). In London MAS, most airports are of this type. Heathrow (LHR), Gatwick (LGW) and Stansted (STN) have been privatized since 1987 and managed by BAA Plc. City/London (LCY) is owned by an Irish entrepreneur, Dermot Desmond (London City, 2005). Table 5.2 compares statistics of some candidate variables between efficient (efficiency score = 1) and inefficient airports (efficiency score > 1). It seems clear that an efficient airport uses less input to produce more output, which can be confirmed by its higher utilization variables. However, it is unclear how the proportion of international passengers associates with performance score. Both groups seem to have comparable figures around 20 - 30%. On the management style, privately-operated airports dominate
- 90 -
Assessment of Productive Efficiency of Airports
in the efficient group. For example, in 2002, five out of seven efficient airports are managed commercially. Table 5.3 shows a Censored Tobit regression model estimation results. The notation is given at the bottom of the Table. This model is a preliminary estimation. It has 11 independent variables, including the constant. Other variables are dropped off for reasons such as high correlation among themselves, being insignificant or having illogical sign. Keep in mind that the lower efficiency score is desirable because it indicates that an airport is more efficient. As a result, a negative sign of estimated coefficient, such as 0.0173 of passengers per aircraft movement (PAX/AIR), contributes to higher efficiency. In this model, passenger market share (PAXSHARE) has an illogical sign. The expected sign should be negative since higher share are more likely associated with higher efficiency. Year 2001 for which it was aimed to test whether September-11 terrorist attack had any effect on efficiency scores; turns out to be insignificant. Possibly, it did not have immediate effect in that year but propagated to the next year 2002, as Y2002 variable is significant. This means that in 2002, an airport became efficient slightly easier than normal. The marginal effects are also shown next to the right of the coefficient’s column. Basically, it indicates the change in efficiency score with respect to unit change of an independent variable. For example, if an airport were able to increase its share of aircraft movements by one percent, it would earn 1.9656 additional units on its efficiency score.
- 91 -
Assessment of Productive Efficiency of Airports
Table 5.2 Comparisons of statistics between efficient and inefficient airports
Variables Land acreage area (LAND) Number of runways (RW) Runway acreage area (RWA) Annual aircraft operations (AIR) Annual total passengers (PAX) AIR/LAND AIR/RW AIR/RWA PAX/LAND PAX/RW PAX/RWA PAX/AIR % of international passenger (INTER) Management style (MANAGE) 2000 Efficient Inefficient (N = 12) (N = 60) 2,261 2,986 (2,896) (3,523) 2.17 2.80 (1.80) (1.12) 67.13 82.05 (68.23) (49.54) 299,963 221,768 (294,845) (159,866) 23,772,828 12,121,727 (28,243,030) (14,946,553) 348 138 (252) (130) 126,123 74,688 (61,830) (35,748) 5,505 2,897 (3,833) (1,496) 12,235 5,426 (11,045) (6,351) 9,442,790 3,945,399 (9,057,704) (4,267,402) 290,332 123,563 (203,776) (118,702) 62.31 45.64 (60.97) (40.42) 33.26 22.81 (39.73) (30.15) 6/6 51/9 2001 Efficient Inefficient (N = 14) (N = 58) 2,161 3,035 (2,714) (3,567) 2.21 2.81 (1.67) (1.13) 68.66 82.32 (64.70) (49.82) 287,397 208,485 (267,031) (150,907) 23,992,367 10,797,422 (26,255,156) (12,955,868) 340 121 (248) (96) 119,697 70,148 (55,985) (33,350) 5,186 2,678 (3,559) (1,242) 13,598 4,637 (10,672) (5,220) 9,399,439 3,513,259 (8,397,929) (3,619,718) 289,295 110,337 (184,177) (103,842) 70.12 42.80 (68.16) (33.24) 27.97 23.88 (37.28) (30.59) 8/6 49/9 2002 Efficient Inefficient (N = 12) (N = 60) 1,615 3,115 (2,163) (3,579) 2.42 2.78 (1.73) (1.18) 73.43 82.49 (68.60) (51.68) 307,916 197,728 (271,078) (142,169) 25,620,505 10,763,918 (27,669,409) (12,608,237) 380 112 (223) (89) 117,094 68,035 (58,651) (30,842) 5,330 2,559 (3,665) (1,123) 15,287 4,666 (11,095) (5,346) 8,670,840 3,668,420 (7,874,927) (3,857,687) 289,748 112,888 (204,542) (104,246) 63.85 46.23 (63.97) (37.68) 24.39 25.09 (36.06) (31.30) 7/5 50/10
Note: N is the number of airports. Standard deviations are shown in parentheses. For Management style, 7/5 in 2002 represents number of noncommercial (7) and commercial airports (5) respectively.
- 92 -
Assessment of Productive Efficiency of Airports
Table 5.3 Censored Tobit regression: preliminary model estimation results
Variables Constant MANAGE PAX/AIR INTER AIRSHARE PAXSHARE Y2001 Y2002 PAX/RWA (million/acre) AIR/AREA (1,000/acre) AIR/RWA (1,000/acre)
Preliminary model 2 R ANOVA = 0.6236 Coefficient 6.8560** (17.901) -0.5964* (-2.409) -0.0173** (-2.917) 0.0236** (6.186) -2.3715** (-3.343) 0.2955 (0.583) -0.3108 (-1.451) -0.4226* (-1.973) -4.4344* (-2.260) -6.0155** (-6.731) -0.5530** (-5.237) Marginal effect 5.6827** (15.802) -0.4943* (-2.419) -0.0144** (-2.899) 0.0195** (6.040) -1.9656** (-3.327) 0.2450 (0.583) -0.2576 (-1.452) -0.3503* (-1.974) -3.6755* (-2.265) -4.9861** (-6.738) -0.4584** (-5.255)
Number of airports = 216 during 2000 - 2002
Notation: Dependent variable = Efficiency score MANAGE = 1 if privately-owned or commercially managed, otherwise = 0 PAX/AIR = Average number of passengers per aircraft movement INTER = Percentage of international passenger (%) AIRSHARE = Market share of annual aircraft movements PAXSHARE = Market share of annual total passengers Y2001 = 1 if compute performance score in year 2001, otherwise = 0 Y2002 = 1 if compute performance score in year 2002, otherwise = 0 PAX/RWA = Annual total passengers per runway area (million/acre) AIR/AREA = Annual aircraft movements per land area (103/acre) AIR/RWA = Annual aircraft movements per runway area (103/acre) ** Estimated coefficient is significant at the 0.01 level (one-tailed) * Estimated coefficient is significant at the 0.05 level (one-tailed)
- 93 -
Assessment of Productive Efficiency of Airports
Table 5.4 Censored Tobit regression: proposed model estimation results
Variables Constant MANAGE PAX/AIR INTER AIR50 PAX50 Y2001 Y2002 PAX/RWA (million/acre) AIR/AREA (1,000/acre) AIR/RWA (1,000/acre)
Proposed model 2 R ANOVA = 0.6206 Coefficient 6.5319** (18.616) -0.6494** (-2.637) -0.0169** (-2.899) 0.0241** (6.424) -0.7025** (-2.727) -0.5858* (-2.405) -0.3374 (-1.593) -0.4634* (-2.184) -4.5793* (-2.349) -5.6768** (-6.551) -0.5760** (-5.515) Marginal effect 5.4222** (16.326) -0.5391** (-2.651) -0.0140** (-2.880) 0.0200** (6.259) -0.5831** (-2.716) -0.4863* (-2.411) -0.2800 (-1.594) -0.3846* (-2.186) -3.8013* (-2.354) -4.7123** (-6.575) -0.4781** (-5.532)
Number of airports = 216 during 2000 - 2002
Notation: Dependent variable = Efficiency score MANAGE = 1 if privately-owned or commercially managed, otherwise = 0 PAX/AIR = Average number of passengers per aircraft movement INTER = Percentage of international passenger (%) AIR50 = 1 if the market share of aircraft movements > 50%, otherwise = 0 PAX50 = 1 if the market share of annual total passengers > 50%, otherwise = 0 Y2001 = 1 if compute performance score in year 2001, otherwise = 0 Y2002 = 1 if compute performance score in year 2002, otherwise = 0 PAX/RWA = Annual total passengers per runway area (million/acre) AIR/AREA = Annual aircraft movements per land area (103/acre) AIR/RWA = Annual aircraft movements per runway area (103/acre) ** Estimated coefficient is significant at the 0.01 level (one-tailed) * Estimated coefficient is significant at the 0.05 level (one-tailed)
- 94 -
Assessment of Productive Efficiency of Airports
Table 5.4 shows the final model estimation results. It is the proposed model that may be used for predicting the efficiency score. Instead of using market share like the preliminary model in Table 5.3, the proposed model considers market dominance as an exploratory variable in order to capture the effects of market characteristics. Market dominance is represented by a dummy variable, equal to 1 if an airport has market share more than 50%. There are two market dominance variables i.e. dominance by aircraft movements share (AIR50) and dominance by passenger share (PAX50). As observed previously, market dominance tends to be associated with efficient airports. They turn out to be significant, as observed. The reason may be that an airport is in a better position to utilize its inputs, given higher traffic. Most of the estimated coefficients, except Y2001, are meaningful and significant at above the 95% confidence level. Privately-operated airports (MANAGE = 0.6494) tend to be more efficient than their publicly-operated counterparts, possibly due to higher risk and higher accountability of the management. All utilization ratio variables contribute to higher efficiency, as expected. Proportion of international passengers is negatively associated with the efficiency score. A higher proportion of international passengers lead to lower efficiency. There might be some effect from anomaly in 2001 as Y2001 becomes stronger, though not yet significant enough. The model captures the anomaly in 2002, a year after September-11, where variable Y2002 is significant. The negative sign indicates that an airport becomes efficient slightly more easily in 2002. The marginal effect suggests that for every additional million passenger per acre, an airport would be more efficient by 3.8013 units. In summary, the first attempt to assess productivity of airports operating in multiple airports systems is presented in this case study. The samples consist of 72 airports in 25
- 95 -
Assessment of Productive Efficiency of Airports
MASs worldwide. The analysis period is during 2000 – 2002. The DEA-Output-VRS model is used as an analytical tool. Five indicators were considered, i.e., land area, number of runways, runway area, number of passengers and number of aircraft movements. The results indicate that there are two groups of efficient or highly productive airports, coined by the busy and the compact. The busy group consists of market leaders in large MASs such as O’Hare International (ORD), Los Angeles International (LAX) and Heathrow/London (LHR). Airports in the compact group are mostly alternative airports with relatively small land area and one or two runways. The reason that both are classified as efficient airports is mainly due to their relatively higher runway utilization. In this respect, larger size of airport does not guarantee high efficiency. It is also found that some airports are under utilized such as Montreal-Mirabel (YMX), Glasgow Prestwick International (GLA), Schoenefeld/Berlin (SXF), and Tempelhof/Berlin (THF). In fact, Montreal-Mirabel (YMX) is a case study of an unsuccessful airport in textbooks (Caves and Gosling, 1999; de Neufville, 1995; de Neufville and Odoni, 2003). Schoenefeld/Berlin (SXF), and Tempelhof/Berlin (THF) and Tegel (TXL, another airport in the Berlin MAS) are planned to be consolidated in 2011. Construction is underway (Berlin Brandenburg International, 2005). In this sense, the proposed models in this case study are useful in pointing out over-investment. Furthermore, a productivity prediction model was developed by using the Censored Tobit Regression. It is found that factors such as utilization of land area and runway area, passengers per aircraft movement, market dominance and privately-operated management style contribute to the enhancement of productivity. Meanwhile a higher proportion of international passengers tends to reduce the productivity. The model also captures anomaly effects in the year 2002 such that an airport could become efficient slightly more easily with
- 96 -
Assessment of Productive Efficiency of Airports
the same utilization rate, possibly due to a global drop in air traffic after September 11 terrorist attacks. Given some planned measures, the model can be used to predict future total productivity of an airport which should be very useful as a tool for planning airport business in a competitive market. An important observation from this case study is that an efficient airport needs to be very busy. Some efficient airports are in fact constrained and show sign of undesirable congestion. Like previous airport studies listed in Tables 2.2, 2.3 and 2.4, the downside of facilities and quality of service are still out of consideration. This may be due to the inapplicability of the DEA models to take into account such measures. In the next case study, this issue will be addressed by considering joint production of desirable and undesirable outputs from airport operations. Given that delay data are available for U.S. airports, the case study will be to assess productivity of U.S. airports using delays as a proxy of undesirable outputs. The results should provide a more complete total productivity index.
- 97 -
Assessment of Productive Efficiency of Airports
CHAPTER 6 CASE STUDY 2 PRODUCTIVITY OF U.S. AIRPORTS
Results from case study 1 indicate that ignoring the downside of facilities and quality of service from the assessment typically leads to the conclusion that efficient airports must be very busy. Such results may provide an inappropriate benchmark for managing other airports. In practice, other important output measures reflecting quality of services are always taken into account. Among them, delay is perhaps a major concern. This chapter describes the second case study where the productivity assessment will take undesirable outputs, i.e., delays into consideration.
6.1 Modeling airport operations
An airport may be viewed as a production unit. As a result, airport operation can be modeled as a production process that requires some inputs for running day-to-day operations in order to produce some target outputs. Necessary inputs include production factors such as capital and labor. Most airport managers set target to maximize movement of aircrafts, passenger throughput and quantity of cargo transported. These outputs are highly desirable and the primary reason for building an airport. However, production of these outputs is always constrained by capacity. As the air traffic volume increases, the likely by-product output is higher delays. An airport manager also wants to make sure that the undesirable byproducts from the airport operation are being kept at the minimal possible level. Furthermore, an airport is bound to comply with rules and regulations which ensure that its operation does
- 98 -
Assessment of Productive Efficiency of Airports
not create unacceptable externalities, notably noise and pollution. In this situation, where on one hand an airport manager wants to maximize desirable outputs and on the other hand minimize undesirable outputs, the directional output distance function is perfectly applicable.
6.2 Inputs and outputs of airport operations
The formulation of the directional output distance function as shown in linear programming in (3.7) requires data on quantity of inputs, desirable outputs, and undesirable outputs, ( x k , y k , b k ) for individual airports. The selection of inputs and outputs is an important decision issue in the assessment of airport productivity. The general suggestion is to include all important measures that are in the interest of the management. Such measures should be common for all airports so that the performance would provide meaningful interpretation. In practice, the main problem is the availability of the data across all airports rather than model limitations. After all, three common physical inputs are considered in this analysis:
x1 = Land area, acre x 2 = Number of runways
x3 = Runway area, acre These inputs may be rather limited due to the availability of data, but they are necessary infrastructure for all airports. Land area (acre) represent considerable share of capital investment that an airport should fully utilize. Number of runways counts all existing runways at the airport regardless of their utilization level. Runway area is the summation of the length x width product of all runways. Runway area is included to reflect the effect of design configuration such as length, width, and separation on productivity. - 99 -
Assessment of Productive Efficiency of Airports
For the set of desirable outputs, it is assumed that airport managers aim at producing the following three outputs as much as possible:
y1 = Non-delayed flights y 2 = Passengers
y 3 = Cargo throughput All outputs are considered on annual basis. A flight is counted as non-delayed if it is operated no later than 15 minutes from the scheduled time according to Federal Aviation Administration (FAA)’s definition (FAA, 2005 – 2006). The non-delayed flights include all kinds of movements, i.e., commercial aircrafts, cargo aircrafts, general aviation, and others. The number of passengers counts both arriving and departing passengers for all type of commercial passengers, i.e., international, domestic and direct transit passengers. Cargo throughput is measured in metric tones of both loaded and unloaded freight which includes international freight, domestic freight, and mail. On the set of undesirable output, it is assumed that an airport manager wants to minimize the following two outputs:
b1 = delayed flights b2 = time delays
Again, both undesirable outputs are on annual basis. Delayed flights are those movements that are operated more than 15 minutes later than the scheduled time. One might argue that delayed flights are not necessarily undesirable from the economics perspective because the passengers still can get to their destinations as they wish. However, airport
- 100 -
Assessment of Productive Efficiency of Airports
managers and passengers may have a different perspective. If they have a choice,, they probably prefer to avoid experiencing delayed flights. As a result, the delayed flights are treated as undesirable outputs. As a matter of fact, individual delayed flights may incur different delayed times; more of delayed flights does not necessarily mean lower operational efficiency. It also depends on time duration of total delays. Therefore, the time delay (b2 ) is included to reflect another perspective of delays. Time delays are the accumulation of delays experienced by individual delayed flights. Given the above-selected three inputs, three desirable outputs and two undesirable outputs, the assessment may be perceived as the measurement of productive efficiency of airside operation. A question may arise regarding the selection of land area as an input measure. One might argue that acquisition of land area is not purely for improving airside operation, but for other purposes as well. For instance, an airport may prefer to possess more land than needed as noise buffer or for future expansion. It is possible that an airport may seek investment opportunities beyond aeronautical activities such as land value appreciation, or even commercial development. Nevertheless, land area is somehow under management control and the manager can affect productivity of airside operation by managing it efficiently. More importantly, the assessment of productivity is an exploratory analysis which provides information for further judgment, not an ultimate conclusion for implementation. If an airport is found to be inefficient because of inefficient use of any measure, the manager still can reason against the findings. To account for its effect on productivity, the scenario analysis will be done for with and without consideration of land area.
- 101 -
Assessment of Productive Efficiency of Airports
6.3 6.3.1
Sample characteristics Size of sample
The case study is to assess the relative productivity of major U.S. commercial airports and examine the impact of the inclusion of undesirable outputs on the productivity, ranking and productivity index. Due to readily available data, part of samples are taken from previous airport productivity studies (Pathomsiri and Haghani, 2004; Pathomsiri, Haghani and Schonfeld, 2005; Pathomsiri, Haghani, Dresner and Windle, 2006a). Additional samples are collected in order to increase the sample size. Overall, there are 56 airports in the dataset. This is a relatively high number compared to most previous studies (Abbott and Wu, 2002; Adler and Berechman, 2001; Bazargan and Vasigh, 2003; Fernandes and Pacheco, 2001, 2002, 2005; Gillen and Lall, 1997, 1998; Hooper and Hensher, 1997; Martin and Roman, 2001; Martin-Cejas (2002); Murillo-Melchor, 1999; Nyshadham and Rao, 2000; Oum, Yu and Fu, 2003; Pacheco and Fernandes 2003; Parker, 1999; Pels, Nijkamp and Rietveld, 2001, 2003; Salazar de la Cruz,1999; Sarkis, 2000; Sarkis and Tulluri, 2004; Yoshida, 2004; Yu, 2004). Table 6.1 lists all 56 airports along with the International Civil Aviation Organization (ICAO) airport codes. Nevertheless, the number of samples is checked against several applicable rules of thumb to guarantee the sufficiency and meaningful interpretation. For non-parametric approach, DEA provides a very good guideline. As mentioned earlier in Chapter 4, an excessive number of measures with respect to the total number of DMUs may deteriorate the discriminatory power of DEA model. The larger the number of input and output measures for a given number of airports the less discriminatory the DEA model becomes. Given a certain set of samples, this means that the addition of measures will reduce the discriminatory power
- 102 -
Assessment of Productive Efficiency of Airports
of the DEA model. Essentially, this is because it is possible that an airport may dominate all others on one measure, which in turn makes it looked equally efficient to other efficient airports. This is a major issue encountered by Parker (1999), Adler and Berechman (2001) and Yu (2004). To avoid this problem, the straightforward way is to guarantee that there will be sufficient number of airports for comparison with, regarding any measures. Boussofiane, Dyson and Thanassoulis (1991) recommend that the total number of DMUs be much greater than the number of inputs times the number of outputs. Compared to this analysis where three inputs and five (three desirable plus two undesirable) outputs are selected for the assessment of airport productivity; the number of samples needs to be much more than 3 x 5 or 15 airports in order to reduce the chance that an airport may be too dominant compared to the others on a particular measure. According to the recommendation, the sample size of 56 airports is deemed satisfactory. To avoid losing discriminatory power, Cooper, Seiford and Tone (2000: page 103) recommend that the desired number of DMUs exceed m + s several times. They suggest a more stringent rule of thumb in the following formula (Cooper, Seiford and Tone, 2000: page 252). n ? max{m x s, 3(m + s )} (6.1)
where n is the number of DMUs, while m and s are the numbers of input and output measures, respectively. Substituting m and s , yields the minimum number of samples: n ? max{3 x 5, 3(3 + 5)} = max{15,24} = 24 (6.2)
- 103 -
Assessment of Productive Efficiency of Airports
Again, the sample size of 56 satisfies this recommendation. After all, it can be concluded that the sample size is sufficient for the analysis. Figure 6.1 shows the locations of the sample. Note that these 56 airports are major U.S. commercial airports which regularly appeared at the top of the published statistics from trade publications (ACI, 2002 – 2004). Many of them are in the top twenty according to statistics on annual aircraft movement, passengers, and cargo throughput.
Alaska
Hawaii
Figure 6.1 Locations of 56 airports 6.3.2 Analysis period
In order to obtain timely results, a recent 4-year panel data for the years 2000 through 2003 are collected. Coincidently, the period spans the critical time before and after September 11 terrorist attacks which severely affected the aviation industry worldwide. As a result, it allows for analyzing airport productivity to understand its effect on productivity of U.S. airports.
- 104 -
Assessment of Productive Efficiency of Airports
6.3.3 Data source and definition
Input measures were collected from the Airport Master Record database (FAA, 2004) which records physical characteristics of U.S. airports. The database is revised on a regular basis to reflect input changes at airports (e.g., the addition of runways). The data was verified with airport managers, airports’ websites and reports, to determine if there were major changes in runway characteristics during the analysis period. There was no change in land area during the analysis period. However, it was found that some airports had improved their runways. For example, George Bush Intercontinental (IAH) expanded and extended runway 15R/33L from 6038’ x 100’ to 10000’ x 150’ in 2002. Detroit Metropolitan Wayne County (DTW) opened its 6th runway on December 11, 2001. The number of runways and runway area were then edited accordingly. They were computed precisely by the time the runway improvement was in service during the year, rounding down in month. In the case of DTW, it was recorded that the airport has 5 and 6 runways in 2001 and 2002 respectively. All airports in the dataset are well-established, having been built and served their respective markets for a number of years. This knowledge helps to relieve concerns about possible sudden productivity drops during the early years after initial lumpy investments. It may be assumed, with caution, that any temporal changes in productivity that might be observed result from operational performance. Data on the three desirable outputs, i.e., annual statistics on number of passengers, cargo throughput, and aircraft movements are published by the Airports Council International (ACI, 2002 - 2004). Note that aircraft movements include both delayed and non-delayed flights. Given the assumption that an airport manager wishes to maximize only non-delayed flights, the data on delayed flights needed to be collected as well.
- 105 -
Assessment of Productive Efficiency of Airports
There are two available delay database; the Airline Service Quality Performance (ASQP) database maintained by the Bureau of Transportation Statistics (BTS) and the FAA’s Operational Network (OPSNET) database (FAA, 2005). In both databases, a flight is counted as delayed if it is operated more than 15 minutes later than the scheduled time, according to the FAA’s definition, The ASQP database contains delays reported by 18 certified U.S. air carriers that have at least one percent of total domestic scheduled-service passenger revenues, plus other carriers that report on a voluntarily basis. The reports cover non-stop scheduled-service flights between points within the U.S., including territories. So far, the carriers report monthly on operations at 31 U.S. airports that account for at least one percent of the nation’s total domestic scheduled-service enplanements. The up-to-date list of reportable airlines and airports can be viewed at BTS’s website (BTS, 2006). In June 2003, the airlines began reporting the causes of delays in five broad categories, namely (BTS 2004 - 2006; FAA 2006): 1) Air carrier: the cause of the cancellation and delay was due to circumstances within the airline’s control (e.g., maintenance or crew problems, aircraft cleaning, baggage loading, fueling, etc.). 2) Extreme weather: significant meteorological conditions (actual and forecast) that, in the judgment of the carrier, delays or prevents the operation of a flight (e.g., tornado, blizzard, hurricane, etc.). 3) National aviation system (NAS): delays and cancellations attributable to the national aviation system that refer to a broad set of conditions – non-extreme weather conditions, airport operations, heavy traffic volume, air traffic control, etc. - 106 -
Assessment of Productive Efficiency of Airports
4) Late-arriving aircraft: a previous flight with the same aircraft arrived late, causing the present flight to depart late. 5) Security: delays or cancellations caused by evacuation of a terminal or concourse, re-boarding of aircraft because of security breach, inoperative screening equipment and/or long lines in excess of 29 minutes at screening areas. In broad classification, delays caused by 1) air carrier, 2) extreme weather and 4) latearriving aircraft are attributable to operation of airlines. Meanwhile, 3) NAS and 5) security delays are from airport operation. According to the recent statistics (BTS, 2005), the proportion of delay causes are 25.6, 4.8, 39.6, 29.7 and 0.3 % for air carrier, extreme weather, NAS, late-arriving aircraft and security respectively. This means that about 60% of delayed flights are caused by airline operation and the rest of 40% are from airport operation. The other delay database, OPSET is an official source of historical National Aviation System (NAS) air traffic delays and covers a broad set of causes, such as non-extreme weather conditions, airport operations, heavy traffic volume, terminal volume, air traffic control, runway, equipment and others. The OPSNET database is chosen because of two main reasons. First, the analysis focuses on airport operation, rather than airline operations being the source of the delay as in the BTS database. Second, the OPSET database is more complete, covering all flights, all flight types (both domestic and international), all airlines (U.S. and non-U.S. carriers), and all airports in the sample. On the contrary, if ASQP database was used, there would be a large number of under-reported flights since it is based on the sampling report. The numbers of under-reported flights are varying depending on airports. For instance, ASQP database reported only 79.97% of flights at O'Hare International airport in 2003. The missing 21.03% of flights were resulted from non-reportable airlines,
- 107 -
Assessment of Productive Efficiency of Airports
international flights, unclear report, incomplete data, missing value and so on. The figure is quite different at the Baltimore/Washington International airport where 50.86% of total movements are not reported. Inherently, the ASQP database contained biased delay data due to different sampling rates across airports.
- 108 -
Assessment of Productive Efficiency of Airports
Table 6.1 List of 56 US airports under consideration and their outputs in 2003
Airport name 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Hartsfield-Jackson Atlanta International O'Hare International Los Angeles International, CA Dallas/Fort Worth International, TX Denver International, CO Phoenix Sky Harbor International, AZ McCarran International, NV George Bush Intercontinental, TX Minneapolis/St. Paul International, MN Detroit Metropolitan Wayne County, MI John F. Kennedy International, NY Miami International, FL Newark Liberty International, NJ San Francisco International, CA Orlando International, FL Seattle Tacoma International, WA Philadelphia International, PA Charlotte/Douglas International, NC Boston Logan International, MA LaGuardia, NY Covington/Cincinnati/Northern Kentucky International, KY Lambert-St. Louis International, MO Baltimore/Washington International, MD Honolulu International, HI Salt Lake City International, UT Midway International, IL Fort Lauderdale - Hollywood International, FL Washington Dulles International, VA Tampa International, FL San Diego International, CA
Airport code ATL ORD LAX DFW DEN PHX LAS IAH MSP DTW JFK MIA EWR SFO MCO SEA PHL CLT BOS LGA CVG STL BWI HNL SLC MDW FLL IAD TPA SAN
Total passengers 79,086,792 69,508,672 54,982,838 53,253,607 37,505,138 37,412,165 36,285,932 34,154,574 33,201,860 32,664,620 31,732,371 29,595,618 29,431,061 29,313,271 27,319,223 26,755,888 24,671,075 23,062,570 22,791,169 22,482,770 21,228,402 20,427,317 20,094,756 19,732,556 18,466,756 18,426,397 17,938,046 16,767,767 15,523,568 15,260,791
Cargo (tons) 798,501 1,510,746 1,833,300 667,574 325,350 288,350 82,153 381,926 315,987 220,246 1,626,722 1,637,278 874,641 573,523 193,037 351,418 524,485 140,085 363,082 28,402 392,695 115,574 235,576 421,930 216,870 23,266 156,449 285,352 93,457 135,547
Aircraft movements 911,723 928,691 622,378 765,296 499,794 541,771 501,029 474,913 510,382 491,073 280,302 417,423 405,808 334,515 295,542 354,770 446,529 443,394 373,304 374,952 505,557 379,772 299,469 319,989 400,452 328,035 287,593 335,397 233,601 203,285
Non delayed flights 874,203 859,506 620,178 755,873 498,469 529,971 494,332 458,924 503,049 486,231 274,217 412,559 381,159 325,205 294,300 352,786 432,902 440,079 369,452 357,054 498,577 374,984 297,733 319,976 399,680 323,041 283,700 329,552 232,471 202,506
Delayed flights 37,520 69,185 2,200 9,423 1,325 11,800 6,697 15,989 7,333 4,842 6,085 4,864 24,649 9,310 1,242 1,984 13,627 3,315 3,852 17,898 6,980 4,788 1,736 13 772 4,994 3,893 5,845 1,130 779
- 109 -
Assessment of Productive Efficiency of Airports
Table 6.1 List of 56 US airports under consideration and their outputs in 2003 (Continued)
Airport name 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56
Pittsburg International, PA Ronald Reagan Washington National, DC Oakland International, CA Portland International, OR Memphis International, TN Mineta San Jose International, CA Cleveland Hopkins International, OH Kansas City International, MO Louis Armstrong New Orleans International, LA John Wayne, CA William P. Hobby, TX Ontario International, CA Port Columbus International, OH Albuquerque International Sunport Airport, NM Palm Beach International, FL Jacksonville International, FL Anchorage International, AK Bob Hope, CA Norfolk International, VA Long Beach, CA Birmingham International, AL Pensacola Regional, FL Palm Spring International, CA Jackson International, MS Santa Barbara, CA Stewart International, NY Total
Airport Code PIT DCA OAK PDX MEM SJC CLE MCI MSY SNA HOU ONT CMH ABQ PBI JAX ANC BUR ORF LGB BHM PNS PSP JAN SBA SWF
Total passengers 14,266,984 14,214,803 13,548,363 12,395,938 11,437,307 10,677,903 10,555,387 9,715,411 9,275,690 8,535,130 7,803,330 6,547,877 6,252,061 6,051,879 6,010,820 4,883,329 4,791,431 4,729,936 3,436,391 2,875,703 2,672,637 1,361,758 1,246,842 1,215,093 752,762 393,530 1,094,725,865
Cargo (tons) 121,536 5,774 597,383 239,265 3,390,515 108,622 95,761 136,687 80,831 12,050 5,775 518,710 10,766 71,599 18,300 70,650 2,102,025 44,654 32,283 50,873 34,184 4,569 103 10,957 2,825 19,024 22,599,243
Aircraft movements 361,329 250,802 342,871 267,052 402,258 198,082 258,460 170,758 137,312 350,074 242,635 146,413 237,979 221,003 171,692 121,143 277,361 178,079 121,373 338,807 154,849 127,197 93,068 79,377 152,485 112,284 18,781,482
Non delayed flights 360,619 249,056 342,567 266,872 400,683 197,855 256,993 170,722 137,094 348,475 242,084 146,212 237,915 220,962 169,836 121,043 277,165 177,902 121,330 338,727 154,781 127,195 93,032 79,376 152,434 112,277 18,485,876
Delayed flights 710 1,746 304 180 1,575 227 1,467 36 218 1,599 551 201 64 41 1,856 100 196 177 43 80 68 2 36 1 51 7 295,606
- 110 -
Assessment of Productive Efficiency of Airports
The OPSNET database recorded both the number of delayed flights and time delays, which are two undesirable outputs in the analysis. Given the number of delayed flights from OPSNET database, the number of non-delayed flights (one of desirable outputs) is simply the difference between aircraft movements and number of delayed flights. Table 6.1 presents output measures of 56 airports in 2003. The figures are ordered by number of annual passengers. At the top of the list, Hartsfield-Jackson Atlanta (ATL) is the busiest airport in terms of passengers. O’Hare International (ORD) serviced the highest number of aircraft movements. Memphis International (MEM), the FedEx hub, had the highest cargo throughput. On the downside, ORD experienced the highest number of delayed flights. As shown in Figure 6.2, the number of delayed flights tends to increase with number of passengers serviced at the airport. In fact, there are always externalities inherent in airport operations, notably delay and noise that increase, ceteris paribus, with airport volume. These externalities are also outputs from the production process, although undesirable. In Figure 6.3 density of aircraft movement (number of flights per runway area) is plotted against average delay per passenger, computed for 56 airports in the dataset during 2000 – 2003. This graph shows that higher density of traffic is associated with higher average delay. According to these results, airport efficiency may come at the cost of numerous delays. This situation may be undesirable from the viewpoints of airports, regulators, airlines, and passengers.
- 111 -
Assessment of Productive Efficiency of Airports
Delayed Flights V.S. Passengers
120,000 100,000 Delayed Flights 80,000 60,000 40,000 20,000 0 0 20,000,000 40,000,000 60,000,000 80,000,000 Passengers
Figure 6.2 Scatter plot between number of delayed flights and number of passengers, 2003
Delay/Passenger V.S. Movements/Runway Area
80,000 Delay/Passenger (minutes) 70,000 60,000 50,000 40,000 30,000 20,000 10,000 0 0 2,500 5,000 7,500 10,000 12,500 15,000 17,500 Movements/Runway Area (flights/acre)
Figure 6.3 Scatter plot between delay/passenger and density of movements, 2000 – 2003
- 112 -
Assessment of Productive Efficiency of Airports
Table 6.2 summarizes the descriptive statistics on input and output measures. All measures show large standard deviations suggesting that airports in the sample vary in both scale and scope of operations. During the study period, all airports experienced at least some flight delays and delays are positively associated with air traffic volume, as scatter plot shows in Figure 6.2. This suggests that delays are important undesirable byproducts that should be taken into consideration when assessing airport productivity. In the next chapter, results from assessing productivity of 56 airports by the directional output distance function will be presented. The assessment will be discussed with respect to operational efficiency of individual airports. In addition, changes of productivity over analysis period, as well as the impact of the inclusion of undesirable outputs (i.e., delayed flights and time delays) on productivity and ranking of airports will be discussed.
- 113 -
Assessment of Productive Efficiency of Airports
Table 6.2 Descriptive statistics of samples 2000 – 2003
Input Statistics Land area (acre) 501 33,422 32,921 4,381 2,650 5,298 Number of runways 1.00 7.00 6.00 3.35 3.00 1.21 Runway area (acre) 24.60
Desirable outputs Total passengers 362,017 # of non delayed flights 79,376 874,203 794,827 343,324 326,086 176,881 Cargo throughput (tons) 74 3,390,800 3,390,726 401,667 171,349 591,702
Undesirable outputs # of delayed flights 1 96,346 96,345 5,818 1,355 11,917 Time delays (minutes) 20 5,398,921 5,398,901 259,558 57,200 611,968
Minimum Maximum Range Mean Median Standard deviation
305.87 80,162,407 281.26 79,800,390 104.21 20,009,558 99.56 16,225,655 51.65 16,924,416
- 114 -
Assessment of Productive Efficiency of Airports
CHAPTER 7 CASE STUDY 2 RESULTS AND DISCUSSION
7.1 Impact of the inclusion of undesirable outputs
For each year (2000 – 2003), the directional output distance function in formulation (3.7) is solved 56 times; i.e., one time for each airport, to determine the optimal efficiency score ? * . ? * , or the distance from the efficient frontier, measures the level of inefficiency. An efficient airport will have ? * = 0. The higher value of ? * shows greater inefficiencies. There are several cases in the case study. Case 1 is for the case that considers delayed flights and time delays as undesirable outputs. The sets of inputs and outputs are as follows: Case 1: with consideration of undesirable outputs Input = {land area, number of runway, runway area} Desirable outputs = {non-delayed flights, passengers, cargo} Undesirable outputs = {delayed flights, time delays} In order to analyze the effect of the inclusion of undesirable outputs on productive efficiency, a model that ignores undesirable outputs is also solved. Suppose that Case 2 is for the case that does not consider undesirable outputs. The sets of inputs and outputs are as follows: - 115 -
Assessment of Productive Efficiency of Airports
Case 2: without undesirable outputs or Case 1 – {delayed flights, time delays} Input = {land area, number of runway, runway area} Desirable output = {aircraft movements, passengers, cargo} Undesirable output = {none} Aircraft movements are the total number of operations both landings and takeoffs, regardless of delay status. To assess airport productivity for this case, the model (3.7) needs to be modified accordingly. Specifically, the model in (3.7) is modified by taking out the constraints associated with undesirable outputs, resulting in the following model: max ? s.t.
k?K
?? ??
k
y km ? (1 + ? ) y km , m = 1,..........., M , x kn ? x kn , n = 1,.............., N ,
(7.1)
k?K
k
? k ? 0, k = 1,.........., K
The model described in (7.1) is also solved 56 times, each time for an individual airport. Table 7.1 shows efficiency scores for the two cases annually during the period 2000 – 2003. In the Table, airports are ordered alphabetically by their corresponding airport codes. An efficient airport must yield a score of zero, implying that increases in desirable outputs or decreases in undesirable outputs and inputs from current levels are not necessary. In Table 7.1 the efficient airports are highlighted with bold typeface. Several observations can be made from the results. They are discussed in the following sections.
- 116 -
Assessment of Productive Efficiency of Airports
7.1.1
Classification of efficient airports
When delayed flights and time delays are ignored (Case 2), the results are typical of those reported in past studies which suggest that operational efficiency is associated with busy airports (Adler and Berechman, 2001; Bazargan and Vasigh, 2003; Fernandes and Pacheco, 2002; Gillen and Lall, 1997, 1998; Martin and Roman, 2001; Oum and Yu, 2003; Pacheco and Fernandes, 2003; Pathomsiri and Haghani, 2004; Pathomsiri, Haghani and Schonfeld, 2005; Pathomsiri, Haghani, Dresner and Windle, 2006; Pels, Nijkamp and Rietveld, 2003; Sarkis, 2000; as well as Sarkis and Talluri, 2004. As is evident from the 2003 data, six efficient airports are also very busy. For examples, Hartsfield-Jackson Atlanta (ATL) and Memphis (MEM), respectively, are the busiest airports in the world in terms of number of passengers and cargo throughput. LaGuardia (LGA) is one of the most chronically congested airports in the U.S. (CRA, 2001). John Wayne airport (SNA) constrains the number of passengers using its facilities. Other well-known busy airports such as Anchorage International (ANC), Newark Liberty International (EWR), John F. Kennedy International (JFK), Midway International (MDW), Miami International (MIA), O'Hare International (ORD), Seattle Tacoma International (SEA) and Lambert-St. Louis International (STL), though not classified as efficient, show very low inefficiency level. They all earn relatively low efficient scores (less than 0.5). The implication of these results is that an airport that is very busy or constrained is generally determined to be efficient.
- 117 -
Assessment of Productive Efficiency of Airports
Table 7.1 Efficiency scores for Case 1 and Case 2
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM 2000 Case 1 Case 2 0.2034 2.6026 0.2314 0.0000 0.0000 0.0000 2.0490 0.0000 0.5044 0.9495 0.0108 1.4154 1.7999 0.0000 0.4323 1.4892 0.3502 0.6097 1.0697 0.0000 0.4044 0.5496 0.0102 0.7610 0.2956 1.2381 0.7486 0.9718 0.6938 1.2993 0.0892 0.0000 0.5941 0.8739 1.6907 0.0000 0.3367 2.0979 0.4777 0.6289 0.5179 0.8696 4.2904 0.0000 0.5331 2.2758 0.3099 0.3469 0.5919 0.6444 0.0000 0.0000 0.0000 0.0000 1.1960 0.0000 0.2827 2.4069 0.2619 0.9268 0.0118 0.3364 0.0000 0.0000 2001 Case 1 Case 2 2.3279 0.0000 0.0506 0.0000 0.0000 0.0000 1.8211 0.0000 0.5901 1.0133 1.3685 0.0000 1.6086 0.0000 0.4745 1.7279 0.1170 0.4057 0.7945 0.0000 0.3123 0.7237 0.3270 0.9666 0.0263 1.2805 0.6133 0.9455 0.4017 1.1319 0.2055 0.0000 0.1505 0.7538 1.6235 0.0000 0.3014 2.0753 0.2531 0.6822 0.8391 0.0000 3.6530 0.0000 0.3636 2.2219 0.2767 0.4730 0.0045 0.6896 0.0000 0.0000 0.0000 0.0000 1.2235 0.0000 0.1933 2.1817 1.0137 0.0000 0.2882 0.0000 0.0000 0.0000 2002 Case 1 Case 2 2.1394 0.0000 0.2719 0.0000 0.0000 0.0000 1.8425 0.0000 0.5707 1.1142 1.2516 0.0000 1.6914 0.0000 0.6814 2.1009 0.2693 0.4162 0.7010 0.0000 0.2028 0.3720 0.0831 1.0714 0.0723 1.2264 0.7380 1.0190 0.6783 1.6990 0.0362 0.1541 0.3577 0.7183 1.5242 0.0000 0.4408 2.0687 0.4155 0.7778 0.5056 0.9481 4.1563 0.0000 0.5007 2.4574 0.3642 0.4246 0.2138 0.6718 0.0000 0.0000 0.0000 0.0000 1.2579 0.0000 0.2980 2.4846 0.0383 1.1632 0.1665 0.0000 0.0000 0.0000 2003 Case 1 Case 2 0.1459 2.5656 0.1607 0.0000 0.0000 0.0000 1.7062 0.0000 0.5366 1.1994 0.9675 0.0000 1.7879 0.0000 0.5329 1.9452 0.2139 0.4722 0.8564 0.0000 0.1445 0.3526 0.8962 0.0000 0.0370 1.4323 0.5632 1.0848 0.5520 1.7536 0.1417 0.0000 0.3649 0.7067 1.5904 0.0000 0.3063 2.0455 0.4017 1.0388 0.6074 0.9998 4.4356 0.0000 0.4274 2.6094 0.2737 0.3759 0.4974 0.6483 0.0000 0.0000 0.0000 0.0000 1.2281 0.0000 0.3422 2.9981 0.1094 1.1712 0.1177 0.0000 0.0000 0.0000
- 118 -
Assessment of Productive Efficiency of Airports
Table 7.1 Efficiency scores for Case 1 and Case 2 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Average score Number of efficient airports 2000 Case 1 Case 2 0.0301 0.0000 0.1833 0.3488 0.1553 2.6039 0.5276 0.0000 1.2123 0.0000 0.2600 0.5309 0.2944 2.5854 0.6936 2.2824 1.1367 0.0000 0.7166 0.6512 0.0000 0.0000 0.2264 1.2968 0.1525 2.7395 0.4099 4.0613 0.0000 0.0000 2.4880 0.0000 0.0821 0.0000 0.5650 0.7716 0.8666 0.0000 0.6842 1.6591 0.0000 0.0000 0.3992 0.0000 0.0147 2.5810 0.4392 1.5896 0.2208 1.1813 23 7 2001 Case 1 Case 2 0.0000 0.0000 0.0393 0.2947 2.4819 0.0000 0.5457 0.0000 0.2388 1.2000 0.1223 0.4958 2.4231 0.0000 0.2441 2.1645 1.1653 0.0000 0.5415 0.6293 0.1491 0.0000 0.1214 0.9713 2.4767 0.0000 0.7247 3.7163 0.0000 0.0000 2.5483 0.0000 0.0873 0.0000 0.4686 0.9830 0.8341 0.0000 0.2175 1.3735 0.0000 0.0000 0.2189 0.3213 0.0512 2.8170 0.0294 1.4895 0.1326 1.1296 29 7 2002 Case 1 Case 2 0.0229 0.0000 0.2002 0.2750 2.4938 0.0000 0.6508 0.0000 1.0546 0.0000 0.4646 0.0000 0.0331 2.2075 0.6013 2.5192 1.2548 0.0000 0.5628 0.0000 0.1590 0.0000 0.3194 1.0942 2.0497 0.0000 3.5666 0.0000 0.0000 0.0000 2.4674 0.0000 0.1816 0.0000 0.7539 1.1199 0.3169 1.1352 0.1123 1.1670 0.0000 0.0000 0.3945 0.4253 2.4968 0.0000 0.1627 1.6490 0.1672 1.1591 29 6 2003 Case 1 Case 2 0.1650 0.0000 0.1709 0.2884 0.1366 2.6035 0.7313 0.0000 0.1085 1.0359 0.4776 0.0000 0.1202 2.2866 0.6686 2.4350 1.3765 0.0000 0.6064 0.5875 0.1739 0.0000 0.3257 1.5232 2.0909 0.0000 3.0955 0.0000 0.0000 0.0000 2.4552 0.0000 0.2607 0.0000 0.7767 1.2892 1.3895 0.0000 0.1898 1.2490 0.0000 0.0000 0.5103 0.6358 2.9040 0.0000 0.3630 1.8012 0.1792 1.2168 28 6
Note: An efficient airport has a zero score as highlighted by bold typeface. The input set of both cases are the same. The
output set of Case 2 consist of passengers, aircraft movements, and cargo throughput. The output set of Case 1 include passengers, non-delayed flights, cargo throughput, delayed flights, and time delays.
- 119 -
Assessment of Productive Efficiency of Airports
On the contrary, when delayed flights and time delays are also considered (Case 1), the results show a greater number of efficient airports, including less-congested airports. In 2003, totally 28 airports are identified as efficient, as shown in Table 7.1. The additional 22 airports received credit due to their relatively low numbers of delayed flights and total time delays. The different classification is a result of overemphasis on increasing desirable outputs and not giving credit to airports with good performance on controlling undesirable outputs. The results indicate that there may be a balance between quantity and quality of outputs in the achievement of efficient outcomes; i.e., airports can trade-off utilization levels for reduced flight and time delays. For certain stakeholders, this option may be an optimal strategy. Passengers and shippers receive services with fewer flight delays. The FAA, as the regulator, has less concern over congestion and safety. Meanwhile, airport managers are able to balance traffic volume with customer satisfaction. By all accounts, the inclusion of undesirable outputs in the analysis appears to provide a fairer assessment of airport productivity.
7.1.2 The number of efficient airports
The results also show that the number of efficient airports increases as the number of measured outputs increases. The results are in line with Salazar de la Cruz (1999) who also found that as the number of variables (inputs and outputs) increases, the number of efficient airports is likely to be more. In 2000, without consideration of delayed flights and time delays (Case 2), only seven efficient airports are identified i.e., HartsfieldJackson Atlanta (ATL), Los Angeles International (LAX), LaGuardia (LGA), Phoenix Sky Harbor International (PHX), San Diego International (SAN) and John Wayne (SNA).
- 120 -
Assessment of Productive Efficiency of Airports
With the consideration of delayed flights and time delays as undesirable outputs (Case 1), there are 23 efficient airports or 16 more. The increase in the number of efficient airports as outputs are added is partly due to the way efficient units are calculated using nonparametric linear programming methods. The greater the number of outputs, the less likely an airport is to be dominated on all outputs; thus the more likely it is to be on the efficient frontier. As pointed out earlier in Chapter 3 the best way to avoid domination is to have a sufficiently large sample size so that an airport at least has some peers for comparison. This is of course a case in this dissertation.
7.1.3 Difference in efficiency scores
In order to show that efficiency scores are significantly different between Cases 1 and 2, several statistical tests are applied to the results. Tests are performed on both yearly basis and all years. Table 7.2 provides the results from paired-sample t-tests by treating efficiency scores as random variables. The results strongly support the assertion of differences. The efficiency scores in Case 2 are statistically higher than in Case 1. To avoid restricted assumptions of t-test, the non-parametric Wilcoxon signed-rank test and sign test are also performed. The results are shown in Table 7.3. They confirm that the difference in efficiency scores between cases with and without consideration of undesirable outputs is significant.
7.1.4 Ranking
The efficiency scores can be used to rank the performance of airports. All efficient airports (score = 0) are equally efficient since they all are on the efficient frontier. The inefficient airports are ranked in descending order by their efficiency scores. Table 7.4 shows ranking of airport productivity during 2000 – 2003. As shown in the - 121 -
Assessment of Productive Efficiency of Airports
Table rankings are drastically different when cases with and without consideration of undesirable outputs are compared. For example, by accounting for delays (Case 1) the operational efficiency of Boston Logan International (BOS) is ranked 25th, 54th, 46th, and 49th during 2000 – 2003 as compared to 29th, 31st, 29th, and 31st when ignoring delays (Case 2). This indicates that by not accounting for delayed flights and time delays, the performance ranking of airports can be distorted.
Table 7.2
Comparisons of efficiency scores between Cases 1 and 2 by paired sample t-test Paired differences Cases 1 and 2 95% confidence interval of the Std. Std. error difference deviation mean Lower Upper 1.0109 0.1351 0.6898 1.2312 0.9211 0.1231 0.7504 1.2437 0.9571 0.1279 0.7356 1.2482 0.9872 0.1319 0.7733 1.3020 0.9635 0.0644 0.8699 1.1236
Paired-sample t-test Pair 1: year 2000 Pair 2: year 2001 Pair 3: year 2002 Pair 4: year 2003 Pair 5: 2000 – 2003
Mean 0.9605 0.9971 0.9919 1.0377 0.9968
t 7.110 8.100 7.755 7.866 15.484
- 122 -
Assessment of Productive Efficiency of Airports
Table 7.3
Comparisons of efficiency scores between Cases 1 and 2 by nonparametric paired tests Asymptotic significance (2-tailed) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Nonparametric paired test A. Wilcoxon Signed-Rank test Pair 1: year 2000 Pair 2: year 2001 Pair 3: year 2002 Pair 4: year 2003 Pair 5: 2000 – 2003 Pair 1: year 2000 Pair 2: year 2001 Pair 3: year 2002 Pair 4: year 2003 Pair 5: 2000 – 2003
Z -6.053a -6.093a -6.154a -6.144a -12.189a -6.571 -6.857 -6.930 -6.647 -13.716
B. Sign test
a
Based on positive ranks.
- 123 -
Assessment of Productive Efficiency of Airports
Table 7.4 Ranking of airport productivity
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW
2000 2001 2002 2003 Case 1 Case 2 Case 1 Case 2 Case 1 Case 2 Case 1 Case 2 1 52 1 49 1 46 34 53 1 11 1 8 1 11 1 8 1 1 1 1 1 1 1 1 46 44 1 44 1 44 1 44 25 29 54 31 46 29 49 31 1 38 1 38 26 38 1 38 43 43 1 41 1 43 1 41 40 39 52 43 43 39 48 43 1 19 35 15 40 19 37 15 41 31 1 24 1 31 1 24 24 18 47 22 41 18 33 22 37 23 48 28 25 23 1 28 56 35 31 37 37 35 29 37 54 30 55 27 56 30 51 27 1 37 50 33 54 37 50 33 51 10 1 11 1 10 1 11 1 27 38 23 51 27 43 23 39 42 1 42 1 42 1 42 45 45 46 45 39 45 39 45 47 20 44 20 45 20 44 20 1 26 1 26 47 26 53 26 48 56 1 55 1 56 1 55 38 46 49 48 48 47 45 48 50 13 45 16 38 13 38 16 1 21 30 21 50 21 46 21 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 35 33 1 36 1 33 1 36 34 48 39 47 35 49 41 47 26 28 1 32 34 28 30 32 1 12 1 12 27 12 1 12
- 124 -
Assessment of Productive Efficiency of Airports
Table 7.4 Ranking of airport productivity (Continued)
Airport code MEM MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Sum Average 56 airports Average inefficiency # of efficient airports
2000 2001 2002 2003 Case 1 Case 2 Case 1 Case 2 Case 1 Case 2 Case 1 Case 2 1 1 1 1 1 1 1 1 30 8 1 1 1 8 1 1 29 14 33 13 31 14 35 13 1 53 1 52 30 53 32 51 1 16 1 18 1 16 1 18 33 34 42 35 1 34 29 35 36 17 37 17 33 17 1 17 53 51 1 50 36 52 31 49 1 47 43 46 53 48 54 46 55 32 1 34 1 32 1 34 1 22 53 19 55 22 52 19 32 1 1 10 1 1 1 10 28 36 36 29 32 36 40 29 42 54 1 51 29 54 1 50 1 55 56 56 42 55 1 56 1 1 1 1 1 1 1 1 1 49 1 53 1 50 1 52 49 9 1 9 1 9 1 9 1 24 51 30 49 24 55 30 52 25 1 25 1 25 1 25 1 41 40 39 52 41 36 39 1 1 1 1 1 1 1 1 27 15 41 14 1 15 47 14 44 50 34 54 28 51 1 54 31 40 32 40 44 40 42 40 1,343 1,575 1,190 1575 1320 1,575 1191 1575 0.2208 1.1813 0.1326 1.1296 0.2172 1.1731 0.1792 1.1339 0.3747 1.3501 0.2749 1.2910 0.3801 1.3406 0.3135 1.2958 23 7 29 7 24 7 28 7
Note: An efficient airport has a ranking = 1.
At the bottom of Table 7.4, the average efficiency scores across inefficient airports are shown. In 2003, the average scores are 0.3135 and 1.2958 for Cases 1 and 2 respectively. These average scores show that the performance of the inefficient airports is about four times poorer if delays are ignored. This result casts a doubt if these airports are really performing that poorly. While these figures may be a result of using incomplete
- 125 -
Assessment of Productive Efficiency of Airports
output measures, they also may be due to ignoring delays that cause unrealistic inefficiency level. This observation leads to another interesting insight next.
7.1.5 Maximum possible production outputs
Another interesting and insightful observation is that by ignoring delayed flights and time delays as outputs, the level of inefficiency may be overestimated. Recall that the terms (1 + ? ) y km plus the corresponding output slacks and (1 ? ? )bkj in (3.7) give the projection of desirable and undesirable outputs onto the frontier. For inefficient airports, these terms represent the maximum possible production outputs or highest potential outputs that an airport could have produced. For an efficient airport with ? = 0 , the terms are simply ( y km , bkj ) or the current level of outputs. It can be seen in Table 7.1 that efficiency scores in Case 2 are much higher than in Case 1. For example, in 2003, Albuquerque International Sunport (ABQ) has a score of 2.5656 in Case 2; implying that ABQ could accommodate at least 2 256.56% more passenger trips, aircraft movements and cargo throughput. Meanwhile, in Case 1, ABQ receives a relatively lower score of 0.1459, implying that ABQ would only need to increase all outputs by 14.59% in order to be on the efficient frontier. Overall, in 2003, the average score for these 56 airports suggest that the U.S. airport system should increase all outputs by 17.92% according to the calculations in Case 1 in order to achieve maximum possible production. On the contrary, the system would have had to produce as high as 121.68% more in Case 2. Tables 7.5, 7.6, 7.7 and 7.8 compare estimated maximum possible production of each airport during 2000 – 2003, for Cases 1 and 2. In each case, the percentage increase
2
The maximum possible production outputs may be higher depending on the value of output slacks.
- 126 -
Assessment of Productive Efficiency of Airports
from current levels of outputs is also computed. Let’s look at results of the most recent year 2003 (Table 7.8) for explanation. In Case 2, Albuquerque International Sunport (ABQ) had the potential to produce 54,411,318 passengers rather than 6,051,879 that was actually produced in 2003, a 799% increase in passengers. In practice, if ABQ were to produce this high output, it is likely that the number of delayed flights and time delays would be very high and unacceptable. However, after consideration of delayed flights and time delays (Case 1), the maximum possible output at ABQ is just 6,935,011 passengers, or a 14.59% percent increase over the current level. In general, ignoring undesirable outputs may yield unrealistic maximum possible production outputs. The unrealistic figures occur mainly because the model neglects the relationship between traffic volume, capacity and delay. In practice, delays play a major role in determining acceptable traffic volume and vice versa. The joint consideration of capacity and delays is therefore necessary (de la Cruz, 1999). In certain situations, the capacity of airside operation is limited by environmental considerations (Pels, Nijkamp and Rietveld, 2003). Based on the results of Case 1, it is suggested that the 56 airports grossly have the potential to increase passengers, aircraft movements (delayed plus non-delayed flights) and cargo throughput by 23.03%, 20.19%, and 34.54%, respectively. If the undesirable outputs are not considered (Case 2), the increases are 133.50%, 90.98% and 363.68%, respectively. The numbers are shown at the end of Table 7.8. The difference of the estimation between Cases 1 and 2 may be interpreted as amount of output loss due to cleaning up delayed flights and time delays or keeping them at relatively low levels.
- 127 -
Assessment of Productive Efficiency of Airports
Table 7.5 Maximum possible passengers, aircraft movements and cargo throughput in 2000
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM Case 2
31,337,627 11,630,465 80,162,407 23,268,112 54,053,755 11,470,266 54,885,529 33,076,352 39,402,365 27,704,471 46,087,079 27,978,866 86,731,133 119,828,724 81,705,167 39,629,328 29,719,867 65,387,989 28,205,634 48,445,074 65,906,186 25,459,430 26,717,442 44,252,913 60,622,286 67,303,182 25,374,868 18,625,264 44,451,860 59,390,735 20,945,697 11,769,213
Total passengers % add Case 1
398 131 0 658 95 142 180 149 71 303 104 76 124 97 130 16 87 184 210 141 87 1,772 405 35 64 0 0 2,820 260 93 34 0 7,572,239 5,030,557 80,162,407 3,067,777 41,712,092 4,800,070 19,602,609 24,396,829 31,173,199 6,873,998 41,581,182 16,050,706 50,206,299 106,262,017 70,620,630 34,188,468 25,282,740 23,016,542 12,170,935 42,098,078 53,509,212 1,360,280 8,106,485 43,038,655 58,688,340 67,303,182 25,374,868 637,853 15,841,801 38,897,365 15,857,218 11,769,213
% add
20 0 0 0 50 1 0 84 35 0 84 1 30 75 99 0 59 0 34 109 52 0 53 31 59 0 0 0 28 26 1 0
Case 2
841,169 355,766 915,454 469,295 951,353 388,327 886,740 826,156 727,592 492,609 740,464 524,560 1,144,754 1,651,938 1,276,964 490,392 548,365 930,378 788,203 743,506 904,086 480,811 487,422 645,887 857,227 783,433 383,325 833,152 743,773 691,833 398,415 388,412
Aircraft movements % add Case 1
260 23 0 205 95 142 180 149 61 107 55 76 124 97 130 9 87 169 210 63 87 429 228 87 64 0 0 120 241 93 34 0 280,925 288,919 915,454 153,917 726,532 162,504 316,703 474,220 607,944 238,011 665,523 300,889 739,636 1,438,268 931,597 450,229 459,262 345,771 339,653 657,193 714,538 90,883 227,940 648,506 815,812 783,433 383,325 379,399 279,970 540,853 301,469 388,412
% add
20 0 0 0 49 1 0 43 34 0 39 1 45 72 68 0 57 0 33 44 48 0 53 88 56 0 0 0 28 51 1 0
Case 2
310,571 2,221,661 868,286 198,815 925,906 90,332 660,899 297,305 351,435 251,273 605,615 76,112 1,055,298 1,780,819 782,621 1,178,964 443,491 1,187,052 95,890 625,271 688,946 224,726 239,602 2,449,730 524,591 2,038,784 71,149 108,514 513,023 639,250 56,849 2,489,078
Cargo (tons) % add Case 1
260 23 0 388 95 142 180 149 78 1,013 55 102 124 97 163 9 87 169 1,136 63 87 1,236 293 35 426 0 0 120 241 136 169 0 130,936 1,804,221 868,286 40,722 714,502 62,433 236,043 211,276 287,627 22,572 644,908 177,857 610,884 1,579,199 1,325,558 1,082,407 515,706 441,163 150,134 567,217 1,146,512 16,815 93,435 2,382,512 1,060,137 2,038,784 71,149 49,415 244,619 545,822 99,464 2,489,078
% add
52 0 0 0 50 67 0 77 46 0 65 371 30 75 345 0 118 0 1,836 48 211 0 53 31 963 0 0 0 62 101 371 0
- 128 -
Assessment of Productive Efficiency of Airports
Table 7.5 Maximum possible passengers, aircraft movements and cargo throughput in 2000 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Total Case 2
41,331,587 49,589,802 35,585,902 17,638,778 27,630,613 110,445,560 19,038,966 22,028,846 34,705,402 41,289,840 36,044,635 64,079,351 17,784,319 13,963,990 14,868,547 12,042,199 32,471,210 72,724,391 24,447,399 56,770,120 7,772,801 42,762,111 26,957,370 41,546,269 2,275,049,295
Total passengers % add Case 1
23 35 260 61 309 53 524 277 152 66 0 223 1,572 990 0 1,450 14 77 87 185 0 40 4,958 159 91 33,621,273 43,503,104 11,407,888 10,963,802 6,757,398 90,903,872 3,946,234 9,894,942 13,790,115 45,168,996 36,044,635 26,063,445 3,769,595 3,991,725 14,868,547 776,904 28,408,553 64,240,596 13,097,259 34,339,104 7,772,801 30,561,387 8,144,375 23,090,278 1,551,380,674
% add
0 18 16 0 0 26 29 69 0 81 0 32 254 212 0 0 0 56 0 73 0 0 1,428 44 30
Case 2
533,000 705,627 575,108 685,977 344,014 1,391,570 447,069 636,289 671,733 799,671 579,816 1,030,784 440,476 420,399 206,289 583,806 482,282 839,091 535,850 975,732 387,862 673,060 488,683 719,560 38,445,502
Aircraft movements % add Case 1
3 35 260 53 121 53 259 228 114 65 0 130 274 406 0 249 8 95 87 166 0 40 258 159 86 517,440 616,656 184,325 449,050 155,501 1,156,282 161,378 324,876 314,378 792,038 579,816 550,159 135,749 117,090 206,289 167,376 445,677 778,647 287,072 614,373 387,862 481,025 138,470 399,174 26,008,394
% add
0 18 16 0 0 27 29 68 0 64 0 23 15 41 0 0 0 81 0 67 0 0 1 44 26
Case 2
1,692,142 498,911 311,060 1,047,067 1,026,868 2,248,207 148,807 146,000 602,591 923,560 340,352 605,070 133,971 88,797 139,107 27,912 494,449 1,545,324 276,125 682,341 15,589 277,661 242,439 383,839 37,950,051
Cargo (tons) % add Case 1
3 35 260 53 121 53 414 596 114 65 0 312 2,196 67,684 0 840 8 77 87 166 0 113 647 273 59 1,642,744 1,045,794 164,659 685,425 464,164 2,437,829 57,061 434,258 282,019 960,161 340,352 180,292 77,276 80,433 139,107 2,970 456,920 1,707,269 147,929 432,166 15,589 130,152 156,004 421,808 34,173,774
% add
0 183 91 0 0 66 97 1,971 0 72 0 23 1,224 61,299 0 0 0 96 0 68 0 0 381 309 43
Note: Cases 1 and 2 are with and without consideration of undesirable outputs, respectively. Aircraft movements include both delayed
and non-delayed flights. % add is the percentage increase from current level of the corresponding output.
- 129 -
Assessment of Productive Efficiency of Airports
Table 7.6 Maximum possible passengers, aircraft movements and cargo throughput in 2001
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM Case 2
47,465,694 9,209,894 75,858,500 27,516,240 48,722,818 10,628,250 57,724,171 36,686,611 47,920,880 33,859,280 56,893,875 25,901,064 90,510,082 129,575,647 94,823,125 36,838,382 28,775,415 71,693,282 26,562,043 56,870,278 64,963,714 30,870,493 32,545,265 43,231,654 59,438,371 61,606,204 21,933,000 15,548,312 56,581,449 56,893,875 20,201,325 11,808,247
Total passengers % add Case 1
668 80 0 813 101 137 183 209 107 407 229 97 151 135 194 21 75 256 208 218 87 2,304 541 47 69 0 0 2,547 370 101 29 0 6,183,606 5,107,311 75,858,500 3,012,729 38,481,290 4,487,335 20,369,923 17,510,135 25,890,177 6,680,897 22,664,832 17,476,832 37,043,730 88,971,994 45,266,906 30,558,000 18,877,562 20,151,936 11,240,147 29,224,387 34,803,580 1,284,311 6,926,081 37,470,750 35,338,310 61,606,204 21,933,000 587,473 14,358,821 28,253,061 15,681,966 11,808,247
% add
0 0 0 0 59 0 0 47 12 0 31 33 3 61 40 0 15 0 30 64 0 0 36 28 0 0 0 0 19 0 0 0
Case 2
807,793 298,831 890,494 419,971 915,317 378,261 845,347 795,772 648,419 436,424 667,871 479,876 1,103,768 1,524,400 1,113,118 526,114 508,702 857,904 766,691 667,618 866,084 429,945 433,576 616,485 834,169 738,114 365,716 797,136 667,193 667,871 359,062 394,826
Aircraft movements % add Case 1
438 5 0 597 101 137 161 173 154 2,015 72 147 144 95 284 21 75 162 2,031 68 84 1,868 406 47 562 0 0 122 287 148 205 0 242,733 284,441 890,494 148,869 703,723 159,705 324,065 428,372 514,675 243,201 506,019 321,934 540,903 1,243,721 725,372 436,420 333,261 327,006 323,801 495,596 470,916 92,402 183,398 598,329 530,328 738,114 365,716 358,508 250,145 315,752 278,734 394,826
% add
0 0 0 0 55 0 0 47 12 0 31 32 12 59 39 0 15 0 30 25 0 0 36 105 7 0 0 0 19 0 0 0
Case 2
392,202 1,968,545 739,927 246,803 795,525 77,872 587,145 278,267 450,829 322,739 554,945 62,278 874,863 1,525,449 924,909 959,095 319,020 885,779 128,895 556,658 621,341 288,059 307,492 2,107,489 530,144 1,774,402 52,148 118,267 551,320 554,945 47,809 2,631,631
Cargo (tons) % add Case 1
438 5 0 597 101 137 161 173 154 2,015 72 147 144 95 284 21 75 162 2,031 68 84 1,868 406 47 562 0 0 122 287 148 205 0 72,876 1,873,750 739,927 35,433 628,306 32,878 225,083 150,411 292,366 15,260 493,936 132,195 420,832 1,378,600 1,442,812 795,584 209,287 337,631 74,212 414,667 337,842 14,634 82,896 1,826,652 517,649 1,774,402 52,148 53,190 236,252 223,545 15,684 2,631,631
% add
0 0 0 0 59 0 0 47 65 0 53 425 17 76 499 0 15 0 1,127 25 0 0 36 28 547 0 0 0 66 0 0 0
- 130 -
Assessment of Productive Efficiency of Airports
Table 7.6 Maximum possible passengers, aircraft movements and cargo throughput in 2001 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Total Case 2
31,668,450 50,445,737 33,313,785 18,105,026 25,557,671 100,892,176 21,572,361 24,791,783 44,798,994 42,669,451 45,702,538 75,858,500 19,880,293 16,548,642 15,184,332 11,496,714 37,026,987 68,676,642 24,006,781 72,645,066 7,324,557 43,567,735 32,864,676 48,378,795 2,402,635,134
Total passengers % add Case 1
0 50 248 55 281 50 628 317 253 74 29 280 1,781 1,309 0 1,485 37 98 83 286 0 63 8,052 204 116 31,668,450 35,058,695 9,567,651 11,713,225 8,302,839 75,694,280 2,963,223 7,388,932 12,703,676 37,847,971 35,439,051 22,365,971 1,057,150 2,025,856 15,184,332 725,140 27,036,073 50,861,183 13,088,997 22,911,625 7,324,557 36,378,058 1,918,917 16,355,131 1,280,691,015
% add
0 4 0 0 24 12 0 24 0 54 0 12 0 72 0 0 0 47 0 22 0 36 376 3 15
Case 2
471,008 649,297 506,842 611,557 340,376 1,364,091 408,409 598,280 630,363 760,845 635,807 890,494 405,037 396,255 206,988 569,460 435,600 813,708 468,616 883,528 378,903 626,503 434,268 649,412 35,958,516
Aircraft movements % add Case 1
0 40 248 55 120 50 522 786 117 63 48 432 3,107 129,894 0 857 9 98 83 224 0 219 1,463 471 67 471,008 520,651 145,564 395,653 191,525 1,063,376 119,309 234,992 291,117 699,371 553,310 506,271 116,501 144,738 206,988 160,486 400,635 730,017 255,499 452,824 378,903 574,174 119,598 268,481 22,772,468
% add
0 4 0 0 24 17 0 24 0 50 0 12 0 72 0 0 0 88 0 22 0 21 5 3 16
Case 2
1,639,760 476,155 263,581 917,575 921,894 1,944,048 178,978 182,456 526,102 873,728 418,347 739,927 159,600 119,595 134,689 28,190 435,452 1,261,208 263,955 702,641 14,849 389,536 311,198 456,142 35,596,398
Cargo (tons) % add Case 1
0 40 248 55 120 50 522 786 117 63 48 432 3,107 129,894 0 857 9 98 83 224 0 219 1,463 471 67 1,639,760 353,018 75,700 593,634 519,100 1,567,465 28,786 31,619 242,967 826,639 283,337 155,931 4,976 21,966 134,689 2,946 400,499 934,037 143,914 313,835 14,849 395,215 20,924 217,161 26,455,537
% add
0 4 0 0 24 21 0 53 0 54 0 12 0 23,776 0 0 0 47 0 45 0 223 5 172 24
Note: Cases 1 and 2 are with and without consideration of undesirable outputs, respectively. Aircraft movements include both delayed
and non-delayed flights. % add is the percentage increase from current level of the corresponding output.
- 131 -
Assessment of Productive Efficiency of Airports
Table 7.7 Maximum possible passengers, aircraft movements and cargo throughput in 2002
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM Case 2
47,082,564 15,526,628 76,876,128 27,861,486 47,983,915 10,403,788 56,209,470 36,782,412 48,704,535 34,377,363 57,657,096 26,662,417 91,791,939 132,833,476 112,639,482 33,702,666 29,275,757 65,692,610 24,937,323 56,988,954 76,876,128 31,395,381 33,066,340 42,657,320 58,997,163 56,223,843 21,986,679 17,041,398 57,345,382 57,657,096 20,262,756 11,141,594
Total passengers % add Case 1
665 216 0 891 111 125 196 252 106 410 177 107 157 151 247 15 72 233 210 234 127 2,471 568 42 69 0 0 1,073 458 116 17 0 6,151,129 4,914,539 76,876,128 2,810,791 35,647,876 4,620,683 19,012,529 17,578,887 29,952,090 6,741,354 37,620,386 13,941,695 38,229,992 91,816,824 54,507,100 30,258,481 23,131,236 19,749,905 11,577,573 31,569,526 51,048,711 1,221,138 7,432,943 40,849,751 42,492,287 56,223,843 21,986,679 1,453,412 13,342,966 27,674,633 17,371,036 11,141,594
% add
0 0 0 0 57 0 0 68 27 0 81 8 7 74 68 4 36 0 44 85 51 0 50 36 21 0 0 0 30 4 0 0
Case 2
800,139 352,657 889,966 416,577 828,929 365,229 820,678 780,673 645,088 434,829 667,474 446,776 1,101,697 1,544,739 1,324,897 468,057 482,401 817,160 755,604 662,482 889,966 427,372 431,550 571,199 830,639 645,424 362,439 791,612 666,695 667,475 354,961 398,769
Aircraft movements % add Case 1
214 27 0 184 111 125 169 210 42 70 37 107 123 102 170 15 72 152 207 78 95 416 246 99 67 0 0 126 248 131 17 0 254,874 277,267 889,966 146,555 610,900 162,211 304,921 420,585 576,384 255,630 582,466 252,786 558,511 1,302,154 815,124 485,889 379,747 323,726 354,136 524,229 743,794 82,883 187,219 584,615 602,418 645,424 362,439 350,603 248,280 403,408 304,304 398,769
% add
0 0 0 0 56 0 0 67 27 0 20 17 13 70 66 20 35 0 44 41 63 0 50 103 21 0 0 0 30 39 0 0
Case 2
439,749 2,253,301 734,083 251,104 820,220 89,502 676,506 314,417 446,975 320,056 550,562 45,457 867,812 1,353,342 1,081,237 981,039 283,600 1,047,424 147,858 577,566 734,083 285,553 304,887 2,264,634 575,288 1,779,855 32,223 120,470 546,955 550,562 30,689 3,390,800
Cargo (tons) % add Case 1
491 27 0 676 111 125 169 210 179 2,891 57 675 161 102 364 15 72 152 2,654 78 123 1,960 342 42 602 0 0 126 303 178 17 0 74,460 1,771,595 734,083 32,353 609,352 39,751 251,354 170,482 280,897 10,700 420,995 131,868 506,884 1,165,008 764,041 880,784 224,076 414,947 99,629 459,860 514,101 13,863 103,415 2,168,672 983,493 1,779,855 32,223 53,356 257,344 358,923 26,309 3,390,800
% add
0 0 0 0 57 0 0 68 75 0 20 2,150 52 74 228 4 36 0 1,756 42 56 0 50 36 1,100 0 0 0 89 81 0 0
- 132 -
Assessment of Productive Efficiency of Airports
Table 7.7 Maximum possible passengers, aircraft movements and cargo throughput in 2002 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Total Case 2
30,748,013 50,905,293 32,323,453 21,470,312 18,244,259 100,910,331 22,022,803 25,628,109 45,742,898 38,756,685 45,950,621 76,876,128 20,355,106 16,923,015 14,931,854 12,363,316 36,563,308 66,684,332 23,734,248 73,670,012 7,903,066 43,496,722 33,385,024 49,161,407 2,427,391,404
Total passengers % add Case 1
2 56 249 65 180 52 536 367 274 56 29 326 1,412 1,426 0 1,598 37 112 114 295 0 70 9,122 217 123 30,060,241 39,838,014 9,251,773 13,005,642 6,517,050 66,565,952 3,579,015 8,780,743 12,241,975 24,799,470 35,547,167 23,785,515 1,345,970 1,108,695 14,931,854 728,307 26,690,843 55,169,980 14,638,695 20,741,521 7,903,066 38,825,764 362,017 18,015,387 1,323,382,375
% add
0 22 0 0 0 0 3 60 0 0 0 32 0 0 0 0 0 75 32 11 0 52 0 16 22
Case 2
456,445 647,262 486,649 614,095 306,737 1,351,570 402,935 587,383 624,291 723,839 632,547 889,966 398,979 389,267 206,380 554,213 430,974 769,286 443,072 881,949 368,627 623,023 432,347 646,230 35,512,220
Aircraft movements % add Case 1
2 27 249 65 105 46 221 252 125 56 16 109 205 357 0 247 18 119 114 117 0 43 250 165 86 446,235 605,830 139,291 371,988 149,292 922,817 129,782 265,888 276,877 463,167 545,771 559,932 130,826 85,243 206,380 159,835 364,735 756,932 296,477 452,571 368,627 604,053 123,642 283,457 23,101,862
% add
0 19 0 0 0 0 3 59 0 0 0 32 0 0 0 0 0 115 43 11 0 38 0 16 21
Case 2
1,661,404 483,388 294,000 1,047,698 1,020,211 2,158,811 180,394 179,964 552,719 845,537 433,703 734,083 160,392 124,395 151,644 26,481 442,811 1,250,166 299,251 696,986 13,730 416,451 308,574 452,261 37,832,866
Cargo (tons) % add Case 1
2 51 249 65 105 46 449 906 125 56 45 424 3,450 168,002 0 835 18 112 114 222 0 223 2,228 393 70 1,624,242 383,425 84,150 634,643 496,547 1,473,980 52,005 54,643 245,134 541,039 298,945 184,943 4,518 74 151,644 2,832 374,753 1,034,300 184,570 297,407 13,730 352,447 13,257 268,010 27,496,714
% add
0 20 0 0 0 0 58 205 0 0 0 32 0 0 0 0 0 75 32 38 0 173 0 192 24
Note: Cases 1 and 2 are with and without consideration of undesirable outputs, respectively. Aircraft movements include both delayed
and non-delayed flights. % add is the percentage increase from current level of the corresponding output.
- 133 -
Assessment of Productive Efficiency of Airports
Table 7.8 Maximum possible passengers, aircraft movements and cargo throughput in 2003
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM Case 2
54,411,318 11,266,457 79,086,792 31,529,895 50,127,482 12,125,357 59,299,554 38,512,358 50,223,775 35,419,757 59,315,094 26,953,931 105,449,056 138,401,886 115,815,897 33,602,832 30,614,102 67,258,823 38,225,686 59,315,094 82,382,075 32,391,557 34,088,414 43,660,391 62,644,718 54,982,838 22,482,770 18,447,560 58,998,549 59,315,094 20,595,835 11,437,307
Total passengers % add Case 1
799.08 135.14 0.00 1,079.73 119.94 156.35 195.10 264.86 117.77 466.53 179.41 89.62 181.16 159.89 254.56 14.17 70.67 240.85 389.86 253.74 141.20 2,565.77 598.06 37.59 72.64 0.00 0.00 541.50 507.27 117.12 11.77 0.00 6,935,011 4,791,431 79,086,792 2,672,637 35,020,347 4,729,936 20,094,756 16,180,712 31,635,406 6,252,061 38,539,652 14,214,803 38,894,115 83,244,552 50,696,543 29,431,061 24,484,401 19,732,556 10,193,688 28,154,745 54,898,411 1,215,093 6,970,535 40,417,649 54,332,887 54,982,838 22,482,770 2,875,703 13,040,110 30,308,286 18,426,397 11,437,307
% add
14.59 0.00 0.00 0.00 53.66 0.00 0.00 53.29 37.17 0.00 81.55 0.00 3.70 56.32 55.20 0.00 36.49 0.00 30.63 67.91 60.74 0.00 42.74 27.37 49.74 0.00 0.00 0.00 34.22 10.94 0.00 0.00
Case 2
788,011 321,940 911,723 419,055 821,054 350,368 834,883 761,211 652,776 441,793 683,792 475,568 1,215,631 1,595,515 1,352,208 463,330 490,823 828,895 738,946 683,792 949,711 431,462 437,251 565,144 825,834 622,378 374,952 754,909 682,712 683,792 366,656 402,258
Aircraft movements % add Case 1
256.56 16.07 0.00 170.62 119.94 96.75 178.79 194.52 47.22 85.64 35.26 89.62 143.23 108.48 175.36 14.17 70.67 159.04 204.55 103.88 99.98 443.56 260.94 101.62 64.83 0.00 0.00 122.81 299.81 131.37 11.77 0.00 253,241 277,361 911,723 154,849 569,476 178,079 299,469 394,639 536,801 237,979 576,589 250,802 541,268 1,185,676 756,815 405,808 389,706 319,989 316,623 465,444 743,930 79,377 172,836 570,852 743,556 622,378 374,952 338,807 229,168 380,232 328,035 402,258
% add
14.59 0.00 0.00 0.00 52.55 0.00 0.00 52.69 21.07 0.00 14.05 0.00 8.30 54.93 54.11 0.00 35.51 0.00 30.49 38.77 56.65 0.00 42.67 103.66 48.41 0.00 0.00 0.00 34.21 28.66 0.00 0.00
Case 2
331,992 2,439,876 798,501 233,382 798,572 87,856 656,757 282,033 485,352 347,759 598,876 37,820 1,064,668 1,397,377 1,157,774 998,619 267,005 1,092,962 56,298 598,876 831,772 309,945 331,134 2,238,198 523,744 1,833,300 28,402 113,352 594,923 598,876 26,160 3,390,515
Cargo (tons) % add Case 1
363.68 16.07 0.00 582.72 119.94 96.75 178.79 194.52 246.47 3,130.16 52.50 555.01 227.24 109.32 425.67 14.17 70.67 159.04 874.86 109.87 117.78 2,728.74 368.70 37.59 537.52 0.00 0.00 122.81 335.24 210.24 12.44 0.00 133,954 2,102,025 798,501 34,184 943,554 44,654 235,576 272,944 334,987 10,766 449,437 5,774 884,820 1,043,533 710,246 874,641 226,674 421,930 155,183 399,990 683,510 10,957 104,553 2,071,962 510,720 1,833,300 28,402 50,873 265,216 860,080 23,266 3,390,515
% add
87.09 0.00 0.00 0.00 159.87 0.00 0.00 185.03 139.13 0.00 14.45 0.00 171.96 56.32 222.48 0.00 44.89 0.00 2,587.15 40.17 78.96 0.00 47.99 27.37 521.67 0.00 0.00 0.00 94.03 345.55 0.00 0.00
- 134 -
Assessment of Productive Efficiency of Airports
Table 7.8 Maximum possible passengers, aircraft movements and cargo throughput in 2003 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Total Case 2
34,479,734 54,502,729 33,425,071 23,456,948 17,639,338 103,836,235 24,266,049 31,753,180 47,653,087 39,166,142 49,730,463 79,086,792 22,273,756 20,361,442 15,260,791 13,319,297 38,400,771 67,104,325 25,514,563 75,830,984 8,535,130 47,057,418 34,412,038 50,687,729 2,556,136,266
Total passengers % add Case 1
16.50 64.16 260.35 73.13 169.39 49.39 606.15 428.27 284.43 58.75 32.93 454.33 1,535.66 1,533.04 0.00 1,669.39 43.52 128.92 138.95 310.64 0.00 130.37 8,644.45 226.52 133.50 29,595,618 41,741,976 10,542,505 13,548,363 7,258,006 69,508,672 3,849,419 10,029,678 12,395,938 39,631,312 37,412,165 18,913,688 1,361,758 1,246,842 15,260,791 752,762 26,755,888 52,079,506 10,677,903 21,972,152 8,535,130 35,870,254 393,530 21,158,079 1,346,865,126
% add
0.00 25.72 13.66 0.00 10.85 0.00 12.02 66.86 0.00 60.64 0.00 32.57 0.00 0.00 0.00 0.00 0.00 77.67 0.00 18.98 0.00 75.60 0.00 36.30 23.03
Case 2
486,310 657,595 494,806 593,629 298,085 1,372,264 398,907 589,762 634,652 708,879 635,974 911,723 393,148 381,163 203,285 526,873 447,256 767,885 473,312 900,616 350,074 621,218 438,355 654,359 35,868,503
Aircraft movements % add Case 1
16.50 28.84 260.35 73.13 103.59 47.76 228.66 243.50 137.65 58.75 17.39 152.32 209.09 309.55 0.00 245.52 26.07 129.55 138.95 124.90 0.00 63.58 290.40 180.12 90.98 417,423 595,117 156,040 342,871 162,248 928,691 135,951 284,004 267,052 700,772 541,771 478,550 127,197 93,068 203,285 152,485 354,770 600,665 198,082 476,174 350,074 568,680 112,284 317,570 22,573,542
% add
0.00 16.60 13.64 0.00 10.82 0.00 12.01 65.41 0.00 56.94 0.00 32.44 0.00 0.00 0.00 0.00 0.00 79.56 0.00 18.91 0.00 49.74 0.00 35.95 20.19
Case 2
1,907,475 529,522 291,276 1,034,279 1,056,052 2,232,328 174,884 254,710 568,616 832,637 415,223 798,501 157,215 95,380 135,547 24,526 443,031 1,312,916 259,549 757,846 12,050 378,835 335,175 491,146 39,051,395
Cargo (tons) % add Case 1
16.50 67.58 260.35 73.13 103.59 47.76 441.72 1,291.86 137.65 58.75 44.00 557.01 3,340.92 92,501.67 0.00 768.17 26.07 128.92 138.95 249.45 0.00 227.79 1,661.85 425.53 363.68 1,637,278 370,000 253,278 597,383 574,965 1,510,746 36,821 175,200 239,265 842,526 288,350 294,139 4,569 103 135,547 2,825 351,418 1,682,768 108,622 402,086 12,050 299,115 19,024 649,238 30,404,043
% add
0.00 17.09 213.34 0.00 10.85 0.00 14.06 857.38 0.00 60.64 0.00 142.02 0.00 0.00 0.00 0.00 0.00 193.41 0.00 85.40 0.00 158.81 0.00 594.69 34.54
Note: Cases 1 and 2 are with and without consideration of undesirable outputs, respectively. Aircraft movements include both delayed
and non-delayed flights. % add is the percentage increase from current level of the corresponding output.
- 135 -
Assessment of Productive Efficiency of Airports
7.2
Lumpiness of airport investment
Due to the fact that airport inputs are relatively fixed and air traffic tends to grow over a long period of time, one may expect a rather stable operational efficiency during the analysis. In most cases, this is true as long as there is no asset selling or drastic change in air traffic. Nevertheless, it is possible to see sharp decline in efficiency at some point in time due to the opening of a new facility. In the early years of an asset’s life it is likely that excess capacity will prevail and hence show up as a contributing factor to low annual productivity. In the later years of an asset’s life it might show up through an impact on high levels of congestion and hence a shortage of capacity which can reduce output and hence affect productivity in a different way (Hooper and Hensher, 1997). In his study, Parker (1999) explained that the sharp decline during 1991/92 at Stansted, London was associated with the opening of a new terminal in that year, leading to further excess capacity. The technical efficiency can be expected to rise over time and favor later airports over the earlier ones. Similarly, a newly delivered runway may therefore have a capacity that far exceeds realized demand. The lumpiness of runway investments can signify a productivity drop in the early years after the investment, as results indicate in the case of Detroit Metropolitan Wayne County (DTW) which opened its 6th runway on December 11, 2001. During 2000 – 2001, its productivity scores in Table 7.1 are rather stable (i.e., 1.2993 and 1.1319), but downgraded significantly in 2002 and 2003 (i.e., 1.6990 and 1.7536) after the new runway was completed. The same situation occurs at George Bush Intercontinental (IAH) which expanded and extended runway 15R/33L from 6038’ x 100’ to 10000’ x 150’. The scores are rather stable during
- 136 -
Assessment of Productive Efficiency of Airports
2000 – 2001, i.e., 0.8696 and 0.8391) and downgraded afterwards (i.e., 0.9481 and 0.9998 in 2002 and 2003 respectively). Therefore, once the sharp drop is detected, it should not be presumed that it is due to poor management.
7.3 Changes in productivity over time
In the past only Gillen and Lall (1998) studied the productivity growth of U.S. airports during 1989 – 1993 by computing the Malmquist index. As reviewed in Chapter 2, the study did not consider any undesirable outputs. It was explained in Chapter 3 that Malmquist index is not appropriate when there is an undesirable output. Here, the Luenberger (L) productivity index (equation 3.20) is computed with more comprehensive output measures during recent years. Table 7.9 and 7.10 shows the results for Cases 1 and 2 respectively. Table 7.9 shows computed changes in productivity for each airport and the overall average for 56 airports. Along with Luenberger index, its two components, i.e., efficiency change (LEFFCH) and technical change (LTECHCH) are also shown. Note that these indexes signal improvements with values greater than zero, and declines in productivity with values less than zero. The zero value indicates no productivity change between two years.
- 137 -
Assessment of Productive Efficiency of Airports
Table 7.9 Luenberger productivity indexes, Case 1
Airport code
ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM
L
0.079 0.000 0.310 -0.002 -0.079 -0.001 0.015 -0.127 0.040 0.000 -0.170 -0.220 -0.019 0.023 0.060 0.011 0.102 0.000 -0.073 -0.048 0.104 0.077 0.042 -0.116 0.235 0.000 0.475 -0.014 -0.025 -0.015 -0.002 0.000
2000 – 2001 LEFFCH LTECHCH
0.203 0.000 0.000 0.000 -0.086 0.011 0.000 -0.042 0.233 0.000 0.092 -0.317 0.269 0.135 0.292 0.000 0.444 0.000 0.035 0.225 0.518 0.000 0.170 0.033 0.587 0.000 0.000 0.000 0.089 0.262 0.012 0.000 -0.124 0.000 0.310 -0.002 0.007 -0.012 0.015 -0.085 -0.193 0.000 -0.262 0.096 -0.288 -0.112 -0.232 0.011 -0.342 0.000 -0.108 -0.273 -0.414 0.077 -0.127 -0.149 -0.353 0.000 0.475 -0.014 -0.114 -0.277 -0.014 0.000
L
0.009 0.000 0.000 0.000 -0.023 0.000 0.096 -0.068 -0.003 0.012 0.219 0.091 0.033 -0.001 -0.086 0.124 -0.045 0.000 0.001 -0.064 -0.102 0.026 -0.029 0.089 0.017 0.000 0.001 0.000 -0.062 -0.009 0.000 0.000
2001 - 2002 LEFFCH LTECHCH
0.000 0.000 0.000 0.000 0.019 0.000 0.000 -0.207 -0.152 0.000 0.110 0.244 -0.046 -0.125 -0.277 -0.036 -0.207 0.000 -0.139 -0.162 -0.506 0.000 -0.137 -0.088 -0.209 0.000 0.000 0.000 -0.105 -0.038 0.000 0.000 0.009 0.000 0.000 0.000 -0.042 0.000 0.096 0.139 0.150 0.012 0.109 -0.153 0.079 0.124 0.190 0.160 0.162 0.000 0.140 0.098 0.404 0.026 0.108 0.176 0.226 0.000 0.001 0.000 0.042 0.029 0.000 0.000
L
-0.090 0.000 0.000 0.029 -0.029 0.000 0.000 0.035 -0.021 -0.012 0.042 0.060 -0.008 0.055 -0.052 0.050 0.009 0.000 0.007 -0.055 -0.032 0.000 -0.029 0.043 -0.111 0.000 0.173 -0.001 -0.072 -0.021 0.000 0.000
2002 - 2003 LEFFCH LTECHCH
-0.146 0.000 0.000 0.000 0.034 0.000 0.000 0.148 0.055 0.000 0.058 0.083 0.035 0.175 0.126 0.036 -0.007 0.000 0.134 0.014 -0.102 0.000 0.073 0.091 -0.284 0.000 0.000 0.000 -0.044 -0.071 0.000 0.000 0.056 0.000 0.000 0.029 -0.063 0.000 0.000 -0.114 -0.076 -0.012 -0.016 -0.023 -0.043 -0.120 -0.178 0.014 0.016 0.000 -0.128 -0.069 0.070 0.000 -0.103 -0.048 0.173 0.000 0.173 -0.001 -0.028 0.050 0.000 0.000
- 138 -
Assessment of Productive Efficiency of Airports
Table 7.9 Luenberger productivity indexes, Case 1 (Continued)
Airport code
MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Average index Number of regress Number of no change Number of progress
L
0.000 -0.001 -0.012 -0.013 -0.077 0.264 0.162 -0.078 -0.015 0.204 0.265 -0.016 0.023 -0.056 0.000 -0.010 -0.026 -0.105 0.000 -0.061 0.000 -0.109 -0.070 0.066 0.018 28 9 19
2000 - 2001 LEFFCH LTECHCH
0.000 0.144 0.155 0.000 -0.239 0.138 0.294 0.450 0.000 0.175 0.000 0.105 0.152 -0.315 0.000 0.000 0.000 0.096 0.000 0.467 0.000 -0.219 -0.037 0.410 0.088 7 21 28 0.000 -0.145 -0.168 -0.013 0.162 0.126 -0.132 -0.527 -0.015 0.029 0.265 -0.121 -0.130 0.259 0.000 -0.010 -0.026 -0.202 0.000 -0.528 0.000 0.109 -0.034 -0.344 -0.070 34 9 13
L
0.000 -0.030 0.046 0.000 0.069 0.170 -0.017 -0.138 0.000 0.047 0.020 -0.066 0.001 0.314 0.000 -0.002 0.000 0.014 -0.158 0.096 0.000 -0.087 0.191 -0.015 0.012 19 15 22
2001 - 2002 LEFFCH LTECHCH
0.000 -0.161 0.000 0.000 0.239 0.122 -0.033 -0.357 0.000 0.541 0.000 -0.198 0.000 0.725 0.000 0.000 0.000 -0.285 -0.317 0.105 0.000 -0.176 0.051 -0.133 -0.035 23 24 9 0.000 0.131 0.046 0.000 -0.170 0.047 0.017 0.219 0.000 -0.495 0.020 0.132 0.001 -0.411 0.000 -0.002 0.000 0.299 0.158 -0.009 0.000 0.089 0.139 0.119 0.047 7 15 34
L
0.000 -0.002 -0.068 0.000 -0.054 0.134 -0.047 -0.009 -0.003 -0.116 0.008 -0.138 0.002 0.000 0.000 -0.018 0.000 -0.056 0.019 -0.056 0.000 -0.171 -0.005 -0.093 -0.013 27 15 14
2002 - 2003 LEFFCH LTECHCH
0.000 0.029 -0.137 0.000 -0.108 0.000 -0.087 -0.067 0.000 -0.606 0.000 -0.006 0.000 0.000 0.000 0.000 0.000 -0.023 0.317 -0.078 0.000 -0.116 0.000 -0.200 -0.012 16 25 15 0.000 -0.031 0.068 0.000 0.054 0.134 0.040 0.058 -0.003 0.491 0.008 -0.132 0.002 0.000 0.000 -0.018 0.000 -0.033 -0.298 0.022 0.000 -0.056 -0.005 0.107 -0.001 23 15 18
Note: Negative index indicates regressed productivity. Zero value means that there is no change in productivity between two
years. Positive index indicates productivity growth.
- 139 -
Assessment of Productive Efficiency of Airports
According to the results, the airport system on average had productivity gains in two periods during 2000 – 2002 and productivity loss in 2003. The amount of changes regardless of progress or regress is rather low. Between 2000 and 2001, the overall average rise in efficiency was 1.8 percent; and continued to increase by 1.2 percent in the next period before falling down slightly 1.3 percent during 2002 – 2003. The results are suggesting that productivity of U.S. airports during 2000 – 2003 is more or less the same. At the airport level, only five airports, i.e., Newark Liberty International (EWR), LaGuardia (LGA), O'Hare International (ORD), Phoenix Sky Harbor International (PHX), and Pensacola Regional (PNS) show progress in all periods. For all airports and all periods, it is found that there were productivity losses in 74 cases; productivity remains the same in 39 cases and there were productivity gains in 55 cases. About 32% of all cases show productivity gains. The productivity loss at many airports, especially between 2000 and 2001, may be associated with the September 11 terrorist attacks which shook aviation industry worldwide. Decomposition of Luenberger index into efficiency change (LEFFCH) and technical change (LTECHCH) can help explain source of productivity gain or loss. The equations for computing LEFFCH and LTECHCH are given in (3.21) and (3.22) respectively. An airport which has been efficient in time period t and t + 1 , will naturally show no change in relative efficiency. Only Covington/Cincinnati/Northern Kentucky International (CVG) achieved productivity gains in all time periods. For the sample as a whole, efficiency gains occur between 2000 and 2001; then drop afterwards in 2002 and 2003. Between 2000 and 2001 the overall average rise in efficiency was 8.8 percent. This
- 140 -
Assessment of Productive Efficiency of Airports
was followed by a 3.5 % and 1.2 % drops in the two subsequent periods. For all airports and all periods, it is found that there were productivity losses in 46 cases; productivity remains the same in 70 cases and there were productivity gains in 52 cases. About 33% of all cases show productivity gains. The computed LTECHCH is shown next to LEFFCH column in Table 7.9. LTECHCH measures the average shifts in the efficient frontier from time period t to time period t + 1 . This corresponds to the term in equation (3.22). The results show productivity gain in one period, i.e., 2001 – 2002; and two periods, i.e., 2000 - 2001 and 2002 - 2003 with productivity loss. Between 2000 and 2001, high negative value of LTECHCH (-0.070) indicates that the efficient frontier shifted backward. In other words, for a given level of inputs in 2000, the airport system produces lower outputs in 2001 than in 2000. For all airports and all periods, it is found that there were productivity losses in 64 cases; productivity remains the same in 39 cases and there were productivity gains in 65 cases. About 33% of all cases show productivity gains. About 39% of all cases show productivity gains. In conclusion, the productivity gains between 2000 and 2001 are mainly from efficiency change (LEFFCH = 0.088) which compensates the productivity loss from frontier shifted backward (LTECHCH = -0.070). The situation is opposite in 2001/2002 period where productivity gains resulted from frontier shift (LTECHCH = 0.047). Between 2002 and 2003 both efficiency loss and frontier backward shift collectively contribute to the overall productivity loss (L = -0.013).
- 141 -
Assessment of Productive Efficiency of Airports
Productivity indexes for Case 2 are also computed in order to analyze the impact of considering undesirable outputs in the assessment. The results are shown in Table 7.10. For convenient comparisons, productivity indexes in Cases 1 and 2 are presented side by side in Table 7.11. For individual airports, the results do not show any recognizable pattern. Moreover, the overall picture of the airport system indicates a different conclusion. Instead of showing productivity gains between 2000 and 2001 (L = 0.018), Case 2 shows the opposite result, i.e., productivity loss (L = -0.041). The classification of productivity changes is also very different in all periods. It clearly shows the two sets of results are drastically different. Again, this confirms previous findings that ignoring undesirable outputs in the assessment really creates problematic results. Several statistical tests are performed on the Luenberger indexes to check if the difference between Cases 1 and 2 are statistical different. Table 7.12 provides the results from paired-sample t-tests which strongly support the assertion of differences. To avoid the restricted assumptions of the t-test, the non-parametric Wilcoxon signed-rank test and sign test are also performed. The results are shown in Table 7.13. They confirm that the difference in Luenberger productivity indexes between Cases 1 and 2 is significant. There is one exception in period 2002/03 where Z-statistics from both Wilcoxon signed rank test (Z = -2.213) and sign test (Z = 0.099) are not significant at 95%.
- 142 -
Assessment of Productive Efficiency of Airports
Table 7.10 Luenberger productivity indexes, Case 2
Airport code
ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM
L
0.125 0.014 0.000 -0.098 -0.201 -0.041 0.040 -0.333 0.031 0.042 -0.325 -0.319 -0.165 -0.137 -0.135 -0.119 -0.032 -0.203 -0.068 -0.224 -0.053 0.082 -0.327 -0.177 -0.078 0.000 0.000 -0.095 -0.132 -0.139 -0.055 0.000
2000 – 2001 LEFFCH LTECHCH
0.275 0.181 0.000 0.228 -0.064 0.047 0.191 -0.239 0.204 0.275 -0.174 -0.206 -0.042 0.026 0.167 -0.116 0.120 0.067 0.023 -0.053 0.030 0.637 0.054 -0.126 -0.045 0.000 0.000 -0.028 0.225 -0.087 0.048 0.000 -0.150 -0.167 0.000 -0.326 -0.138 -0.088 -0.151 -0.094 -0.173 -0.234 -0.151 -0.113 -0.122 -0.163 -0.302 -0.003 -0.152 -0.270 -0.090 -0.170 -0.084 -0.555 -0.381 -0.051 -0.033 0.000 0.000 -0.067 -0.358 -0.052 -0.104 0.000
L
0.158 -0.029 0.000 -0.044 -0.159 0.061 -0.112 -0.416 -0.018 0.087 0.350 -0.106 0.050 -0.073 -0.569 -0.017 -0.005 0.039 -0.047 -0.098 -0.114 -0.533 -0.251 0.070 0.010 0.000 0.000 -0.043 -0.305 -0.092 0.096 0.078
2001 – 2002 LEFFCH LTECHCH
0.189 -0.221 0.000 -0.021 -0.101 0.117 -0.083 -0.373 -0.010 0.093 0.352 -0.105 0.054 -0.073 -0.567 0.051 0.035 0.099 0.007 -0.096 -0.109 -0.503 -0.235 0.048 0.018 0.000 0.000 -0.034 -0.303 -0.149 0.122 0.000 -0.031 0.192 0.000 -0.023 -0.059 -0.056 -0.029 -0.043 -0.007 -0.006 -0.002 -0.002 -0.004 0.001 -0.002 -0.068 -0.041 -0.060 -0.054 -0.003 -0.005 -0.029 -0.016 0.022 -0.008 0.000 0.000 -0.009 -0.002 0.058 -0.026 0.078
L
-0.477 0.073 0.012 0.153 -0.046 0.189 -0.061 0.066 -0.039 -0.127 0.052 0.191 -0.154 0.000 0.001 0.009 0.038 -0.018 -0.045 -0.194 -0.004 -0.229 -0.106 0.034 0.014 0.000 0.008 -0.077 -0.425 0.041 0.062 0.000
2002 - 2003 LEFFCH LTECHCH
-0.426 0.111 0.000 0.136 -0.085 0.284 -0.096 0.156 -0.056 -0.155 0.019 0.175 -0.206 -0.066 -0.055 0.012 0.012 -0.066 0.023 -0.261 -0.052 -0.279 -0.152 0.049 0.024 0.000 0.000 0.030 -0.514 -0.008 0.049 0.000 -0.051 -0.038 0.012 0.016 0.039 -0.095 0.036 -0.090 0.017 0.028 0.033 0.015 0.052 0.066 0.056 -0.003 0.027 0.048 -0.068 0.066 0.048 0.050 0.046 -0.015 -0.010 0.000 0.008 -0.107 0.089 0.049 0.013 0.000
- 143 -
Assessment of Productive Efficiency of Airports
Table 7.10 Luenberger productivity indexes, Case 2 (Continued)
Airport code
MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Average index Number of regress Number of no change Number of progress
L
-0.023 -0.051 -0.195 -0.146 -0.132 -0.066 -0.155 -0.081 -0.178 -0.058 -0.201 0.014 -0.040 0.056 0.000 -0.148 -0.093 -0.264 -0.124 0.017 -0.007 -0.021 -0.675 -0.141 -0.104 42 4 10
2000 – 2001 LEFFCH LTECHCH
0.030 0.054 0.122 -0.018 0.012 0.035 0.162 0.118 -0.029 0.022 -0.149 0.326 0.263 0.345 0.000 -0.060 -0.005 -0.211 0.032 0.286 0.000 0.078 -0.236 0.100 0.052 18 6 32 -0.054 -0.105 -0.317 -0.128 -0.144 -0.101 -0.317 -0.199 -0.149 -0.080 -0.052 -0.312 -0.303 -0.289 0.000 -0.088 -0.088 -0.053 -0.156 -0.268 -0.007 -0.099 -0.439 -0.242 -0.156 51 4 1
L
-0.011 0.016 -0.063 0.011 0.088 0.038 0.171 -0.416 -0.093 -0.009 -0.016 -0.124 0.378 0.067 0.000 -0.014 -0.097 -0.139 -0.263 0.202 -0.007 -0.112 0.304 -0.172 -0.041 33 4 19
2001 - 2002 LEFFCH LTECHCH
-0.023 0.020 -0.012 -0.105 0.145 0.031 0.216 -0.355 -0.089 0.066 -0.010 -0.123 0.427 0.150 0.000 0.081 -0.094 -0.137 -0.301 0.207 0.000 -0.104 0.320 -0.160 -0.029 28 6 22 0.011 -0.004 -0.051 0.116 -0.057 0.007 -0.045 -0.061 -0.003 -0.075 -0.006 -0.001 -0.049 -0.083 0.000 -0.095 -0.003 -0.002 0.038 -0.004 -0.007 -0.008 -0.016 -0.013 -0.012 43 4 9
L
-0.132 0.002 -0.024 -0.064 0.001 0.013 -0.112 0.012 -0.078 -0.050 -0.009 -0.373 -0.086 0.380 0.000 -0.163 -0.036 -0.098 -0.184 -0.036 -0.012 -0.215 -0.356 -0.118 -0.050 32 3 21
2002 - 2003 LEFFCH LTECHCH
-0.142 -0.013 -0.110 -0.081 0.019 -0.013 -0.079 0.084 -0.122 -0.025 -0.015 -0.429 -0.041 0.471 0.000 0.012 -0.079 -0.169 -0.254 -0.082 0.000 -0.210 -0.407 -0.152 -0.058 33 6 17 0.010 0.016 0.086 0.017 -0.018 0.026 -0.033 -0.072 0.044 -0.026 0.006 0.056 -0.045 -0.091 0.000 -0.175 0.043 0.072 0.071 0.046 -0.012 -0.004 0.051 0.034 0.008 18 3 35
Note: The negative index indicates regressed productivity. Zero value means that there is no change in the productivity
between two years. The positive index indicates productivity growth.
- 144 -
Assessment of Productive Efficiency of Airports
Table 7.11 Comparisons of Luenberger productivity indexes between Cases 1 and 2
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM
2000 - 2001 Case 1 Case 2 0.079 0.125 0.000 0.014 0.310 0.000 -0.002 -0.098 -0.079 -0.201 -0.001 -0.041 0.015 0.040 -0.127 -0.333 0.040 0.031 0.000 0.042 -0.170 -0.325 -0.220 -0.319 -0.019 -0.165 0.023 -0.137 0.060 -0.135 0.011 -0.119 0.102 -0.032 0.000 -0.203 -0.073 -0.068 -0.048 -0.224 0.104 -0.053 0.077 0.082 0.042 -0.327 -0.116 -0.177 0.235 -0.078 0.000 0.000 0.475 0.000 -0.014 -0.095 -0.025 -0.132 -0.015 -0.139 -0.002 -0.055 0.000 0.000
2001 - 2002 Case 1 Case 2 0.009 0.158 0.000 -0.029 0.000 0.000 0.000 -0.044 -0.023 -0.159 0.000 0.061 0.096 -0.112 -0.068 -0.416 -0.003 -0.018 0.012 0.087 0.219 0.350 0.091 -0.106 0.033 0.050 -0.001 -0.073 -0.086 -0.569 0.124 -0.017 -0.045 -0.005 0.000 0.039 0.001 -0.047 -0.064 -0.098 -0.102 -0.114 0.026 -0.533 -0.029 -0.251 0.089 0.070 0.017 0.010 0.000 0.000 0.001 0.000 0.000 -0.043 -0.062 -0.305 -0.009 -0.092 0.000 0.096 0.000 0.078
2002 - 2003 Case 1 Case 2 -0.090 -0.477 0.000 0.073 0.000 0.012 0.029 0.153 -0.029 -0.046 0.000 0.189 0.000 -0.061 0.035 0.066 -0.021 -0.039 -0.012 -0.127 0.042 0.052 0.060 0.191 -0.008 -0.154 0.055 0.000 -0.052 0.001 0.050 0.009 0.009 0.038 0.000 -0.018 0.007 -0.045 -0.055 -0.194 -0.032 -0.004 0.000 -0.229 -0.029 -0.106 0.043 0.034 -0.111 0.014 0.000 0.000 0.173 0.008 -0.001 -0.077 -0.072 -0.425 -0.021 0.041 0.000 0.062 0.000 0.000
- 145 -
Assessment of Productive Efficiency of Airports
Table 7.11 Comparisons of Luenberger productivity indexes (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Average index Number of regress Number of no change Number of progress
2000 - 2001 Case 1 Case 2 0.000 -0.023 -0.001 -0.051 -0.012 -0.195 -0.013 -0.146 -0.077 -0.132 0.264 -0.066 0.162 -0.155 -0.078 -0.081 -0.015 -0.178 0.204 -0.058 0.265 -0.201 -0.016 0.014 0.023 -0.040 -0.056 0.056 0.000 0.000 -0.010 -0.148 -0.026 -0.093 -0.105 -0.264 0.000 -0.124 -0.061 0.017 0.000 -0.007 -0.109 -0.021 -0.070 -0.675 0.066 -0.141 0.018 -0.104 28 42 9 4 19 10
2001 - 2002 Case 1 Case 2 0.000 -0.011 -0.030 0.016 0.046 -0.063 0.000 0.011 0.069 0.088 0.170 0.038 -0.017 0.171 -0.138 -0.416 0.000 -0.093 0.047 -0.009 0.020 -0.016 -0.066 -0.124 0.001 0.378 0.314 0.067 0.000 0.000 -0.002 -0.014 0.000 -0.097 0.014 -0.139 -0.158 -0.263 0.096 0.202 0.000 -0.007 -0.087 -0.112 0.191 0.304 -0.015 -0.172 0.012 -0.041 19 33 15 4 22 19
2002 - 2003 Case 1 Case 2 0.000 -0.132 -0.002 0.002 -0.068 -0.024 0.000 -0.064 -0.054 0.001 0.134 0.013 -0.047 -0.112 -0.009 0.012 -0.003 -0.078 -0.116 -0.050 0.008 -0.009 -0.138 -0.373 0.002 -0.086 0.000 0.380 0.000 0.000 -0.018 -0.163 0.000 -0.036 -0.056 -0.098 0.019 -0.184 -0.056 -0.036 0.000 -0.012 -0.171 -0.215 -0.005 -0.356 -0.093 -0.118 -0.013 -0.050 27 32 15 3 14 21
Note: The negative index indicates regressed productivity. Zero value means that there is
no change in the productivity between two years. The positive index indicates productivity growth.
- 146 -
Assessment of Productive Efficiency of Airports
Table 7.12 Comparisons of Luenberger productivity indexes by paired sample t-test
Paired-sample t-test Pair 1: 2000/01 Pair 2: 2001/02 Pair 3: 2002/03 Pair 4: 2000 – 03 Mean 0.1219 0.0531 0.0373 0.0708
Paired differences Cases 1 and 2 95% confidence interval of the Std. Std. error difference deviation mean Lower Upper 0.1458 0.0194 0.0829 0.1610 0.1527 0.0204 0.0122 0.0940 0.1287 0.0172 0.0029 0.0718 0.1466 0.0113 0.0484 0.0931
t 6.2605 2.6014 2.1741 6.2607
Table 7.13 Comparisons of Luenberger productivity indexes by nonparametric tests
Nonparametric paired test A. Wilcoxon Signed-Rank test Pair 1: years 2000/01 Pair 2: years 2001/02 Pair 3: years 2002/03 Pair 4: years 2000 - 03 Pair 1: years 2000/01 Pair 2: years 2001/02 Pair 3: years 2002/03 Pair 4: years 2000 - 03
Z -5.307a -2.598a -2.213a -6.027a -4.395 -2.747 -1.648 -5.234
B. Sign test
Asymptotic significance (2-tailed) 0.000 0.009 0.026 0.000 0.000 0.006 0.099 0.000
a
Based on positive ranks.
- 147 -
Assessment of Productive Efficiency of Airports
7.4
Scenario analysis
Having shown that ignoring undesirable outputs while assessing productivity of airports can cause drastically different interpretations (sometimes unrealistic), it is interesting to analyze the impact of other measures on productivity. The results will provide more insights on the sensitivity of estimated productivity. Many different sets of possible input and output measures can be analyzed. Pathomsiri, Haghani, Windle and Dresner (2006) have analyzed the impact of cargo throughput and a single undesirable output – delayed flights, on the productivity of 56 U.S. airports (the same dataset that is used in this dissertation). Their work notes that the addition of the undesirable output (delayed flights) into the model made much more difference than the addition of another desirable output (cargo throughput); i.e., higher number of efficient airports are identified. The results suggest that consideration of undesirable output is at least as important as the consideration of additional desirable outputs in determining relative productivity of airports. Here let’s consider the effects of an input measure, i.e., land area. Recall that the preceding sections consider land area as an operational input along with number of runways and runway area. The resulting estimated productivity indicates the operational efficiency of airside operation from utilizing these three inputs. One might argue that land area is not being used solely for airside operation which is probably true. Nowadays many airports are more enthusiastic to provide non-aeronautical services including concessions, rentals and car parking (Francis and Humphreys, 2001; Francis, Humphreys and Fry (2002); Hooper and Hensher, 1997; Humphreys, 1999; Humphreys and Francis, 2002; Nyshadham and Rao, 2000; Oum and Yu, 2004; Oum, Yu and Fu,
- 148 -
Assessment of Productive Efficiency of Airports
2003). In some airports such as Honolulu, Vancouver and Sydney, non-aeronautical revenues account for as much as 70% of their total revenues (Oum and Yu, 2004). Airport management has become more diversified (Francis and Humphreys, 2001). Land area is being utilized beyond just aeronautical services. Furthermore, airport may own more excessive land for other reasons such as noise abatement, planned future expansion, land appreciation, and other investment. In such cases, inclusion of land area in the set of input may give biased results in favor of airports with less land area and limited nonaeronautical activities. To analyze the impact of land area, Case 3 is then set up as follow: Case 3: without land area as an input Input = {number of runway, runway area} Desirable outputs = {non-delayed flights, passengers, cargo} Undesirable outputs = {delayed flights, time delays} The directional output distance function in (3.7) is then solved again 56 times, each for an individual airport to estimate the efficiency score, ? . Table 7.14 shows the results. The estimated efficiency scores in Case 1 are copied from Table 7.1 for comparison purposes. There are several interesting observations for discussion. First, on the efficiency scores, average scores from both cases (at the bottom of the Table) are not much different. The difference is relatively much smaller than the comparison between Cases 1 (with undesirable outputs) and 2 (without desirable outputs). This is because at the individual airports, scores are not different. Several
- 149 -
Assessment of Productive Efficiency of Airports
airports even earned identical scores regardless of cases. In 2003, 45 airports have identical pairs of scores. Only 11 airports (i.e., ABQ, BOS, DCA, FLL, HOU, LGB, MDW, MSY, ONT, and ORF) show different scores. For all of them, the productivity is downgraded. These 11 airports are relatively smaller in land size than those 45. The range of area is between 650 (at MDW) and 2039 (at ABQ) acres. Inclusion of land area as an input clearly favors airports with relatively smaller size. Second, on the classification of efficient airports, the results show that all efficient airports in Case 3 are also efficient in Case 1. In other words, the set of efficient airports in Case 3 is a subset of efficient airports in Case 1. Excluding land area never decreases inefficiency level in Case 1. Inefficient airports in Case 1 are still identified as (more) inefficient in Case 3. Note that one should not expect this finding to be always true in other applications. It is possible that excluding (or including) an input measure from the consideration may decrease (or increase) inefficiency levels if that input measure is a dominant one. Third, on the number of efficient airports, the annual figures are not much different. Case 1 identifies 23, 29, 29 and 28 efficient airports whereas Case 3 identifies 22, 24, 24 and 25 airports in 2000, 2001, 2002 and 2003 respectively. Both cases identify almost the same set of efficient airports. Case 1 with more total measures (number of inputs plus outputs) identifies more efficient airports. This finding is in line with previous observations (Parker, 1999; Salazar de la Cruz, 1999; Pathomsiri, Haghani, Dresner and Windle, 2006b, 2006c; Pathomsiri, Haghani, Windle and Dresner, 2006).
- 150 -
Assessment of Productive Efficiency of Airports
Table 7.14 Efficiency scores for Cases 1 and 3
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM 2000 Case 1 Case 3 0.2034 0.5771 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5044 0.5048 0.0108 0.1478 0.0000 0.0000 0.4323 0.4763 0.3502 0.3502 0.0000 0.0000 0.4044 0.4044 0.0102 0.0570 0.2956 0.2956 0.7486 0.7486 0.6938 0.6938 0.0000 0.0000 0.5941 0.5941 0.0000 0.0000 0.3367 0.4391 0.4777 0.4777 0.5179 0.5179 0.0000 0.0000 0.5331 0.5331 0.3099 0.3099 0.5919 0.6252 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.2827 0.2827 0.2619 0.2619 0.0118 0.6070 0.0000 0.0000 2001 Case 1 Case 3 0.0193 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5901 0.5901 0.0500 0.0000 0.0000 0.0000 0.4745 0.5047 0.1170 0.1170 0.0000 0.0000 0.3123 0.3123 0.3270 0.3445 0.0263 0.0263 0.6133 0.6133 0.4017 0.4017 0.0000 0.0000 0.1505 0.1505 0.0000 0.0000 0.3014 0.4138 0.2531 0.2531 0.0000 0.0000 0.0000 0.0000 0.3636 0.3636 0.2767 0.2767 0.0045 0.0429 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.1933 0.1933 0.0000 0.0000 0.1707 0.0000 0.0000 0.0000 2002 Case 1 Case 3 0.1729 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5707 0.5713 0.0000 0.0000 0.0000 0.0000 0.6814 0.6958 0.2693 0.2693 0.0000 0.0000 0.2028 0.2028 0.0831 0.1696 0.0723 0.0723 0.7380 0.7380 0.6783 0.6783 0.0362 0.1011 0.3577 0.3577 0.0000 0.0000 0.4408 0.4880 0.4155 0.4155 0.5056 0.5056 0.0000 0.0000 0.5007 0.5007 0.3642 0.3642 0.2138 0.2595 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.2980 0.2980 0.0383 0.0383 0.4767 0.0000 0.0000 0.0000 2003 Case 1 Case 3 0.1459 0.4162 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5366 0.5407 0.0000 0.0000 0.0000 0.0000 0.5329 0.6372 0.2139 0.2139 0.0000 0.0000 0.1445 0.1445 0.0428 0.0000 0.0370 0.0370 0.5632 0.5632 0.5520 0.5520 0.0000 0.0000 0.3649 0.3654 0.0000 0.0000 0.3063 0.4503 0.4017 0.4017 0.6074 0.6074 0.0000 0.0000 0.4274 0.4274 0.2737 0.2737 0.4974 0.4974 0.0000 0.0000 0.0000 0.0000 0.0306 0.0000 0.3422 0.3422 0.1094 0.1094 0.5131 0.0000 0.0000 0.0000
- 151 -
Assessment of Productive Efficiency of Airports
Table 7.14 Efficiency scores for Cases 1 and 3 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Average score Number of efficient airports 2000 Case 1 Case 3 0.0000 0.0000 0.1833 0.1833 0.1553 0.2757 0.0000 0.0000 0.1934 0.0000 0.2600 0.2600 0.2944 0.2944 0.6936 0.6936 0.0000 0.0000 0.7166 0.7526 0.0000 0.0000 0.2264 0.2264 0.1525 0.1525 0.4099 0.4372 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5650 0.5650 0.0000 0.0000 0.6842 0.6842 0.0000 0.0000 0.0000 0.0000 0.0147 0.0147 0.4392 0.4392 0.2208 0.2514 23 22 2001 Case 1 Case 3 0.0000 0.0000 0.0393 0.0438 0.1037 0.0000 0.0000 0.0000 0.2388 0.3104 0.1223 0.1223 0.0000 0.0000 0.2441 0.2441 0.0000 0.0000 0.5415 0.5620 0.0000 0.0000 0.1214 0.1214 0.0000 0.0000 0.7247 0.7255 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.4686 0.4686 0.0729 0.0000 0.2175 0.2175 0.0000 0.0000 0.2189 0.2189 0.0512 0.0512 0.0294 0.0294 0.1326 0.1453 29 24 2002 Case 1 Case 3 0.0000 0.0000 0.2002 0.2002 0.0468 0.0000 0.0000 0.0000 0.1477 0.0000 0.0000 0.0000 0.0331 0.0408 0.6013 0.6013 0.0000 0.0000 0.6188 0.0000 0.0000 0.0000 0.3194 0.3194 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.7539 0.7539 0.3169 0.4290 0.1123 0.1123 0.0000 0.0000 0.3945 0.3945 0.0000 0.0000 0.1627 0.1627 0.1672 0.2001 29 24 2003 Case 1 Case 3 0.0000 0.0000 0.1709 0.1709 0.1366 0.2220 0.0000 0.0000 0.1085 0.2135 0.0000 0.0000 0.1202 0.1316 0.6686 0.6686 0.0000 0.0000 0.6064 0.6064 0.0000 0.0000 0.3257 0.3257 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.7767 0.7767 0.0000 0.0000 0.1898 0.1898 0.0000 0.0000 0.5103 0.5103 0.0000 0.0000 0.3630 0.3630 0.1792 0.2026 28 25
Note: An efficient airport has a zero score as highlighted by bold typeface. The output sets of Cases 1 and 3 are the same. The input
set of Case 3 is {number of runways, runway area}. The input set of Case 1 is {land area, number of runways, runway area}
- 152 -
Assessment of Productive Efficiency of Airports
Both parametric and nonparametric statistical tests are performed to determine whether the score differences between Cases 1 and 3 are significant. Table 7.15 shows results from paired-sample t-test whereas results from nonparametric Wilcoxon signed rank test and sign test are shown in Table 7.16. In brief, paired-sample t-tests infer that efficiency scores in Cases 1 and 3 are still significantly different. However, the level of significance is not as strong as in the comparison between Cases 1 and 2. The t-statistics are as low as -2.1708 (Pair 4: year 2003). Results from nonparametric tests (Table 7.16) also provide the same statistical inference. Both Wilcoxon signed-rank test and sign test indicate the difference in scores in Cases 1 and 3 are statistically significant at 99% level.
Table 7.15
Comparisons of efficiency scores between Cases 1 and 3 by paired sample t-test Paired differences Cases 1 and 3 95% confidence interval of the Std. Std. error difference deviation mean Lower Upper 0.0981 0.0131 -0.0568 -0.0042 0.0330 0.0044 -0.0215 -0.0038 0.1072 0.0143 -0.0616 -0.0041 0.0807 0.0107 -0.0450 -0.0018 0.0845 0.0056 -0.0360 -0.0137
Paired-sample t-test Pair 1: year 2000 Pair 2: year 2001 Pair 3: year 2002 Pair 4: year 2003 Pair 5: 2000 – 2003 Mean -0.0305 -0.0127 -0.0328 -0.0234 -0.0248
t -2.3273 -2.8848 -2.2956 -2.1708 -4.4057
- 153 -
Assessment of Productive Efficiency of Airports
Table 7.16
Comparisons of efficiency scores between Cases 1 and 3 by nonparametric paired tests Asymptotic significance (2-tailed) 0.0022 0.0014 0.0014 0.0033 0.0000 .0004 b .0002 b .0002 b .0009 b .0000 b
Nonparametric paired test A. Wilcoxon Signed-Rank test Pair 1: year 2000 Pair 2: year 2001 Pair 3: year 2002 Pair 4: year 2003 Pair 5: 2000 – 2003 Pair 1: year 2000 Pair 2: year 2001 Pair 3: year 2002 Pair 4: year 2003 Pair 5: 2000 – 2003
Z -3.0594a -3.1798a -3.1798a -2.9340a -6.0927a -6.8571
B. Sign test
a b
Based on negative ranks. Binomial distribution used. Fourth, on the maximum possible production, Tables 7.17, 7.18, 7.19 and 7.20
compare estimated potential outputs from Cases 1 and 3 in 2000, 2001, 2002, and 2003 respectively. Since the two cases tend to identify similar set of efficient airports, it is not surprising to see that the figures are very similar between cases. The difference between two cases only occurs at some airports. For example, in 2003, 11 airports (i.e., ABQ, BOS, DCA, FLL, HOU, LGB, MDW, MSY, ONT, and ORF - those with different efficiency scores) show different maximum possible production. Consequently, the figures for overall system are slightly different. Without land area (Case 3), the system may have potential to accommodate 25, 22, and 42% of passengers, movements and cargo throughput as compared to 23, 20 and 35% when having land area as another input (Case 1). Seemingly, ignoring land area does not drastically change the results. The assessment of 56 U.S. airports tends to be robust. The addition of undesirable outputs
- 154 -
Assessment of Productive Efficiency of Airports
(i.e., delayed flights and time delays) into the model made a much greater difference than the addition of another input (i.e., land area). In the latter case, the model only identifies very few more efficient airports. This finding suggests that consideration of undesirable outputs is at least as important as the consideration of additional inputs in determining the relative productivity of airports. Fifth, on the productivity growth indexes, Table 7.21 compares computed Luenberger productivity indexes between Cases 1 and 3. Since inputs to the computation (i.e., efficiency scores) are rather similar, the resulting indexes are therefore only slightly different. Between 2000 and 2001, the overall system growth is 0.9% in Case 3 as compared to 1.8% in Case 1. The gaps are narrower in the the two subsequent periods where Case 1 shows growth of 1.2% and -1.3% as compared to 1.3% and -1.0% in Case 3. The significant difference of efficiency scores between two cases are largely hidden in the computation of Luenberger indexes. A caution should be raised here if one were to perform statistical tests on the difference of Luenberger indexes.
- 155 -
Assessment of Productive Efficiency of Airports
Table 7.17 Maximum possible passengers, aircraft movements and cargo throughput in 2000, Cases 1 and 3
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM Case 1 7,572,239 5,030,557 80,162,407 3,067,777 41,712,092 4,800,070 19,602,609 24,396,829 31,173,199 6,873,998 41,581,182 16,050,706 50,206,299 106,262,017 70,620,630 34,188,468 25,282,740 23,016,542 12,170,935 42,098,078 53,509,212 1,360,280 8,106,485 43,038,655 58,688,340 67,303,182 25,374,868 637,853 15,841,801 38,897,365 15,857,218 11,769,213 Total passengers % add Case 3 20 17,571,012 0 5,030,557 0 80,162,407 0 3,067,777 50 41,722,480 1 5,508,388 0 19,602,609 84 21,725,198 35 31,173,199 0 6,873,998 84 41,581,182 1 16,793,926 30 50,206,299 75 106,262,017 99 70,620,630 0 34,188,468 59 25,282,740 0 23,016,542 34 14,351,543 109 42,098,078 52 53,509,212 0 1,360,280 53 8,106,485 31 43,038,655 59 59,913,693 0 67,303,182 0 25,374,868 0 637,853 28 15,841,801 26 38,897,365 1 25,185,292 0 11,769,213 % add 179 0 0 0 50 16 0 63 35 0 84 6 30 75 99 0 59 0 58 109 52 0 53 31 63 0 0 0 28 26 61 0 Case 1 280,925 288,919 915,454 153,917 726,532 162,504 316,703 474,220 607,944 238,011 665,523 300,889 739,636 1,438,268 931,597 450,229 459,262 345,771 339,653 657,193 714,538 90,883 227,940 648,506 815,812 783,433 383,325 379,399 279,970 540,853 301,469 388,412 Aircraft movements % add Case 3 20 368,092 0 288,919 0 915,454 0 153,917 49 726,709 1 184,494 0 316,703 43 488,684 34 607,944 0 238,011 39 665,523 1 314,654 45 739,636 72 1,438,268 68 931,597 0 450,229 57 459,262 0 345,771 33 365,560 44 657,193 48 714,538 0 90,883 53 227,940 88 648,506 56 832,349 0 783,433 0 383,325 0 379,399 28 279,970 51 540,853 1 471,021 0 388,412 % add 58 0 0 0 49 15 0 47 34 0 39 6 45 72 68 0 57 0 44 44 48 0 53 88 60 0 0 0 28 51 58 0 Case 1 130,936 1,804,221 868,286 40,722 714,502 62,433 236,043 211,276 287,627 22,572 644,908 177,857 610,884 1,579,199 1,325,558 1,082,407 515,706 441,163 150,134 567,217 1,146,512 16,815 93,435 2,382,512 1,060,137 2,038,784 71,149 49,415 244,619 545,822 99,464 2,489,078 Cargo (tons) % add Case 3 52 288,345 0 1,804,221 0 868,286 0 40,722 50 714,680 67 49,499 0 236,043 77 407,440 46 287,627 0 22,572 65 644,908 371 295,216 30 610,884 75 1,579,199 345 1,325,558 0 1,082,407 118 515,706 0 441,163 1,836 145,628 48 567,217 211 1,146,512 0 16,815 53 93,435 31 2,382,512 963 1,195,525 0 2,038,784 0 71,149 0 49,415 62 244,619 101 545,822 371 428,643 0 2,489,078 % add 234 0 0 0 50 32 0 241 46 0 65 682 30 75 345 0 118 0 1,778 48 211 0 53 31 1,099 0 0 0 62 101 1,932 0
- 156 -
Assessment of Productive Efficiency of Airports
Table 7.17 Maximum possible passengers, aircraft movements and cargo throughput in 2000, Cases 1 and 3 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Total Case 1 33,621,273 43,503,104 11,407,888 10,963,802 6,757,398 90,903,872 3,946,234 9,894,942 13,790,115 45,168,996 36,044,635 26,063,445 3,769,595 3,991,725 14,868,547 776,904 28,408,553 64,240,596 13,097,259 34,339,104 7,772,801 30,561,387 8,144,375 23,090,278 1,551,380,674 Total passengers % add Case 3 0 33,621,273 18 43,503,104 16 12,596,173 0 10,963,802 0 8,064,476 26 90,903,872 29 3,946,234 69 9,894,942 0 13,790,115 81 47,507,035 0 36,044,635 32 26,063,445 254 3,769,595 212 5,016,855 0 14,868,547 0 776,904 0 28,408,553 56 64,240,596 0 13,097,259 73 34,339,104 0 7,772,801 0 30,561,387 1,428 8,144,375 44 23,090,278 30 1,578,762,310 % add 0 18 28 0 19 26 29 69 0 91 0 32 254 292 0 0 0 56 0 73 0 0 1,428 44 33 Case 1 517,440 616,656 184,325 449,050 155,501 1,156,282 161,378 324,876 314,378 792,038 579,816 550,159 135,749 117,090 206,289 167,376 445,677 778,647 287,072 614,373 387,862 481,025 138,470 399,174 26,008,394 Aircraft movements % add Case 3 0 517,440 18 616,656 16 203,499 0 449,050 0 185,542 27 1,156,282 29 161,378 68 324,876 0 314,378 64 807,482 0 579,816 23 550,159 15 135,749 41 119,352 0 206,289 0 167,376 0 445,677 81 778,647 0 287,072 67 614,373 0 387,862 0 481,025 1 138,470 44 399,174 26 26,424,875 % add 0 18 28 0 19 27 29 68 0 67 0 23 15 44 0 0 0 81 0 67 0 0 1 44 28 Case 1 1,642,744 1,045,794 164,659 685,425 464,164 2,437,829 57,061 434,258 282,019 960,161 340,352 180,292 77,276 80,433 139,107 2,970 456,920 1,707,269 147,929 432,166 15,589 130,152 156,004 421,808 34,173,774 Cargo (tons) % add Case 3 0 1,642,744 183 1,045,794 91 204,213 0 685,425 0 553,947 66 2,437,829 97 57,061 1,971 434,258 0 282,019 72 980,276 0 340,352 23 180,292 1,224 77,276 61,299 92,445 0 139,107 0 2,970 0 456,920 96 1,707,269 0 147,929 68 432,166 0 15,589 0 130,152 381 156,004 309 421,808 43 35,253,474 % add 0 183 137 0 19 66 97 1,971 0 75 0 23 1,224 70,468 0 0 0 96 0 68 0 0 381 309 48
Note: Case 3 differs from case 1 in that it drops land area from the set of inputs. Aircraft movements include both delayed and non-
delayed flights. % add is the percentage increase from current level of the corresponding output.
- 157 -
Assessment of Productive Efficiency of Airports
Table 7.18 Maximum possible passengers, aircraft movements and cargo throughput in 2001, Cases 1 and 3
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM Case 1 6,183,606 5,107,311 75,858,500 3,012,729 38,481,290 4,487,335 20,369,923 17,510,135 25,890,177 6,680,897 22,664,832 17,476,832 37,043,730 88,971,994 45,266,906 30,558,000 18,877,562 20,151,936 11,240,147 29,224,387 34,803,580 1,284,311 6,926,081 37,470,750 35,338,310 61,606,204 21,933,000 587,473 14,358,821 28,253,061 15,681,966 11,808,247 Total passengers % add Case 3 0 6,302,898 0 5,107,311 0 75,858,500 0 3,012,729 59 38,481,290 0 4,711,491 0 20,369,923 47 17,868,321 12 25,890,177 0 6,680,897 31 22,664,832 33 17,707,265 3 37,043,730 61 88,971,994 40 45,266,906 0 30,558,000 15 18,877,562 0 20,151,936 30 12,211,045 64 29,224,387 0 34,803,580 0 1,284,311 36 6,926,081 28 37,470,750 0 36,688,573 0 61,606,204 0 21,933,000 0 587,473 19 14,358,821 0 28,253,061 0 18,358,628 0 11,808,247 % add 2 0 0 0 59 5 0 50 12 0 31 34 3 61 40 0 15 0 41 64 0 0 36 28 4 0 0 0 19 0 17 0 Case 1 242,733 284,441 890,494 148,869 703,723 159,705 324,065 428,372 514,675 243,201 506,019 321,934 540,903 1,243,721 725,372 436,420 333,261 327,006 323,801 495,596 470,916 92,402 183,398 598,329 530,328 738,114 365,716 358,508 250,145 315,752 278,734 394,826 Aircraft movements % add Case 3 0 247,415 0 284,441 0 890,494 0 148,869 55 703,723 0 167,673 0 324,065 47 437,059 12 514,675 0 243,201 31 506,019 32 326,103 12 540,903 59 1,243,721 39 725,372 0 436,420 15 333,261 0 327,006 30 351,588 25 495,596 0 470,916 0 92,402 36 183,398 105 598,329 7 564,386 0 738,114 0 365,716 0 358,508 19 250,145 0 315,752 0 325,549 0 394,826 % add 2 0 0 0 55 5 0 50 12 0 31 34 12 59 39 0 15 0 41 25 0 0 36 105 14 0 0 0 19 0 17 0 Case 1 72,876 1,873,750 739,927 35,433 628,306 32,878 225,083 150,411 292,366 15,260 493,936 132,195 420,832 1,378,600 1,442,812 795,584 209,287 337,631 74,212 414,667 337,842 14,634 82,896 1,826,652 517,649 1,774,402 52,148 53,190 236,252 223,545 15,684 2,631,631 Cargo (tons) % add Case 3 0 87,662 0 1,873,750 0 739,927 0 35,433 59 628,306 0 47,052 0 225,083 47 153,488 65 292,366 0 15,260 53 493,936 425 155,913 17 420,832 76 1,378,600 499 1,442,812 0 795,584 15 209,287 0 337,631 1,127 63,929 25 414,667 0 337,842 0 14,634 36 82,896 28 1,826,652 547 699,128 0 1,774,402 0 52,148 0 53,190 66 236,252 0 223,545 0 187,012 0 2,631,631 % add 20 0 0 0 59 43 0 50 65 0 53 519 17 76 499 0 15 0 957 25 0 0 36 28 773 0 0 0 66 0 1,092 0
- 158 -
Assessment of Productive Efficiency of Airports
Table 7.18 Maximum possible passengers, aircraft movements and cargo throughput in 2001, Cases 1 and 3 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Total Case 1 31,668,450 35,058,695 9,567,651 11,713,225 8,302,839 75,694,280 2,963,223 7,388,932 12,703,676 37,847,971 35,439,051 22,365,971 1,057,150 2,025,856 15,184,332 725,140 27,036,073 50,861,183 13,088,997 22,911,625 7,324,557 36,378,058 1,918,917 16,355,131 1,280,691,015 Total passengers % add Case 3 0 31,668,450 4 35,211,933 0 10,559,396 0 11,713,225 24 8,782,872 12 75,694,280 0 2,963,223 24 7,388,932 0 12,703,676 54 38,352,450 0 35,439,051 12 22,365,971 0 1,057,150 72 2,026,702 0 15,184,332 0 725,140 0 27,036,073 47 50,861,183 0 14,043,080 22 22,911,625 0 7,324,557 36 36,378,058 376 1,918,917 3 16,355,131 15 1,289,705,328 % add 0 4 10 0 31 12 0 24 0 56 0 12 0 73 0 0 0 47 7 22 0 36 376 3 16 Case 1 471,008 520,651 145,564 395,653 191,525 1,063,376 119,309 234,992 291,117 699,371 553,310 506,271 116,501 144,738 206,988 160,486 400,635 730,017 255,499 452,824 378,903 574,174 119,598 268,481 22,772,468 Aircraft movements % add Case 3 0 471,008 4 522,863 0 213,523 0 395,653 24 202,566 17 1,063,376 0 119,309 24 234,992 0 291,117 50 708,189 0 553,310 12 506,271 0 116,501 72 144,798 0 206,988 0 160,486 0 400,635 88 730,017 0 273,871 22 452,824 0 378,903 21 574,174 5 119,598 3 268,481 16 23,015,097 % add 0 4 47 0 31 17 0 24 0 52 0 12 0 72 0 0 0 88 7 22 0 21 5 3 17 Case 1 1,639,760 353,018 75,700 593,634 519,100 1,567,465 28,786 31,619 242,967 826,639 283,337 155,931 4,976 21,966 134,689 2,946 400,499 934,037 143,914 313,835 14,849 395,215 20,924 217,161 26,455,537 Cargo (tons) % add Case 3 0 1,639,760 4 354,561 0 184,203 0 593,634 24 549,112 21 1,567,465 0 28,786 53 31,619 0 242,967 54 837,658 0 283,337 12 155,931 0 4,976 23,776 19,408 0 134,689 0 2,946 0 400,499 47 934,037 0 514,508 45 313,835 0 14,849 223 395,215 5 20,924 172 217,161 24 27,372,929 % add 0 4 143 0 31 21 0 53 0 56 0 12 0 20,996 0 0 0 47 258 45 0 223 5 172 29
Note: Case 3 differs from case 1 in that it drops land area from the set of inputs. Aircraft movements include both delayed and non-
delayed flights. % add is the percentage increase from current level of the corresponding output.
- 159 -
Assessment of Productive Efficiency of Airports
Table 7.19 Maximum possible passengers, aircraft movements and cargo throughput in 2002, Cases 1 and 3
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM Case 1 6,151,129 4,914,539 76,876,128 2,810,791 35,647,876 4,620,683 19,012,529 17,578,887 29,952,090 6,741,354 37,620,386 13,941,695 38,229,992 91,816,824 54,507,100 30,258,481 23,131,236 19,749,905 11,577,573 31,569,526 51,048,711 1,221,138 7,432,943 40,849,751 42,492,287 56,223,843 21,986,679 1,453,412 13,342,966 27,674,633 17,371,036 11,141,594 Total passengers % add Case 3 0 16,538,636 0 4,914,539 0 76,876,128 0 2,810,791 57 35,662,137 0 4,620,683 0 19,012,529 68 17,729,702 27 29,952,090 0 6,741,354 81 37,620,386 8 15,055,396 7 38,229,992 74 91,816,824 68 54,507,100 4 32,154,783 36 23,131,236 0 19,749,905 44 11,957,427 85 31,569,526 51 51,048,711 0 1,221,138 50 7,432,943 36 40,849,751 21 44,094,440 0 56,223,843 0 21,986,679 0 1,453,412 30 13,342,966 4 27,674,633 0 25,651,538 0 11,141,594 % add 169 0 0 0 57 0 0 70 27 0 81 17 7 74 68 10 36 0 49 85 51 0 50 36 26 0 0 0 30 4 48 0 Case 1 254,874 277,267 889,966 146,555 610,900 162,211 304,921 420,585 576,384 255,630 582,466 252,786 558,511 1,302,154 815,124 485,889 379,747 323,726 354,136 524,229 743,794 82,883 187,219 584,615 602,418 645,424 362,439 350,603 248,280 403,408 304,304 398,769 Aircraft movements % add Case 3 0 298,929 0 277,267 0 889,966 0 146,555 56 611,140 0 162,211 0 304,921 67 424,159 27 576,384 0 255,630 20 582,466 17 251,933 13 558,511 70 1,302,154 66 815,124 20 492,992 35 379,747 0 323,726 44 365,708 41 524,229 63 743,794 0 82,883 50 187,219 103 584,615 21 623,900 0 645,424 0 362,439 0 350,603 30 248,280 39 403,408 0 446,523 0 398,769 % add 17 0 0 0 56 0 0 68 27 0 20 17 13 70 66 22 35 0 49 41 63 0 50 103 26 0 0 0 30 39 47 0 Case 1 74,460 1,771,595 734,083 32,353 609,352 39,751 251,354 170,482 280,897 10,700 420,995 131,868 506,884 1,165,008 764,041 880,784 224,076 414,947 99,629 459,860 514,101 13,863 103,415 2,168,672 983,493 1,779,855 32,223 53,356 257,344 358,923 26,309 3,390,800 Cargo (tons) % add Case 3 0 340,854 0 1,771,595 0 734,083 0 32,353 57 609,596 0 39,751 0 251,354 68 171,945 75 280,897 0 10,700 20 420,995 2,150 157,961 52 506,884 74 1,165,008 228 764,041 4 935,983 36 224,076 0 414,947 1,756 59,656 42 459,860 56 514,101 0 13,863 50 103,415 36 2,168,672 1,100 621,355 0 1,779,855 0 32,223 0 53,356 89 257,344 81 358,923 0 249,193 0 3,390,800 % add 358 0 0 0 57 0 0 70 75 0 20 2,595 52 74 228 10 36 0 1,011 42 56 0 50 36 658 0 0 0 89 81 847 0
- 160 -
Assessment of Productive Efficiency of Airports
Table 7.19 Maximum possible passengers, aircraft movements and cargo throughput in 2002, Cases 1 and 3 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Total Case 1 30,060,241 39,838,014 9,251,773 13,005,642 6,517,050 66,565,952 3,579,015 8,780,743 12,241,975 24,799,470 35,547,167 23,785,515 1,345,970 1,108,695 14,931,854 728,307 26,690,843 55,169,980 14,638,695 20,741,521 7,903,066 38,825,764 362,017 18,015,387 1,323,382,375 Total passengers % add Case 3 0 30,060,241 22 39,838,014 0 9,684,859 0 13,005,642 0 7,579,751 0 66,565,952 3 3,605,579 60 8,780,743 0 12,241,975 0 40,145,233 0 35,547,167 32 23,785,515 0 1,345,970 0 1,108,695 0 14,931,854 0 728,307 0 26,690,843 75 55,169,980 32 15,884,555 11 20,741,521 0 7,903,066 52 38,825,764 0 362,017 16 18,015,387 22 1,365,321,443 % add 0 22 5 0 16 0 4 60 0 62 0 32 0 0 0 0 0 75 43 11 0 52 0 16 26 Case 1 446,235 605,830 139,291 371,988 149,292 922,817 129,782 265,888 276,877 463,167 545,771 559,932 130,826 85,243 206,380 159,835 364,735 756,932 296,477 452,571 368,627 604,053 123,642 283,457 23,101,862 Aircraft movements % add Case 3 0 446,235 19 605,830 0 156,839 0 371,988 0 171,318 0 922,817 3 130,745 59 265,888 0 276,877 0 729,445 0 545,771 32 559,932 0 130,826 0 85,243 0 206,380 0 159,835 0 364,735 115 756,932 43 295,878 11 452,571 0 368,627 38 604,053 0 123,642 16 283,457 21 23,637,472 % add 0 19 13 0 15 0 4 59 0 57 0 32 0 0 0 0 0 115 43 11 0 38 0 16 24 Case 1 1,624,242 383,425 84,150 634,643 496,547 1,473,980 52,005 54,643 245,134 541,039 298,945 184,943 4,518 74 151,644 2,832 374,753 1,034,300 184,570 297,407 13,730 352,447 13,257 268,010 27,496,714 Cargo (tons) % add Case 3 0 1,624,242 20 383,425 0 196,986 0 634,643 0 569,902 0 1,473,980 58 50,334 205 54,643 0 245,134 0 875,831 0 298,945 32 184,943 0 4,518 0 74 0 151,644 0 2,832 0 374,753 75 1,034,300 32 234,260 38 297,407 0 13,730 173 352,447 0 13,257 192 268,010 24 28,235,881 % add 0 20 134 0 15 0 53 205 0 62 0 32 0 0 0 0 0 75 67 38 0 173 0 192 27
Note: Case 3 differs from case 1 in that it drops land area from the set of inputs. Aircraft movements include both delayed and non-
delayed flights. % add is the percentage increase from current level of the corresponding output.
- 161 -
Assessment of Productive Efficiency of Airports
Table 7.20 Maximum possible passengers, aircraft movements and cargo throughput in 2003, Cases 1 and 3
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM Case 1 6,935,011 4,791,431 79,086,792 2,672,637 35,020,347 4,729,936 20,094,756 16,180,712 31,635,406 6,252,061 38,539,652 14,214,803 38,894,115 83,244,552 50,696,543 29,431,061 24,484,401 19,732,556 10,193,688 28,154,745 54,898,411 1,215,093 6,970,535 40,417,649 54,332,887 54,982,838 22,482,770 2,875,703 13,040,110 30,308,286 18,426,397 11,437,307 Total passengers % add Case 3 15 15,625,565 0 4,791,431 0 79,086,792 0 2,672,637 54 35,114,859 0 4,729,936 0 20,094,756 53 17,281,283 37 31,635,406 0 6,252,061 82 38,539,652 0 14,823,901 4 38,894,115 56 83,244,552 55 50,696,543 0 29,431,061 36 24,493,287 0 19,732,556 31 11,316,924 68 28,154,745 61 54,898,411 0 1,215,093 43 6,970,535 27 40,417,649 50 54,332,887 0 54,982,838 0 22,482,770 0 5,103,201 34 13,040,110 11 30,308,286 0 27,880,730 0 11,437,307 % add 158 0 0 0 54 0 0 64 37 0 82 4 4 56 55 0 37 0 45 68 61 0 43 27 50 0 0 77 34 11 51 0 Case 1 253,241 277,361 911,723 154,849 569,476 178,079 299,469 394,639 536,801 237,979 576,589 250,802 541,268 1,185,676 756,815 405,808 389,706 319,989 316,623 465,444 743,930 79,377 172,836 570,852 743,556 622,378 374,952 338,807 229,168 380,232 328,035 402,258 Aircraft movements % add Case 3 15 312,946 0 277,361 0 911,723 0 154,849 53 570,992 0 178,079 0 299,469 53 421,281 21 536,801 0 237,979 14 576,589 0 261,399 8 541,268 55 1,185,676 54 756,815 0 405,808 36 389,845 0 319,989 30 351,390 39 465,444 57 743,930 0 79,377 43 172,836 104 570,852 48 743,556 0 622,378 0 374,952 0 349,176 34 229,168 29 380,232 0 491,221 0 402,258 % add 42 0 0 0 53 0 0 63 21 0 14 4 8 55 54 0 36 0 45 39 57 0 43 104 48 0 0 3 34 29 50 0 Case 1 133,954 2,102,025 798,501 34,184 943,554 44,654 235,576 272,944 334,987 10,766 449,437 5,774 884,820 1,043,533 710,246 874,641 226,674 421,930 155,183 399,990 683,510 10,957 104,553 2,071,962 510,720 1,833,300 28,402 50,873 265,216 860,080 23,266 3,390,515 Cargo (tons) % add Case 3 87 305,789 0 2,102,025 0 798,501 0 34,184 160 1,001,459 0 44,654 0 235,576 185 458,278 139 334,987 0 10,766 14 449,437 0 403,062 172 884,820 56 1,043,533 222 710,246 0 874,641 45 227,763 0 421,930 2,587 333,060 40 399,990 79 683,510 0 10,957 48 104,553 27 2,071,962 522 510,720 0 1,833,300 0 28,402 0 52,431 94 265,216 346 860,080 0 744,163 0 3,390,515 % add 327 0 0 0 176 0 0 379 139 0 14 6,881 172 56 222 0 46 0 5,667 40 79 0 48 27 522 0 0 3 94 346 3,098 0
- 162 -
Assessment of Productive Efficiency of Airports
Table 7.20 Maximum possible passengers, aircraft movements and cargo throughput in 2003, Cases 1 and 3 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Total Case 1 29,595,618 41,741,976 10,542,505 13,548,363 7,258,006 69,508,672 3,849,419 10,029,678 12,395,938 39,631,312 37,412,165 18,913,688 1,361,758 1,246,842 15,260,791 752,762 26,755,888 52,079,506 10,677,903 21,972,152 8,535,130 35,870,254 393,530 21,158,079 1,346,865,126 Total passengers % add Case 3 0 29,595,618 26 41,741,976 14 11,335,042 0 13,548,363 11 7,945,867 0 69,508,672 12 3,888,687 67 10,029,678 0 12,395,938 61 39,631,312 0 37,412,165 33 18,913,688 0 1,361,758 0 1,246,842 0 15,260,791 0 752,762 0 26,755,888 78 52,079,506 0 10,677,903 19 21,972,152 0 8,535,130 76 35,870,254 0 393,530 36 21,158,079 23 1,371,693,477 % add 0 26 22 0 21 0 13 67 0 61 0 33 0 0 0 0 0 78 0 19 0 76 0 36 25 Case 1 417,423 595,117 156,040 342,871 162,248 928,691 135,951 284,004 267,052 700,772 541,771 478,550 127,197 93,068 203,285 152,485 354,770 600,665 198,082 476,174 350,074 568,680 112,284 317,570 22,573,542 Aircraft movements % add Case 3 0 417,423 17 595,117 14 167,701 0 342,871 11 177,587 0 928,691 12 137,337 65 284,004 0 267,052 57 700,772 0 541,771 32 478,550 0 127,197 0 93,068 0 203,285 0 152,485 0 354,770 80 600,665 0 198,082 19 476,174 0 350,074 50 568,680 0 112,284 36 317,570 20 22,908,850 % add 0 17 22 0 21 0 13 65 0 57 0 32 0 0 0 0 0 80 0 19 0 50 0 36 22 Case 1 1,637,278 370,000 253,278 597,383 574,965 1,510,746 36,821 175,200 239,265 842,526 288,350 294,139 4,569 103 135,547 2,825 351,418 1,682,768 108,622 402,086 12,050 299,115 19,024 649,238 30,404,043 Cargo (tons) % add Case 3 0 1,637,278 17 370,000 213 276,540 0 597,383 11 629,456 0 1,510,746 14 36,532 857 175,200 0 239,265 61 842,526 0 288,350 142 294,139 0 4,569 0 103 0 135,547 0 2,825 0 351,418 193 1,682,768 0 108,622 85 402,086 0 12,050 159 299,115 0 19,024 595 649,238 35 32,195,292 % add 0 17 242 0 21 0 13 857 0 61 0 142 0 0 0 0 0 193 0 85 0 159 0 595 42
Note: Case 3 differs from case 1 in that it drops land area from the set of inputs. Aircraft movements include both delayed and non-
delayed flights. % add is the percentage increase from current level of the corresponding output.
- 163 -
Assessment of Productive Efficiency of Airports
Table 7.21 Luenberger productivity indexes, Cases 1 and 3
Airport code ABQ ANC ATL BHM BOS BUR BWI CLE CLT CMH CVG DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH JAN JAX JFK LAS LAX LGA LGB MCI MCO MDW MEM
2000 - 2001 Case 1 Case 3 0.136 0.079 0.000 0.000 0.310 0.310 0.006 -0.002 -0.078 -0.079 -0.002 -0.001 0.015 0.015 -0.095 -0.127 0.040 0.040 0.000 0.000 -0.170 -0.170 -0.216 -0.220 -0.019 -0.019 0.023 0.023 0.060 0.060 0.011 0.011 0.074 0.102 0.000 0.000 0.002 -0.073 -0.048 -0.048 0.104 0.104 0.077 0.077 0.042 0.042 -0.116 -0.116 0.189 0.235 0.000 0.000 0.202 0.475 -0.016 -0.014 -0.025 -0.025 -0.015 -0.015 -0.103 -0.002 -0.062 0.000
2001 - 2002 Case 1 Case 3 0.093 0.009 0.000 0.000 0.000 0.000 0.000 0.000 -0.027 -0.023 0.079 0.000 0.096 0.096 -0.090 -0.068 -0.003 -0.003 0.012 0.012 0.219 0.219 0.041 0.091 0.033 0.033 -0.001 -0.001 -0.086 -0.086 0.167 0.124 -0.045 -0.045 0.000 0.000 -0.007 0.001 -0.064 -0.064 -0.102 -0.102 0.026 0.026 -0.029 -0.029 0.089 0.089 -0.002 0.017 0.000 0.000 0.030 0.001 0.000 0.000 -0.062 -0.062 -0.009 -0.009 -0.002 0.000 0.000 0.000
2002 - 2003 Case 1 Case 3 -0.079 -0.090 0.000 0.000 0.000 0.000 0.028 0.029 -0.020 -0.029 0.000 0.000 0.000 0.000 0.002 0.035 -0.021 -0.021 -0.012 -0.012 0.042 0.042 0.093 0.060 -0.008 -0.008 0.055 0.055 -0.052 -0.052 0.023 0.050 0.021 0.009 0.000 0.000 -0.014 0.007 -0.055 -0.055 -0.032 -0.032 0.000 0.000 -0.029 -0.029 0.043 0.043 -0.036 -0.111 0.000 0.000 0.009 0.173 -0.018 -0.001 -0.072 -0.072 -0.021 -0.021 0.045 0.000 0.000 0.000
- 164 -
Assessment of Productive Efficiency of Airports
Table 7.21 Luenberger productivity indexes, Cases 1 and 3 (Continued)
Airport code MIA MSP MSY OAK ONT ORD ORF PBI PDX PHL PHX PIT PNS PSP SAN SBA SEA SFO SJC SLC SNA STL SWF TPA Average index Number of regress Number of no change Number of progress
2000 - 2001 Case 1 Case 3 0.000 0.000 -0.003 -0.001 -0.049 -0.012 -0.013 -0.013 -0.032 -0.077 0.555 0.264 0.162 0.162 -0.078 -0.078 -0.015 -0.015 -0.022 0.204 0.075 0.265 -0.016 -0.016 0.023 0.023 -0.027 -0.056 0.000 0.000 -0.010 -0.010 -0.026 -0.026 -0.105 -0.105 -0.047 0.000 -0.061 -0.061 0.000 0.000 -0.109 -0.109 -0.070 -0.070 0.066 0.066 0.009 0.996 28 29 9 7 19 20
2001 - 2002 Case 1 Case 3 0.000 0.000 -0.029 -0.030 0.036 0.046 0.000 0.000 0.043 0.069 0.170 0.170 -0.020 -0.017 -0.138 -0.138 0.000 0.000 0.016 0.047 0.019 0.020 -0.066 -0.066 0.003 0.001 0.270 0.314 0.000 0.000 -0.002 -0.002 -0.003 0.000 0.013 0.014 -0.151 -0.158 0.096 0.096 0.000 0.000 -0.087 -0.087 0.191 0.191 -0.015 -0.015 0.013 0.681 19 24 15 11 22 21
2002 - 2003 Case 1 Case 3 -0.011 0.000 -0.002 -0.002 -0.081 -0.068 0.000 0.000 0.008 -0.054 0.134 0.134 -0.049 -0.047 -0.009 -0.009 -0.003 -0.003 -0.014 -0.116 0.008 0.008 -0.138 -0.138 0.003 0.002 0.000 0.000 0.000 0.000 -0.021 -0.018 0.000 0.000 -0.056 -0.056 0.106 0.019 -0.056 -0.056 -0.001 0.000 -0.171 -0.171 -0.005 -0.005 -0.093 -0.093 -0.010 -0.013 27 29 15 12 14 15
Note: The negative index indicates regressed productivity. Zero value means that there is
no change in the productivity between two years. The positive index indicates productivity growth.
- 165 -
Assessment of Productive Efficiency of Airports
Tables 7.22 and 7.23 show the test results from paired-sample t-test and the two nonparametric tests. The t-statistics in Table 7.22 (0.8926, -0.1367, -0.5812, and 0.5096) indicate that the differences in Luenberger indexes between two cases are not statistically significant. Results from nonparametric tests in Table 7.23 support the same inference. The differences in Luenberger indexes are not statistically significant, although the efficiency scores themselves are statistically different (see results in Tables 7.15 and 7.16). The computation of Luenberger indexes really can conceal the difference in efficiency level and in turn can provide a misleading interpretation.
Table 7.22
Comparisons of Luenberger productivity indexes by paired sample t-test Cases 1and 3 Paired differences Cases 1 and 3 95% confidence interval of the Std. Std. error difference deviation mean Lower Upper 0.0708 0.0094 -0.0105 0.0274 0.0205 0.0027 -0.0058 0.0051 0.0331 0.0044 -0.0114 0.0062 0.0466 0.0036 -0.0052 0.0089
Paired-sample t-test Pair 1: 2000/01 Pair 2: 2001/02 Pair 3: 2002/03 Pair 4: 2000 – 03 Mean 0.0084 -0.0003 -0.0025 0.0018
t 0.8926 -0.1367 -0.5812 0.5096
- 166 -
Assessment of Productive Efficiency of Airports
Table 7.23
Comparisons of Luenberger productivity indexes by nonparametric paired tests Asymptotic significance (2-tailed) 0.4140 0.3477 0.6264 0.5791 0.6636b 0.1892b 1.0000a 0.2077
Nonparametric paired test A. Wilcoxon Signed-Rank test Pair 1: years 2000/01 Pair 2: years 2001/02 Pair 3: years 2002/03 Pair 4: years 2000 - 03 Pair 1: years 2000/01 Pair 2: years 2001/02 Pair 3: years 2002/03 Pair 4: years 2000 – 03
Z -0.8168a -0.9388a -0.4867c -0.5546a -1.2598
B. Sign test
a b c
Based on positive rank Binomial distribution used Based on negative rank
Determination of airport productivity
7.5
A causal model is developed to explain the variations in the efficiency score. For planning and managing an airport, the model will be very useful for predicting future productivity based on given information. The information is treated as exploratory variables which may include number of passengers per runway, passengers per movement, average delay, percentage of international passengers, etc. The dependent variable is the efficiency score. The model presented here is for Case 1 (with consideration of delays) which is considered as a more complete assessment of productive efficiency. By the nature of the directional output distance function, the value of efficiency scores can only be in the range of zero to infinity. Because of this special type of limited dependent variable,
- 167 -
Assessment of Productive Efficiency of Airports
simple regression is not an appropriate model. The issue was discussed earlier in chapter 5. Censored Tobit regression is employed. In this case, efficiency score of airport y i is represented by Equation (7.2)
?? xi + ? i yi = ? ?0
if y i > 0 if y i ? 0
(7.2)
y i is an efficiency score that is observable for values greater than 0 and is censored for values less than or equal to 0. Efficiency scores of all efficient airports are censored at 0, regardless of values of independent variables xi . ? and ? i are the coefficients and the error term of the Tobit model respectively. Coefficients ? can be estimated using Maximum Likelihood (ML) method. ML estimation for the Tobit model involves dividing the observations into two sets. The first set contains uncensored observations. The second set contains censored observations. The log-likelihood function is given in Equation (5.3). Meanwhile the marginal effect with respect to an exploratory variable can be computed using Equation (5.4). To measure the goodness-of-fit, the R2 ANOVA given in Equation (5.5) is computed. LIMDEP version 8.0 (Greene, 2002) is used to estimate the model.
7.6 Factors affecting productive efficiency of U.S. airports
The airport operation is a complex process involving a large number of activities. There are many variables that can affect the operational efficiency. Five groups of variables are investigated. The proxy of each group entering the model is essentially based on data availability.
- 168 -
Assessment of Productive Efficiency of Airports
First, Airport characteristics are represented here by physical characteristics. These are basically input measures that are used in the directional output distance function, i.e., land area (LAND), number of runways (RW) and runway area (RWA). These inputs certainly play a major role in accommodating traffic. However, one should be aware that having more of these inputs does not necessarily mean more outputs. Second, Airport services are mainly represented by outputs of airport operations which consist of number of aircraft movements (AIR), passengers (PAX) and cargo throughput (CARGO). One would expect that more services contribute to higher efficiency. However, this is not necessarily true since efficiency takes into account both inputs and outputs. Accordingly, another group of variables is introduced, i.e., level of utilization. Third, Level of utilization may be a better determinant of operational efficiency since it takes into accounts both input and output measures. This case study considers many ratio variables, such as non-delayed flights/land area, non-delayed flights/ /runway, non-delayed flights/ runway acreage area, annual total passengers/land area, annual total passengers/number of runways, annual total passengers/runway acreage area, annual cargo throughput/runway acreage area etc. Intuitively, higher values of these ratios should result in more efficient operation. Fourth, Market characteristics include target market (e.g., passengers, aircraft operation, cargo, general aviation and military service), market share, market dominance, market focus (e.g., domestic, international, tourist, business passengers), whether the airport is an airport in a multiple airport system, whether the airport is a hub airport
- 169 -
Assessment of Productive Efficiency of Airports
according to FAA definition, and irregularity of time periods. An attempt was made to collect these variables as much as possible. After all, six variables are entered in the estimation. These include the percentage of international passengers, percentage of general aviation, whether the airport is an airport in MAS, whether the airport is any kind of FAA hub (i.e., large, medium, small, or non-hub), whether the airport dominates in its corresponding MAS, and irregularity of years. In addition, several interaction variables (e.g., whether the airport is an airport in an MAS and also dominates the market etc.) are also tested. The fifth group of variables is Service characteristics. Not only the model
considers quantity of airport services, but it also aims to investigate the effects of service quality on the productive efficiency. As pointed out from the results in case study 1, unless quality of services is taken into account, only busy and congested airports will be classified as efficient. This case study takes delays as a proxy to represent service characteristics. Different ratios are computed and entered as exploratory variables. These include percentage of delayed flights, delayed-flights per runway, average delay per passenger and average delay per movement. It is expected that the lower values of these ratios should indicate the higher productive efficiency. In other words, the sign of the coefficients should be positive. Note that ownership/management characteristic, which is one significant variable in case study 1, is not considered here because there is no difference across airports in the dataset. Every airport, except Stewart International (SWF) is publicly owned and
- 170 -
Assessment of Productive Efficiency of Airports
operated. The inclusion of a dummy variable to represent private ownership does not give the meaningful results. The model is estimated from the pooled 4-year data (i.e., the years 2000 – 2003). Airports with incomplete independent variables are taken out. Totally, there are 211 complete observations or 13 samples shorter than the full sample size of 224 (i.e., 56 airports x 4 years). Several models consisting of different combinations of exploratory variable were estimated. Table 7.24 shows final model estimation results. It has nine independent variables, including the constant. Other variables are dropped off for reasons such as high correlation among themselves, being insignificant or having illogical sign. Recall that the lower efficiency score is desirable because it indicates that an airport is more efficient. As a result, a negative sign of the three utilization ratio variables in the model, i.e., non-delayed flights per land area (-0.1451 x 10-2), non-delayed-flights per runway area (-0.5986 x 10-4) and cargo throughput per runway area (-0.7212 x 10-4) contribute to higher productive efficiency. They are statistically significant at above the 95% confidence level. The marginal effects in the last column indicate changes in efficiency scores with respect to the changes in the corresponding exploratory variables. For instance, an increment of one non-delayed flight per acre of runway would result in an airport becoming more efficient by -0.2917 x 10-4 units.
- 171 -
Assessment of Productive Efficiency of Airports
Table 7.24 Censored Tobit regression model estimation results
Variables Constant % International Passengers Non-delayed flights/Land area Non-delayed flights/Runway area Cargo/ Runway area Delay/Passenger Y2001 Y2002 Y2003 Number of observations = 211 Log Likelihood function = -65.8723
Proposed model 2 R ANOVA = 1.3106 Coefficient 0.6167** (6.950) 0.0104** (3.646) -0.1451 x 10-2** (-4.612) -0.5986 x 10-4** (-2.156) -0.7212 x 10-4** (-5.411) 0.1033 x 10-4** (4.896) -0.1724** (-2.730) -0.1200* (-1.893) -0.1171* (-1.857) Marginal 0.3006** (6.276) 0.0051** (3.796) -0.0707 x 10-2** (-4.672) -0.2917 x 10-4** (-2.178) -0.3515 x 10-4** (-6.250) 0.0503 x 10-4** (5.105) -0.0840** (-2.703) -0.0585* (-1.890) -0.0571* (-1.857)
Note: Dependent variable = Efficiency score Y2001 = 1 if compute performance score in year 2001, otherwise = 0 Y2002 = 1 if compute performance score in year 2002, otherwise = 0 Y2003 = 1 if compute performance score in year 2003, otherwise = 0 ** Estimated coefficient is significant at the 0.05 level (one-tailed) * Estimated coefficient is significant at the 0.10 level (one-tailed)
Percentage of international passengers (coefficient = 0.0104) is positively associated with the efficiency score. The higher proportion of international passengers leads to lower efficiency. This may be well explained by the longer service time of this target market in comparison to domestic passengers. In general an airport uses more
- 172 -
Assessment of Productive Efficiency of Airports
resources to service an international passenger than it does to service a domestic passenger. The marginal effect suggests that for every additional percent of international passenger, an airport would be less efficient by 0.0051 units. The average delay per passenger (coefficient = 0.1033 x 10-4) is also positively associated with the efficiency score as expected. It is understandable that the higher delay leads to lower efficiency but this rarely has been quantified in the past. In this model, the estimation of the marginal effect suggests that for every additional minute of average delay per passenger, an airport would become less efficient by 0.0503 x 10-4 units. In addition to the above mentioned variables, there may be some effects from anomalies in 2001, 2002 and 2003 since these dummy variables are also statistically significant. The negative signs for these years indicate that airports become efficient slightly more easily in comparison to year 2000. In summary, the case study has assessed the airport productivity of 56 U.S. airports where joint production of desirable and undesirable outputs is taken into consideration. It also compares results with the case that undesirable outputs are ignored as this is the case in previous studies (See Table 2.4 for the list and description). In the last part of the chapter a productivity prediction model was developed using the Censored Tobit Regression. It is found that the increment of factors such as non-delayed flights per land area, non-delayed flights per runway area, and cargo throughput per runway area contribute to the enhancement of productivity. Meanwhile, the higher proportion of international passengers and average delay per passenger tend to reduce the productivity of airport operations. The model captures anomaly effects in the years 2001, 2002 and
- 173 -
Assessment of Productive Efficiency of Airports
2003 by indicating that airports could become efficient slightly more easily than the year 2000, ceteris paribus. In the next chapter, the important findings and insights will be summarized. Potential future research extensions are also suggested.
- 174 -
Assessment of Productive Efficiency of Airports
CHAPTER 8 CONCLUSIONS AND FUTURE RESEARCH
8.1 Conclusions from assessing productivity of airports in MASs
Arguing that it may be more useful to analyze productivity of airports operating in a similar market structure, the case study focused on airports in multiple airports systems (MASs). The data set consisted of 72 airports from 25 MASs in North America, South America, Europe and Asia. Data Envelopment Analysis (DEA) technique was used to assess the relative efficiencies of these airports. It was assumed that land area, number of runways, and runway area were the proxies of operational inputs whereas number of annual aircraft movements and passengers were two main target outputs from the operations. The analysis period was 2000 – 2002.
8.1.1 Productivity of airports in MASs
The assessment indicates that there are two groups of efficient or highly productive airports, coined by the busy and the compact. The busy group consists of market leaders in large MASs such as O’Hare International (ORD), Los Angeles International (LAX) and Heathrow/London (LHR). Air traffic statistics (ACI, 2002 – 2004) confirm that they are among the busiest airports in the world. Airports in the compact group are alternative airports with relatively small land area and only have one or two runways. Clearly airports in both groups are classified as efficient airports because of their relatively higher runway utilization. In this respect, larger airport size does not
- 175 -
Assessment of Productive Efficiency of Airports
guarantee high efficiency. An implication from this result is that an airport must be very busy; otherwise it would not be regarded as an efficient airport. This may make good sense as long as such high utilization does not create undesirable congestion and delays.
8.1.2 Underutilized airports
It is very difficult for all airports in an MAS to be highly utilized because total air travel demand must somehow be distributed among airports (Caves and Gosling, 1999; de Neufville, 1995; de Neufville and Odoni, 2003; Pathomsiri and Haghani, 2005; Pathomsiri, Mahmassani and Haghani, 2004). The effort to manage them by either coordinating or regulating air traffic has not been successful in most cases (Caves and Gosling, 1999; Charles River, 2001; de Neufville, 1995; de Neufville and Odoni, 2003). Given that the capital investment in airport business is very lumpy, it is extremely difficult to keep all runways in an MAS busy (New York/New Jersey region may be an exception). Consequently, functional failure is followed. This seems to be the case in this case study. It is found that some airports such as Montreal-Mirabel (YMX), Glasgow Prestwick International (GLA), Schoenefeld/Berlin (SXF), and Tempelhof/Berlin (THF) are underutilized. In fact, Montreal-Mirabel (YMX) is a case study of an unsuccessful airport in textbooks (Caves and Gosling, 1999; de Neufville, 1995; de Neufville and Odoni, 2003). Schoenefeld/Berlin (SXF), and Tempelhof/Berlin (THF) and Tegel (TXL, another airport in the Berlin MAS) are planned to be consolidated in 2011. Construction is underway (Berlin Brandenburg International, 2005). In this sense, the proposed models in this case study are useful in pointing out over-investment.
- 176 -
Assessment of Productive Efficiency of Airports
8.1.3
Factors affecting productive efficiency of airports in MASs
The case study estimated a Censored Tobit regression model for explaining variations in efficiency scores of airport operations. Five groups of exploratory variables were investigated, i.e., airport characteristics, airport services, level of utilization, market characteristics, and ownership/management characteristics. It was found that factors such as utilization of land area and runway area, passengers per aircraft movement, market dominance and privately-operated management style contribute to the enhancement of productivity. Meanwhile higher proportion of international passengers tends to reduce the productivity. The model also captured anomaly effects in year 2002 (it was observed that an airport could become efficient slightly more easily with the same utilization rate, possibly due to a global drop in air traffic after the September 11 terrorist attacks). Given some planned measures, the model can be used to predict future productivity of an airport and should be very useful as a tool for planning airport business in a competitive market.
8.2 Conclusions from assessing productivity of U.S. airports
The traditional measurement of productive efficiency of airport operations typically focuses on marketable outputs such as throughput of passengers, aircraft movements and cargo. As confirmed by case study 1, the typical results indicate that efficient airports are very busy airports and frequently they are congested. Reduction in so-called “undesirable outputs” such as delays has never been given credit in the assessment although it is a major concern in airport management. Case study 2 aimed at re-assessing productive efficiency of airport operations by considering joint production of desirable and undesirable outputs.
- 177 -
Assessment of Productive Efficiency of Airports
To estimate relative productivity, airports are viewed as similar production units taking three representatives of capital inputs, i.e., land area, number of runways and runway area; then producing three main desirable outputs, i.e., aircraft movements, passengers and cargo throughput. By the nature of airport operation, there are two byproduct outputs, though undesirable, i.e., delayed flights and time delays. The efficient airports are the ones that both achieve relatively high levels of desirable outputs while keeping the undesirable outputs at relatively low levels. Mathematically speaking, the model ought to simultaneously maximize desirable outputs and minimize undesirable outputs. Data Envelopment Analysis (DEA) seems inappropriate since it seeks to maximize all outputs simultaneously. This dissertation proposed to use the nonparametric direction output distance function. The model is a linear programming problem. Solving it can identify a set of airports that form a linear piecewise efficient production frontier. For inefficient airports, it quantifies the levels of inefficiency. In addition, the maximum possible production can also be estimated to understand how much the potential outputs are. Results are beneficial in many management regards such as performance measurement, benchmarking, ranking and policy development. The approach was applied in case study 2 to assess the productivity of 56 major commercial U.S. airports. A recent panel data from 2000 – 2003 were used. In order to analyze the impact of the inclusion of the undesirable outputs, a model without accounting for delays was also estimated. There are several important findings and
- 178 -
Assessment of Productive Efficiency of Airports
insightful implications as discussed in Chapter 7. This chapter concludes with those discussions.
8.2.1 Productivity of U.S. commercial airports
Among 56 airports, approximately half of them are identified as efficient during 2000 – 2003. These airports include busy ones such as Hartsfield-Jackson Atlanta (ATL), Los Angeles International (LAX), LaGuardia (LGA), Memphis (MEM), Phoenix Sky Harbor International (PHX), San Diego International (SAN), and John Wayne airport (SNA). Other well-known busy airports such as O'Hare International (ORD), Midway International (MDW), Newark Liberty International (EWR), John F. Kennedy International (JFK), Anchorage International (ANC), Miami International (MIA), Seattle Tacoma International (SEA) and Lambert-St. Louis International (STL) though not classified as efficient show very low inefficiency levels. In addition, the model also identified several other less busy airports as efficient They include Birmingham International (BHM), Baltimore/Washington International (BWI), Port Columbus International (CMH), and Oakland International (OAK). These airports are credited because they have relatively low delays. The results indicate that there may be a balance between quantity and quality of outputs in the achievement of efficient outcomes; i.e., airports can trade-off utilization levels for reduced flight and time delays. For certain stakeholders, this option may be an optimal strategy. Passengers and shippers receive service with fewer flight delays. The FAA, as the regulator, has less concern over congestion and safety. Meanwhile, airport managers are able to balance
- 179 -
Assessment of Productive Efficiency of Airports
traffic volume with customer satisfaction. By all accounts, the inclusion of undesirable outputs in the analysis appears to provide a fairer assessment of airport efficiency. In 2003, the overall system (56 airports) has potential to accommodate increases of about 30% (~1,550 million passengers), 26% (~ 26 million movements) and 43% of total passengers, aircraft movements and cargo throughput, respectively. In numbers, these amounts are equivalent to totally 1,550 million passengers, 26 million movements and 34 million tons of cargo. This would make the system operate at the maximum possible production level. The estimated potential outputs vary across airports. For airport planning, the figure provides a good estimation of excess capacity. An airport manager may use this information in planning an airport improvement program. Finally, it is observed that when there is a major new investment in an airport, its productivity decreases during early years after the construction. This is the case for Detroit Metropolitan Wayne County (DTW) and George Bush Intercontinental (IAH). DTW opened its sixth runway in 2001 whereas IAH finished constructing runway expansion and extension in 2002.
8.2.2 Productivity growth of U.S. commercial airports
During the period 2000 – 2003, the changes in productivity are rather modest in the narrow range of -1.3% to +1.8%. The airport system on the average had productivity gains in two periods during 2000/2001 and 2001/2002 and productivity loss in 2002/2003. Between 2000 and 2001, the overall average rise in efficiency was 1.8 percent; and continued to increase by 1.2 percent in the next period before falling down
- 180 -
Assessment of Productive Efficiency of Airports
slightly 1.3 percent during 2002 – 2003. The slow growth and regress may be associated with the September 11 terrorist attacks which shook aviation industry worldwide and still had effects during the analysis period. The netted 1.8% productivity gains between 2000 and 2001 are mainly from efficiency change (airports become +8.8% more efficient) which compensates 7.0% productivity loss from technical change (frontier-shift effect). The situation is opposite in 2001/2002 period when overall 1.2% productivity gains mainly resulted from frontier shift (4.7%). Between 2002 and 2003 both efficiency loss (1.2%) and frontier shift (0.1%) collectively contribute to the overall 1.3% productivity loss.
8.2.3 Impact of delays on airport productivity
It is found out that by ignoring undesirable outputs, i.e., delayed flights and time delays, the results are drastically different in many important aspects. First of all, only a handful of airports (i.e., 6 to 7 depends on the year) are classified as efficient. Exclusively, they are very busy airports that include Hartsfield-Jackson Atlanta (ATL), Los Angeles International (LAX), LaGuardia (LGA), Memphis (MEM), Phoenix Sky Harbor International (PHX), San Diego International (SAN), and John Wayne airport (SNA). All other airports are classified as inefficient with different degrees. Unless traffic is not exceptionally dense, the airport will never be identified as efficient. This is because delays are out of the assessment. Second, the level of inefficiency as reflected through the efficient score is generally much higher. This is proven by several statistical tests. Consequently, airport
- 181 -
Assessment of Productive Efficiency of Airports
performance looks very poorly although it may not be the case from stakeholders’ perception. For instance, the results suggest that Boston Logan International (BOS) handled less than 50% of the level that it should have been able to handle. More precisely, it should have handled about 50 million passengers in 2003 (Table 5.8), rather than just 22.79 million (Table 4.1). The level of inefficiency may be overestimated when delays are not taken into the assessment. Third, the relative ranking of airports may also be distorted. In particular, smaller and less busy airports that are deemed to be inefficient may appear on the efficient frontier when delays are added as undesirable outputs. Fourth, the estimated maximum possible production may not be reasonable and practical. The results indicate potential increases of traffic from current levels at around 133%, 91%, and 364% as compared to around 23%, 20% and 35% when delays are considered. The discrepancy may be interpreted as amount of output loss due to cleaning up delays or keeping them at relatively low levels. It can also represent the tradeoff that an airport has to bear in exchange for higher quality of service. Fifth, the computed productivity indexes are statistically different when delays are accounted for. In many cases, the indexes provided opposite inference regarding the productivity growth. This is actually not surprising since the computation of indexes takes different sets of efficient scores; the resulting indexes are not necessarily similar. The point is that it is crucial to use the right efficiency scores so that the indexes will be meaningful. These are deemed to be the ones with account for undesirable output.
- 182 -
Assessment of Productive Efficiency of Airports
In conclusion, the assessment which does not consider joint production of desirable and undesirable byproduct such as delays will give biased measurements of airport productivity. The interpretation of results can be misleading. Any computation afterwards based on unreasonable efficiency scores including productivity indexes can be confusing. It is strongly recommended to take undesirable outputs into consideration since the results seem to be more reasonable and practical.
8.2.4 Selection of input and output measures
Scenario analyses (i.e., with and without delays, with and without land area) provide insights regarding the effects of chosen measures on the sensitivity of productivity. It is true in general that as the number of input and output measures increase, there will be more airports that are deemed efficient. Note however that the increase in number of airports on the efficient frontier was more dramatic when delays were added than when an additional input (land area) was added. This suggests that consideration of undesirable outputs is at least as important as the consideration of additional input in determining productive efficiency of airports. The failure to include undesirable outputs in the assessment of airport productivity could lead to misleading results. It is concluded here that selection of input, desirable and undesirable outputs should be carefully considered in tandem in order to provide meaningful, yet practical results. Ignoring undesirable outputs (such as delays) could lead to unwise policy choices for managing airports. For example, unless funding agencies or regulators give credit or rewards to airports for keeping delays at low levels, there will be little motivation to
- 183 -
Assessment of Productive Efficiency of Airports
improve quality of service. Instead, airports may prefer to focus on accommodating increasing levels of traffic without considering downside of these services.
8.2.5 Factors affecting productive efficiency of U.S. airports
A Censored Tobit regression model was estimated for explaining the variations in efficiency scores of airport operations. Five groups of exploratory variables were investigated, i.e., airport characteristics, airport services, level of utilization, market characteristics, and service characteristics. It was found that factors such as utilization of land area and runway area contribute to the enhancement of productivity. Meanwhile higher proportion of international passengers and average delay per passenger tend to reduce the productivity. The model also captured anomaly effects in the years 2001, 2002 and 2003 (it was observed that an airport could become efficient slightly more easily with the same utilization rate). Given some planned measures, the model can be used to predict future productivity of an airport and should be very useful as a tool for planning airport business in the U.S. and other geographical regions.
8.3 Suggested future research
This research pioneers the work on the assessment of productivity of airports operating in MASs and when joint production of desirable and undesirable outputs is taken into consideration. It opens up new opportunities for aviation researchers and practitioners to better understand the relation between inputs and outputs of airport operations. There are several potential extensions to this research that could be conducted in the future. Some of them are suggested here.
- 184 -
Assessment of Productive Efficiency of Airports
1) Consideration of comprehensive input and output measures An attempt may be made to collect other input and output measures and take them into consideration for assessing the productivity of U.S. and international airports. The input measures in this dissertation may be rather limited to the airside operation. In fact, one may want to see how other capital inputs such as number of gates, terminal area, and apron area could impact the productivity of airports. Financial inputs are also important for airport operations. Environmental factors (e.g., population density, accessibility, and market condition) also have significant impact on traffic volume which in turn affects productivity of airport operations. On the undesirable outputs, although this dissertation considers perhaps the most conceivable undesirable outputs i.e., delays but there are other undesirable outputs that airport stakeholders are also concerned with such as the number of mishandled baggage and accidents. Even delays could be expanded to encompass a wider number of delay causes. Externality such as noise is perhaps the most frequently-cited undesirable byproduct during the airport planning process. With the current technology, there is no way to get rid of them. However, no study has ever taken them into consideration while assessing productive efficiency of airport operation. It will be very interesting to see how externalities could affect the productivity of airports. Future research may include them into the model. One might argue that unless these inputs and output measures are not accounted for, the performance measures may be misleading. In this line of research extension, a lot of effort and resources are needed to collect the data since there does not seem to be a
- 185 -
Assessment of Productive Efficiency of Airports
consolidated database. As this dissertation showed, different sets of inputs and outputs can lead to very different results. A reasonable question is how to choose a set of input/output measures that yields robust results, yet is meaningful for airport management. 2) Application in the international context It will be very useful if one can extend the study framework to assess productivity of airports in the global context so that the valuable lessons may be learned from truly efficient airports, rather than benchmark among U.S. airports only. However, comparison of airports across nations is not an easy task. There will be several other factors involved that affect the efficiency of airport operations. For example, differences in organizational structure may provide different levels of control to airport managers. The definition and measurement of inputs and outputs are also an issue since different countries may adopt different approaches. Again, data availability will be a major hurdle. 3) Better understanding of factors affecting productive efficiency Many studies have focused on assessing productivity, but relatively few paid attention to the development of prediction models. More research effort may be put forth toward the development of casual models for explaining variation in airport productivity. Such models will enable the managers and policy makers to better understand factors that can enhance operational efficiency. In this area, one may want to investigate effects of other variables beyond those considered in this dissertation. For instance, it is interesting to study the effects of the common ownership of airports in an MAS on their
- 186 -
Assessment of Productive Efficiency of Airports
productivity. One might expect a higher efficiency due to strong coordination, but this has never been studied before. There is a lot more room for research in this direction. 4) Application to other transportation modes It is not an exaggeration to say that transportation activities create undesirable byproducts, regardless of modes. Bus, rail, and water-transport systems all create air pollution. Accidents occur every day on highways. Delays are incurred in all transportation modes. The proposed methodology is certainly well-suited for assessing productivity of other transportation modes. Policy makers may want to know performance of transportation services if these undesirable byproducts are considered. Recently, some researchers (mostly from the economics discipline) started looking at productivity of bus transit by considering joint production of desirable measures and pollutants (e.g., NO2 and CO2) (Noh and McMullen, 2006). Productivity and efficiency of trucking industry accounting for traffic fatalities is studied by Weber and Weber (2004). 5) Theoretical development As for DEA, there is much room for further developing the directional output distance function approach to treat certain cases. For example, it may be adapted to deal with categorical input or output measures such as operating conditions (e.g., snow-belt or not, hub or non-hub, existence of noise abatement program). With this model, it is possible to make an analysis closer to a like comparison which in turn provides fairer and more meaningful results for airport management. Furthermore, the model can be
- 187 -
Assessment of Productive Efficiency of Airports
developed to allow for non-radial expansion and contraction for use in the case that policy makers can reveal preference toward individual input and output measures.
- 188 -
Assessment of Productive Efficiency of Airports
REFERENCES
Abbott M., Wu S. (2002) Total factor productivity and efficiency of Australian airports, The Australian Economic Review, Volume 35, Number 3, pp. 244 – 260. Adler N., Berechman J. (2001) Measuring airport quality from the airlines’ viewpoint: an application of Data Envelopment Analysis. Transport Policy, Volume 8, Issue 3, pp. 171 – 181. Air Transport Research Society (2002) Airport benchmarking report: the global standard for airport excellence, ATRS, Vancouver. Air Transport Research Society (2003) Airport benchmarking report: the global standard for airport excellence, ATRS, Vancouver. Airports Council International (2002) Worldwide Airport Traffic Report – 2001. Airports Council International. (2003) Worldwide Airport Traffic Report – 2002. Airports Council International (2004) Worldwide Airport Traffic Report – 2003. Amemiya T. (1984) Tobit models: A survey, Journal of Econometrics, Volume 24, Number 1 – 2, pp. 3 – 61. Anderson P, Peterson N.C. (1993), A procedure for ranking efficient units in Data Envelopment Analysis, Management Science, Volume 39, Number 10, pp. 1261 – 1264. Ashford N. (1999) Airport finance, Wiley, London.
- 189 -
Assessment of Productive Efficiency of Airports
Banker R.D. (1984) Estimating most productive scale size using data envelopment analysis, European Journal of Operational Research, Volume 17, Issue 1, pp. 35 – 44. Banker R.D., Charnes A., Cooper W.W. (1984) Some models for estimating technical and scale efficiencies in Data Envelopment Analysis, Management Science, Volume 30, Number 9, pp. 1078 – 92. Banker R.D., Charnes A., Cooper W. W., Swarts J., Thomas and D.A. (1989) An introduction to Data Envelopment Analysis with some of its model and their uses, Research in governmental and Nonprofit Accounting, JAI Press, Connecticut 5, pp. 125 – 163. Banker R.D., Morey R.C. (1986) The use of categorical variables in Data Envelopment Analysis, Management Science, Volume 32, Number 12, pp. 1613 – 1627. Banker R.D., Thrall R.M. (1992) Estimation of returns to scale using Data Envelopment Analysis, European Journal of Operational Research, Volume 62, Issue 1, pp. 74 – 84. Bazargan M., Vasigh B. (2003) Size versus efficiency: a case study of US commercial airports, Journal of Air Transport Management, Volume 9, Issue 3, pp. 187 – 193. Berlin Brandenburg International (2005) Schedule,http://www.berlinairport.de/bbi/rubEnglish/rubProjekt/rubZeitplan/index.html, accessed on July 30, 2005.
- 190 -
Assessment of Productive Efficiency of Airports
Boame K.A. (2004) The technical efficiency of Canadian urban transit system, Transportation Research Part E, Volume 40, Issue 5, pp. 401 – 416. Boame K.A., Obeng K. (2005) Sources of productivity change: a Malmquist total factor productivity approach, Transport Reviews, Volume 25, Number 1, pp. 103 – 116. Boussofiane A, Dyson R. G., Thanassoulis E (1991) Applied Data Envelopment Analysis European Journal of Operational Research, Volume 52, Number 1, pp. 1 – 15. Brockett P L., Golony B. (1996) Using rank statistics for determining programmatic efficiency differences in Data Envelopment Analysis. Management Science, Volume 42, Number 3, pp. 466 – 472. Bureau of Transportation Statistics (2004) Transportation Statistics Annual Report. Bureau of Transportation Statistics (2005) Transportation Statistics Annual Report. Bureau of Transportation Statistics (2006) TranStats,http://www.transtats.bts.gov/, Accessed on March 31, 2006. Caves D. W, Christensen L. R., Diewert W. E. (1982) The economic theory of index numbers and the measurement of input, output and productivity, Econometrica 50, pp. 303 – 313. Caves R.E., Gosling G.D. (1999) Strategic Airport Planning, Elsevier Science Ltd., The Netherlands.
- 191 -
Assessment of Productive Efficiency of Airports
Chambers R. G. (1996) A new look at exact input, output, productivity and technical change measurement, Maryland Agricultural Experiment Station, Department of Agricultural and Resource Economics, University of Maryland, College Park. Chambers R. G., Färe R., Grosskopf S. (1996) Productivity growth in APEC countries, Pacific Economic Review, Volume 1, Issue 3, pp. 181 – 190. Charles River Associates Incorporated (2001) Reductions in flight operations as an alternative to runway reconfiguration at San Francisco International Airport, Final Report, CRA No. D02730-00. Charnes A., Cooper W. W., Lewin A. Y., Seiford L. M. (editors) (1994), Data Envelopment Analysis: theory, methodology and applications, Kluwer Acadeic Publishers, USA. Charnes A., Cooper W. W., Rhodes E. (1978) Measuring the efficiency of decision making units, European Journal of Operational Research, Volume 2, Number 6, pp. 429 - 444. Chung Y. H., Färe R., Grosskopf S. (1997) Productivity and undesirable outputs: a directional output distance function approach, Journal of Environmental Management, Volume 51, Issue 3, pp. 229 – 240. Coelli T., Perelman S. (1999) A comparison of parametric and non-parametric distance functions: with application to European railways, European Journal of Operational Research, Volume 117, Issue 2, pp. 326 – 339.
- 192 -
Assessment of Productive Efficiency of Airports
Cook W. D., Zhu J. (2005) Modeling performance measurement: applications and implementation issues in DEA, Springer, USA. Cooper W.W., Seiford L.M., Thanassoulis E., Zanakis S.H. (editors) (2004) DEA and its uses in different countries, European Journal of Operational Research, Volume 154, Issue 2, pp. 337 – 344. Cooper W.W., Seiford L. M., Tone K. (2000) Data Envelopment Analysis a comprehensive text with models, applications, references and DEA-solver software, Kluwer Academic Publishers. Cooper W. W., Seiford L. M., Zhu J. (2004) Data Envelopment Analysis: history, models and interpretations, Handbook on Data Envelopment Analysis, Kluwer Academic Publishers, USA, pp. 1 – 39. de Neufville R. (1995) Management of Multi-Airport Systems A Development Strategy. Journal of Air Transport Management. Volume 2, No. 2, pp. 99 – 110. de Neufville R., Odoni A. (2003) Airport Systems Planning, Design, and Management, McGraw-Hill Inc. Doganis R. (1992) The airport business, Routledge, UK. Doyle J., Green R. (1994) Efficiency and cross-efficiency in DEA: derivations, meanings and uses. Journal of the Operational Research Society, Volume 45, Number 5, pp. 567 – 578.
- 193 -
Assessment of Productive Efficiency of Airports
Emrouznejad A. (2006) DEA homepage,http://www.deazone.com/, accessed June 1, 2006. Färe R., Grosskopf S. (2004a) Modeling undesirable factors in efficiency evaluation: comment, European Journal of Operational Research 157, 242 - 245. Färe R., Grosskopf S. (2004b) New Directions: Efficiency and Productivity, Kluwer Academic Publishers. Färe R., Grosskopf S., Lindgren B., Roos P. (1994a) Productivity Developments in Swedish Hospitals: A Malmquist Output Index Approach, Data Envelopment Analysis theory, methodology and applications, Charnes A., Cooper W., Lewin A. Y., Seiford L. M. (editors), Kluwer Academic Publishers, 253 - 272. Färe R., Grosskopf S., Norris M., Zhang Z. (1994b) Productivity growth, technical progress and efficiency change in industrialized countries. The American Economic Review, Volume 84, Number 1, pp. 66 – 83. Färe R., Grosskopf S., Roos P. (1998) Malmquist productivity indexes: a survey of theory and practice, Index numbers: essays in honour of Sten Malmquist, Färe R., Grosskopf S., Russell R. R. (editors), Kluwer Academic Publishers, 127 - 190. Färe R., Grosskopf S., Whittaker G. (2004) Distance Functions with Applications to DEA, Handbook on Data Envelopment Analysis Directions: Efficiency and Productivity, Cooper W.W., Seiford L.M, Zhu J. (editors), Kluwer Academic Publishers, 139 - 152.
- 194 -
Assessment of Productive Efficiency of Airports
Färe R., Primond D. (1995) Multi-output production and duality: theory and applications, Kluwer Academic Publishers, Boston. Farrell M.J. (1957) The measurement of productive efficiency, Journal of the Royal Statistical Society, Series A, Volume 120, pp. 253 - 351. Federal Aviation Administration (2004a), the Terminal Area Forecast System,http://www.apo.data.faa.gov/faatafall.HTM, Accessed March 1, 2004. Federal Aviation Administration (2004b) Airport Data (5010),http://www.faa.gov/data_statistics/aviation_data_statistics/, Accessed March 1, 2004. Federal Aviation Administration (2005) FAA Operations & Performance Data: Operational Network (OPSNET),http://www.apo.data.faa.gov/, Accessed on August 29, 2005. Federal Aviation Administration (2006) FAA Operations & Performance Data: Airline Service Quality Performance (ASQP),http://www.apo.data.faa.gov/, Accessed on March 31, 2006. Fernandes E., Pacheco R. R (2001) Total airport productivity in Brazil, a paper presented at XIIth World Productivity Congress, Hong Kong, November 2001. Fernandes E., Pacheco R.R. (2002) Efficient use of airport capacity, Transportation Research Part A, Volume 36, Issue 3, pp. 225 – 238.
- 195 -
Assessment of Productive Efficiency of Airports
Fernandes E., Pacheco R.R. (2005) Impact of changes in management style on the performance of Brazilian airports, Proceedings of the Air Transport Research Society (ATRS) conference, Rio de Janeiro, Brazil, July 3 – 7, 2005. Francis G., Humphreys I. (2001) Airport regulation: reflecting on the lessons from BAA plc, Public Money & Management, January – March, pp. 49 – 52. Francis G., Humphreys I., Fry J. (2002) The benchmarking of airport performance, Journal of Air Transport Management 8, pp 239 – 247. Ganley J. A., Cubbin J. S. (1992) Public sector efficiency measurement: applications of Data Envelopment Analysis, Elsevier. Gillen D., Lall A. (1997) Developing measures of airport productivity and performance: an application of Data Envelopment Analysis. Transportation Research Part E, Volume 33, Number 4, 261 – 273. Gillen D., Lall A (1998) Non-parametric measures of efficiency of U.S airports, Proceedings of the Air Transport Research Society (ATRS) conference, Dublin, Ireland, July 20 - 21, 1998. Graham A. (2003) Managing airports: an international perspective, 2nd edition, Elsevier, UK. Greene W.H. (2002) LIMDEP Version 8.0 Econometric Modeling Guide, Volume 2, Econometric Software Inc., USA. Greene W.H. (2003) Econometric analysis, 5th Edition, Prentice Hall, USA.
- 196 -
Assessment of Productive Efficiency of Airports
Hansen M., Weidner T. (1995) Multiple Airport Systems in the United States: Current Status and Future Prospects, Transportation Research Record, Number 1506, pp. 8 – 17. Haynes K.E., Stough R.R., Shroff H.F.E. (1990), New methodology in context: Data Envelopment Analysis, Computers, Environment and Urban Systems, Volume 14, Number 2, pp. 85 – 175. Hooper P.G., Hensher D.A. (1997) Measuring total factor productivity of airports – an index number approach, Transportation Research Part E, Volume 33, Number 4, 249 – 259. Humphreys I. (1999) Privatisation and comercialisation: changes in UL airport ownership patterns, Journal of Transport Geography 7 (2), pp. 121 – 134. Humphreys I., Francis G. (2002) Performance measurement: a review of airports, International Journal of Transport Management 1, pp. 79 – 85. Itoh H. (2002) Efficiency changes at major container ports in Japan: a window application of Data Envelopment Analysis, The Applied Regional Science Conference (ARSC), Volume 14, Number 2, pp. 133 – 152. Jarach D. (2005) Airport marketing: strategies to cope with the new millennium environment, Ashgate Publishing Limited. Kerstens K. (1996) Technical efficiency measurement and explanation of French urban transit companies, Transportation Research Part A, Volume 30, Number 6, pp. 431 – 452.
- 197 -
Assessment of Productive Efficiency of Airports
London City Airport Consultative Committee (2005) Constructing the Airport,http://www.lcacc.org/history/construction.html, accessed on July 31, 2005. Maddala G.S. (1983) Limited-dependent and Qualitative Variables in Econometrics, Cambridge University Press, USA. Malmquist S. (1953) Index numbers and indifference surfaces, Trabajos de Estatistica, Volume 4, pp. 209 – 242. Martin J. C., Roman C. (2001) An application of DEA to measure the efficiency of Spanish airports prior to privatization, Journal of Air Transport Management, Volume 7, Issue 3, 149 – 157. Martin-Cejas R.R. (2002) An approximation to the productive efficiency of the Spanish airports network through a deterministic cost frontier, Journal of Air Transport Management, Volume 8, Issue 4, pp. 233 – 238. Martinez-Budria E., Diaz-Armas R., Navarro-Ibanez M., Ravelo-Mesa T. (1999) A study of the efficiency of Spanish port authorities using Data Envelopment Analysis, International Journal of Transport Economics, Volume XXVI, Number 2, pp. 237 – 253. Murillo-Melchor C. (1999) An analysis of technical efficiency and productivity changes in Spanish airports using the Malmquist index, International Journal of Transprot Economics, Volume XXVI, Number 2, pp. 271 – 292.
- 198 -
Assessment of Productive Efficiency of Airports
Noh D.W., McCullen B. S. (2006), Accounting for environmental standards in the measurement of bus transit efficiency: a directional output distance function approach, Manuscript, submitted. Norlander N. O., Roos P. (1998) Implementing productivity index: the case of the national corporation of Swedish pharmacies, Index numbers: essays in honour of Sten Malmquist, Färe R., Grosskopf S., Russell R. R. (editors), Kluwer Academic Publishers, 191 – 216. Nyshadham E. A., Rao V. K. (2000) Assessing efficiency of European airports: a total factor productivity approach, Public Works Management & Policy, Volume 5, Number 2, pp 106 – 114. Oum T.H., Yu C. (2004) Measuring airports’ operating efficiency: a summary of the 2003 ATRS global airport benchmarking report. Transportation Research Part E 40, 515 – 532. Oum T. H., Yu C., Fu X. (2003) A comparative analysis of productivity performance of the world’s major airports: summary report of the ATRS global airport benchmarking research report – 2002. Journal of Air Transport Management, Volume 9, Issue 5, 285 – 297. Pacheco R.R., Fernandes E. (2003) Managing efficiency of Brazilian airports, Transportation Research Part A, Volume 37, Issue 8, 667 – 680. Parker D. (1999) The performance of BAA before and after privatization, Journal of Transport Economics and Policy, Volume 33, Part 2, pp. 133 – 146.
- 199 -
Assessment of Productive Efficiency of Airports
Pathomsiri S., Haghani A. (2004) Benchmarking efficiency of airports in the competitive multiple-airport systems: the international perspective, Proceedings of the 19th Transport Chicago Conference (CD-ROM), Chicago, IL. Pathomsiri S., Haghani A. (2005) Taste Variations in Airport Choice Models, Journal of Transportation Research Record 1915, Transportation Research Board, National Research Council, Washington DC, pp. 27 – 35. Pathomsiri S., Haghani A., Schonfeld P. S. (2005) Operational efficiency of airports in multiple-airport systems, Transportation Research Board 84th Annual Meeting Compendium of Papers (CD-ROM), Washington DC. Pathomsiri S., Haghani A., Dresner M., Windle R. J. (2006a), Measurement and determination of airport productivity in competitive markets, Transportation Research Board 85th Annual Meeting Compendium of Papers (CD-ROM), Washington DC. Pathomsiri S., Haghani A., Dresner M., Windle R. J. (2006b) Productive and Operational Efficiency of US Airports with Joint Consideration of both Desirable and Undesirable Outputs, the 47th Annual Transportation Research Forum Proceedings (CD-ROM), New York, March 23 – 25, 2006. Pathomsiri S., Haghani A., Dresner M., Windle R. J. (2006c) Assessment of Airport Performance with Joint Consideration of Desirable and Undesirable Outputs, the 10th Annual World Conference Air Transport Research Society (ATRS) Proceedings (CD-ROM), Nagoya, Japan, May 26 – 28, 2006.
- 200 -
Assessment of Productive Efficiency of Airports
Pathomsiri S., Haghani A., Windle R.J., Dresner M. (2006) Role of Freight Transportation Services and Reduction of Delayed Flights on Productivity of US Airports, National Urban Freight Conference 2006 Proceedings (CD-ROM), Long Beach, CA, February 1 – 3, 2006. Pathomsiri S., Mahmassani H.S., Haghani A. (2004) Airport Choice Models for Baltimore-Washington Region, Transportation Research Board 83rd Annual Meeting Compendium of Papers (CD-ROM), Washington DC, January 2004. Pels, E., Nijkamp P., Rietveld P. (2001) Relative efficiency of European airports, Transport Policy, Volume 8, 183 – 192. Pels E., Nijkamp P., Rietveld P. (2003) Inefficiencies and scale economies of European airport operations, Transportation Research Part E, Volume 39, Issue 5, 341 – 361. Pina V., Torres L. (2001) Analysis of the efficiency of local government services delivery: an application to urban public transport, Transportation Research Part A, Volume 35, Issue 10, pp. 929 – 944. Pires Capobianco H.M., Fernandes E. (2004) Capital structure in the world airline industry, Transportation Research Part A, Volume 38, Issue 6, pp. 421 – 434. QMS (2005) Quantitative Micro Software,http://www.eviews.com, accessed on July 30, 2005. Ray S. C. (2004) Data Envelopment Analysis theory and techniques for economics and operations research, Cambridge University Press.
- 201 -
Assessment of Productive Efficiency of Airports
Ross A.D., Droge C. (2004) An analysis of operations efficiency in large-scale distribution systems, Journal of Operations Management, Volume 21, Issue 6, pp. 673 – 688. Rousseau J. J, Semple J. H. (1995) Radii of classification preservation in Data Envelopment Analysis: a case study of ‘program follow-through’. Journal of the Operational Research Society, Volume 46, Number 8, pp 943 – 957. Salazar De La Cruz F. (1999) A DEA approach to the airport productivity function, International Journal of Transport Economics, Volume XXVI, Number 2, pp. 255 – 270. Sarkis J. (2000) An analysis of the operational efficiency of major airports in the United States. Journal of Operations Management, Volume 18, Issue 3, 335 – 351. Sarkis J., Talluri S. (2004) Performance based clustering for benchmarking of U.S. airports. Transportation Research Part A, Volume 38, Issue 5, 329 -346. Schefczyk M. (1993) Operational performance of airlines: an extension of traditional measurement paradigms, Strategic Management Journal, Volume 14, pp. 301 – 317. Scheraga C.A. (2004) Operational efficiency versus financial mobility in the global airline industry: a data envelopment and Tobit analysis, Transportation Research Part A, Volume 38, Issue 5, pp. 383 - 404. Seiford L.M. (1996) Data Envelopment Analysis: the evolution of the state of the art (1978 – 1995), Journal of Productivity Analysis, Volume 7, pp. 99 – 137.
- 202 -
Assessment of Productive Efficiency of Airports
Seiford L.M., Zhu J. (2002) Modeling undesirable factors in efficiency evaluation, European Journal of Operational Research 142, 16 – 20. Seiford L.M., Zhu J. (2005) A response to comments on modeling undesirable factors in efficiency evaluation, European Journal of Operational Research 161, 579 – 581. Shephard R. W. (1970) Theory of cost and production functions, Princeton University Press, New Jersey. StataCorp (2005), Stata Statistical Software for Professionals,http://www.stata.com, accessed on July 30, 2005. Stewart International Airport (2005) Stewart History,http://www.stewartintlairport.com/home.html, accessed on Jul 31, 2005. Tobin J. (1958) Estimation of Relationships for Limited Dependent Variables, Econometrica, Volume 26, Number 1, pp. 24 – 36. Tretheway M. (1995) Book review (The Economic Performance of European Airports R.S. Doganis, A. Graham and A. Lobbenberg), Journal of Air Transport Management, Volume 2, Issues 3-4, p. 207. Tone K, Malmquist productivity index: efficiency changes over time, Handbook on Data Envelopment Analysis directions: efficiency and productivity, Cooper W.W., Seiford L.M, Zhu J. (editors), Kluwer Academic Publishers, 203 – 228. Tongzon J.L. (1995) Determinants of port performance and efficiency, Transportation Research Part A, Volume 29, Number 3, pp. 245 – 252.
- 203 -
Assessment of Productive Efficiency of Airports
Tongzon J. (2001) Efficiency measurement of selected Australian and other international ports using data envelopment analysis, Transportation Research Part A, Volume 35, Issue 2, pp. 107 – 122. Turner H., Windle R., Dresner M. (2004) North American containerport productivity: 1984 – 1997, Transportation Research Part E, Volume 40, Issue 4, pp. 339 – 356. Weber M.M, Weber W.L. (2004) Productivity and efficiency in the trucking industry accounting for traffic fatalities, International Journal of Physical Distribution & Logistics Management, Volume 34, Number 1, pp. 39 – 61. Windle R.J., Dresner M.E. (1992) Partial productivity measures and total factor productivity in the air transport industry: limitations and uses, Transportation Research Part A, Volume 26, Number 6, pp. 435 – 445. Yoshida Y. (2004) Endogenous-weight TFP measurement: methodology and its application to Japanese-airport benchmarking, Transportation Research Part E 40, 151 – 182. Yoshida Y., Fujimoto H. (2004) Japanese-airport benchmarking with the DEA and endogenous- weight TFP methods: testing the criticism of overinvestment in Japanese regional airports, Transportation Research Part E 40, 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, 295 – 303.
- 204 -
Assessment of Productive Efficiency of Airports
Zaim O. (2005) A framework for incorporating environmental indicators to the measurement of human well-being, Research paper No. 2005/05, World Institute for Development Economics Research. Zhu J. (2003) Quantitative models for performance evaluation and benchmarking Data Envelopment Analysis with spreadsheets and DEA Excel Solver, Kluwer Academic Publishers.
- 205 -
Assessment of Productive Efficiency of Airports
Bibliography
Mr. Somchai Pathomsiri was born and raised in Bangkok, the capital city of the Kingdom of Thailand. He earned three engineering degrees, i.e., B.Eng. (Honor) in Civil Engineering from King Mongkut Institute of Technology Thonburi (now King Mongkut's University of Technology Thonburi); M.Eng. in Traffic and Transportation Systems Engineering from the Department of Civil Engineering, Chulalongkorn University, Thailand; and PhD in Transportation Systems Engineering and Planning from the Department of Civil and Environmental Engineering, University of Maryland, USA. All three degrees were fully supported by merit-based scholarship. In addition, he also received an MBA (Executive program) from Thammasat University, Thailand. He began his career in the industry working in various roles including civil engineer, traffic engineer, transport engineer, project engineer, project manager and engineering manager. He had worked extensively in construction and transportation planning projects for both private and public sectors. His experience in transportation spans into all modes (i.e., land, water and air). For example, he worked on several feasibility studies and environmental impact assessment for highway, expressway, port and airports projects. He has been a registered professional engineer (P.E.) in Thailand since 1994. He had been a faculty member at the Department of Civil Engineering, Faculty of Engineering Mahidol University since 1997. He is taking care construction management and transportation program. His research interests include Logistics, transportation and
- 206 -
Assessment of Productive Efficiency of Airports
traffic systems analysis, travel behavior & demand modeling, transportation planning, GIS, aviation systems planning, applications of information technology in traffic management, transportation safety and security, transportation economics and infrastructure management. Presently, he lives happily with his amazing wife, Laddawan and three wonderful children, Nawanont (FeiFei), Nontthida (FinFin) and Phuwanont (FarnFarn) in Bangkok, Thailand.
- 207 -
doc_223500467.pdf