This article was originally published in a journal published by Elsevier, and the attached copy is provided by Elsevier for the author’s benefit and for the benefit of the author’s institution, for non-commercial research and educational use including without limitation use in instruction at your institution, sending it to specific colleagues that you know, and providing a copy to your institution’s administrator. All other uses, reproduction and distribution, including without limitation commercial reprints, selling or licensing copies or access, or posting on open internet sites, your personal or institution’s website or repository, are prohibited. For exceptions, permission may be sought for such use through Elsevier’s permissions site at: http://www.elsevier.com/locate/permissionusematerial

ARTICLE IN PRESS

Journal of Air Transport Management 12 (2006) 342–351 www.elsevier.com/locate/jairtraman

co

L.C. Lin, C.H. Hong

py

Operational performance evaluation of international major airports: An application of data envelopment analysis Department of Logistics Management, National Kaohsiung First University of Science and Technology, Taiwan

al

Abstract

on

This paper uses data envelopment analysis to assess the operational performance of 20 major airports around the world. It is found that the form of ownership and the size of an airport are not apparently correlated with operational performance of airports. In contrast, the existence of a hub airport, the location of the airport, and the economic growth rate of the country in which the airport is located are all related to the operational performance of airports. The 20 airports are put into four groups according to their efficiency values. r 2006 Elsevier Ltd. All rights reserved.

rs

Keywords: International airports; Operational performance evaluation; Data envelopment analysis

pe

1. Introduction

Au

th o

r's

Since the 1970s, the air transportation industry has been undergoing a process of deregulation and privatization. This deregulation has stimulated demands for faster and more efficient processing of aircrafts, passengers and cargos through airports. As the continental markets in North America, Europe, Australia, and Asia become more competitive, airlines have more alternatives from which to choose in deciding where to base their domestic hubs and intercontinental gateways and how to route their connecting flights (Oum et al., 2003). According to the International Civil Aviation Organization (ICAO, 2003), it is estimated that $25 billion will be invested in airport construction before 2010. It is thus essential for airport managers to identify the best practices in a range of airport operations and to provide the best services in the most efficient manner (Chin and Siong, 2001; Forsyth, 2003). However, most of the studies on the efficiency of airports have been confined to a particular country or geographical area. The present study extends research into the operational performance of airports across international borders.

Corresponding author. Fax: +886 7 6011040.

E-mail address: [email protected] (L.C. Lin). 0969-6997/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.jairtraman.2006.08.002

In analyzing multiple decision-making units (DMUs), multiple inputs, and multiple outputs, the data envelopment analysis (DEA) model is a useful tool. The model does not assume a functional form in advance, and it can determine the weight of each variable individually. Because the reference set in the DEA model can be used for benchmarking purposes, the model is especially suitable for evaluation of operational performance. The present study thus evaluates the operational performance of major international airports using the DEA model. The study proposes and tests certain hypotheses about the relationships between airport operational performance and five important features of airports.

2. Conceptual framework and hypotheses The Airport Capacity/Demand Profiles (2003) adopted three criteria (passengers, cargos, and movements) in ranking the most competitive international airports. Under each criterion, a ranking of the top 30 airports was established. Twenty airports appeared on the lists two or more times, and these airports were selected as DMUs for this research. The twenty airports included ten airports in North America (Atlanta Hartsfield, Chicago O’Hare, Los Angeles, Dallas Fort Worth, Denver, McCarran, Newark, Kennedy, Miami, and Pearson), five airports in Europe

ARTICLE IN PRESS L.C. Lin, C.H. Hong / Journal of Air Transport Management 12 (2006) 342–351

H30. There is no significant difference in operational performance between public airports and mixed ownership airports.

Large airports: Atlanta Hartsfield, Chicago O’Hare, London Heathrow, Los Angeles, Dallas Fort Worth, Frankfurt, Paris Charles-de-Gaulle, Amsterdam. Small airports: Denver, McCarran, Madrid Barajas, Newark, Kennedy, Bangkok, Miami, Hong Kong, New Tokyo, Pearson, Sydney, Changi.

on




al

co

py

The stochastic frontier analysis (SFA) was used to analyze 34 airports in Europe, and a high correlation between the size of an airport and its return to scale was found (Pels et al., 2003). In addition, both the ownership and the size of the airport were found to be important factors in determining the operational performance of airports. However, the ownership of a given airport has less impact on performance than does the size of airport (Humphreys et al., 2002). The size of an airport can be measured in units of traffic transported (UT) (MartinCejas, 2002), whereby UT is the number of passengers plus (kilograms of freight/1000). Using these units, airports can be categorized as small airport if UTp4  107, and large if UT44  107. Applying this criterion, the airports are categorized as follows:







Private ownership: London Heathrow, Bangkok, Pearson, Sydney. Mixed public and private ownership: Frankfurt, Denver. Public ownership: Atlanta Hartsfield, Paris Charlesde-Gaulle, Madrid Barajas, Hong Kong, New Tokyo, Changi, Chicago O’Hare, Los Angeles, Dallas Fort Worth, Amsterdam, McCarran, Kennedy, Miami, Newark.

Au





th o

r's

pe

rs

(London Heathrow, Frankfurt, Paris Charles-de-Gaulle, Amsterdam, and Madrid Barajas), one airport in Australia (Sydney), and four airports in Asia (Bangkok, Hong Kong, New Tokyo, and Changi). These airports were assessed in the current study against five factors: ownership of airport, size of airport, hub airport, location of airport, and economic growth rate of country. Each of these is discussed in greater detail below. Since the 1980s, the deregulation of aviation industry and the privatization of airports has been accelerated. These trends have intensified the need for performance monitoring from the perspective of both investors and governments. Moreover, it is important to establish the relationship between privatization and management strategy for performance improvement (Gillen and Waters, 1997). In general, the reasons for governments choosing to promote airport privatization include a desire to: remove airports from the public sector; increase capital investment in existing airports; reduce the number of civil servants; increase tax revenue; protect airport administration from political interference; impose commercial disciplines on airport management; and stimulate private investment to build new airports. According to Inamete (1993), the alternatives to privatization of airports include: the management of government-owned airports by private firms; mixed ownership of an airport by government and private firms; a government-holding corporation; and increased autonomy for government-owned airports. Ashford (1994) also suggested: selling whole or part of an airport; setting up a private company owned by local or central government authorities; and establishing long-term leases, or build–operate–transfer (BOT) arrangements. In summary, three kinds of ownership of airports have been pointed out in recent years: private ownership, mixed public and private ownership, and public ownership with direct government control (Humphreys, 1999). This study therefore categorized the airports into three groups (e.g. following, Morrel and Lu, 2000; Thomson, 2002; Forsyth, 2003):

343

The following hypotheses are proposed with respect to the first factor: H10. There is no significant difference in operational performance between private airports and public airports. H20. There is no significant difference in operational performance between private airports and mixed ownership airports.

The following null hypothesis is proposed with respect to the second factor: H40. There is no significant difference in operational performance between large airports sand small airports.

A hub–spoke network model enables airports to centralize unpopular routes and increase the flow of backbone routes to accomplish economies of scale and reduce the average transportation cost. Such an approach can offer customers lower prices and multiple services. The characteristics of a hub airport include: more frequent flights; more direct flights; more opportunities for sameday return flights; more international flights; and service geared to local market needs (Button et al., 1999). Sarkis (2000) pointed out that the hub airport was more efficient than a non-hub airport, and Gerber (2002) identified the hub–spoke system and airline alliance as two major factors to be considered in the modern competitive aviation industry. To make the comparison, four categories of airports are identified on the basis of traffic intensity and characteristics of country as shown in Fig. 1 (Ashford, 1994):







Category 1: airports with higher transfer/transit traffic intensity. Category 2: airports with lower transfer/transit traffic intensity. Category A: airports in the developed countries. Category B: airports in the developing countries.

ARTICLE IN PRESS L.C. Lin, C.H. Hong / Journal of Air Transport Management 12 (2006) 342–351

344

There is a close relationship between freight transportation and economic growth (Zhang, 2003; Wang, 1995). This relationship was confirmed by the Boeing World Air Cargo Forecast of 2002, that forecast significant future growth in the aviation industry as a result of improving economic activity (Boeing, 2002). The average growth rate in all national economies in 2002 was 1.9% according to International Monetary Fund (2003). The present study therefore categorized the airports as being situated in a high-growth economy if the country’s economic growth exceeded 1.9%, and as being in a low-growth economy if the nation’s economic growth was less than 1.9%. This led to categorizations:

Category 2

py

Non-hub airport

Category 1

co

Category B

H11 0 . Airports in Australia and Asia display no significant difference in operational performance.

Hub airport

Category A

Traffic intensity

Fig. 1. Classification of airports.

The airports satisfying Category 1 and Category A would be grouped as ‘hub airports’, while those satisfying Category 2 and Category B would be grouped as ‘non-hub airports’ as shown in Fig. 1. The airports under consideration were categorized as follows:

on




Hub airport: Atlanta Hartsfield, Chicago O’Hare, London Heathrow, Los Angeles, Dallas Fort Worth, Frankfurt, Paris Charles-de-Gaulle, Amsterdam. Non-hub airport: Bangkok. The null hypothesis used to examine the third factor is:

pe

H50.

There is no significant difference in operational performance between hub airports and non-hub airports.





North America: Atlanta Hartsfield, Chicago O’Hare, Los Angeles, Dallas Fort Worth, Denver, McCarran, Kennedy, Miami, Newark, Pearson; Europe: London Heathrow, Frankfurt, Paris Charlesde-Gaulle, Amsterdam, Madrid Barajas; Australia: Sydney; Asia: Bangkok, Hong Kong, New Tokyo, Changi.

th o




r's

In a study of the operational performance of 44 major US airports, Sarkis (2000) demonstrated that the location of an airport had an influence on its efficiency. Here, airports are categorized into four groups:




High economic growth rate: Atlanta Hartsfield, Chicago O’Hare, Los Angeles, Dallas Fort Worth, Denver, McCarran, Kennedy, Miami, Newark, Pearson, Bangkok, Hong Kong, Changi; Low economic growth rate: London Heathrow, Frankfurt, Paris Charles-de-Gaulle, Amsterdam, Madrid Barajas, Sydney, New Tokyo.




rs







al

Developed/developing countries

H10 0 . Airports in Europe and Asia display no significant difference in operational performance.

Au

The following null hypotheses is explored with respect to the fourth factor: H60. Airports in North America and Europe display no significant difference in operational performance. H70. Airports in North America and Australia display no significant difference in operational performance. H80. Airports in North America and Asia display no significant difference in operational performance. H90. Airports in Europe and Australia display no significant difference in operational performance.

The following null hypothesis is proposed with respect to the fifth factor:

H12 0 . There is no significant difference in operational performance between airports in the countries with high economic growth and airports in the countries with low economic growth. 3. Data envelopment analysis The DEA model was formulated in the 1970s (Charnes et al., 1978) and based on Farrell’s (1957) non-parametric production frontier function. The DEA model makes use of mathematical programming based on the multiple inputs and multiple outputs to estimate the relative efficiency of multiple DMUs. The ‘relative efficiency’ classifies a DMU as being an efficient DMU or an inefficient DMU. An efficient DMU is one that has the most appropriate combinations of input and output variables, which constitutes the efficiency frontier. The relative position of a DMU with respect to this efficiency frontier is used as a measure of the extent of efficiency of an inefficient DMU. As the constituents of DMUs change, the extent of relative efficiency can change accordingly. Because DEA does not require predetermined functional forms, human subjective judgments can be avoided, and deficiencies in traditional analytical methods and regression analyses can be improved. However, the number of DMUs has to be at least twice the sum of the input and output variables; otherwise, DEA is unable to distinguish an efficient unit from an

ARTICLE IN PRESS L.C. Lin, C.H. Hong / Journal of Air Transport Management 12 (2006) 342–351

effects on inefficient DMUs, but in which the efficiency values of efficient DMUs would be re-calculated to be greater than 1. This allows efficient DMUs to be ranked. Deprins et al. (1984) first proposed the free disposable hull (FDH) model. Tulkens (1993) considered that the efficiency of a DMU should be determined by real observational performance of DMUs rather than by the DMUs of a reference set. In effect, Tulkens considered that the production frontier of the FDH model would have the shape of a stepladder, rather than a smooth curve. However, this stepladder shape for a production frontier made most of the DMUs efficient, but difficult to distinguish from good to bad. The operational performances of airports in the present study were thus evaluated on the basis of four DEA analyses described below. First, according to the efficiency value analysis, the DMUs are classified into efficient unit and inefficient unit. The BCC model makes use of index u0 to determine the status of return to scale. Secondly, the inefficient DMUs can improve the utilization of resources by slack variable analysis. Thirdly, with an increase or decrease in the number of DMUs, or removal of the input and output variables one by one, the decision-makers can observe variations of efficiency values in the sensitivity analysis. Fourthly, in the analysis of variable weight, ur represents contribution of each output unit to the relative efficiency value, whereas vi represents the contribution to the relative efficiency for every input unit. The procedure adopted here is summarized in Fig. 2. First, the DMUs were selected. Input and output variables for operational performance of airports were then selected. The five models were then applied on the basis of the data collected. The four analyses were then conducted to provide further insight into airport efficiency. The 12 hypotheses proposed above were then tested. Finally, the efficiency grouping of airports was presented as an aid for decision-making.

co

py

inefficient unit. The present study adopted five DEA models. Each of these is described below. The CCR model was developed to measure production efficiency under constant return to scale (CRS) conditions. The model presumes that DMUk (k ¼ 1; . . . ; n) uses input X ik (i ¼ 1; . . . :; m), and produces output Yrk (r ¼ 1; . . . ; s). The relative efficiency value of DMUk can then be obtained as follows: Maximize Ps ur Y rk hk ¼ Pr¼1 m i¼1 vi X ik

345

subject to: Ps ur Y rj Pr¼1 p1; j ¼ 1; . . . ; n, m i¼1 vi X ij

th o

r's

on

pe

rs

where ur and vi are the weights (suppositional multipliers) of output r and input i, respectively; e represents the extremely small positive number (set as 106) to make all ur, vi positive. The BCC model is named after (Banker et al, 1984), who loosened the restrictions of the CCR model to variable return to scale (VRS), and a new restriction of u0, which was equivalent to an intercept, was introduced. Furthermore, the aggregate efficiency was broken down into pure technical efficiency and scale efficiency. In other words, a lack of efficiency could be due to inappropriate policies having been adopted (that is, technical inefficiency) or an inappropriate production scale having been implemented (that is, scale inefficiency). The simple cross-efficiency model was developed as an extension of the efficiency values obtained from the CCR and BCC models, which are referred to as ‘self-appraisal’. If the concept of ‘peer-appraisal’ is introduced, a crossefficiency matrix is effectively added to the DEA model (Sexton et al., 1986). Ekj represents the optimum multiplier of DMUk used to measure the efficiency value of DMUj, and ej is the average cross-efficiency value by peerappraisal as shown below: Pn k¼1;jak E kj . ej ¼ ðn  1Þ

al

ur ; vi X
X0; r ¼ 1; . . . ; s; i ¼ 1; . . . ; m,

Mk ¼

Au

In the CCR model of self-appraisal, if DMUk is not referred often, there is a greater chance that DMUk is a group outlier. Doyle and Green (1994) therefore established an index for judging group outliers, as shown below: E kk  ek , ek

where Ekk is the CCR model efficiency value, ek is the average cross-efficiency value, k ¼ 1 . . . n. The A&P model is named after Andersen and Petersen (1993). For efficient DMUs, the efficiency values are all counted as 1 in the CCR and BCC models. Andersen and Petersen proposed a new method, which did not have any

4. Result analysis 4.1. Selection of input and output variables Previous analyses suggest a range of input and output that may be relevant.1 From these, a preliminary selection of eight input variables and three output variables was made as shown in Table 1. In choosing these variables, the following three factors were taken into consideration:




the DEA model is more suited to quantitative data than qualitative data;

1 Gillen and Lall (1997); Gillen and Waters (1997); David (1999); Sarkis (2000); Martin and Roman (2001); Pels et al. (2001); Yu (2001); Fernandes and Pacheco (2002); Fernandes and Pacheco (2003); Massoud and Bijan (2003); Pels et al. (2003); Lin and Chen (2004); Sarkis and Talluri (2004); and Yu (2004).

ARTICLE IN PRESS L.C. Lin, C.H. Hong / Journal of Air Transport Management 12 (2006) 342–351

346

Selection of DMUs

Selection of input and output variables

FDH model

A & P model

SCE model

BCC model

CCR model

py

Supplement model

co

Basic model

Ownership of airport

Efficiency value analysis

al

DEA analysis Size of airport

Slack variable analysis

on

Hypothesis tests of airport features Hub airport

Efficiency grouping of airports

Analysis of variable weight

rs

Location of airport

Sensitivity analysis

Economic growth rate

Conclusion and suggestion

Table 1 Pearson coefficients of input and output variables Number of employees

Number of check in counters

0.635*

0.047

Number of passengers Cargo Movements

0.536* 0.293

th o

Output

0.465* 0.024

Number of runways

0.453*

0.784*

0.291 0.776*

0.185 0.839*




Number of baggage collection belts

Number of aprons

Number of boarding gages

Terminal area

0.292

0.406

0.300

0.159

0.033 0.456*

0.040 0.537*

0.115 0.421

0.592* 0.006

**

It is relatively significant when significance level is set to be 0.01.

Au

*It is relatively significant when significance level is set to be 0.05.




Number of parking spaces

r's

Input

pe

Fig. 2. Research procedure.

the number of DMUs must be at least twice the sum of input and output variables; and the selected variables must be comprehensive enough to provide an effective evaluation of the operational performance of international airports.

The data source for the input and output variables was Airport Capacity/Demand Profiles (Airport Council International, 2003). Because DEA is incapable of distinguishing the significance level of input and output variables in advance, all variables have an equal chance of affecting efficiency value.

In addition, the selection of variables must be isotonic, i.e., the value of output variables cannot decrease when input variables are increased. This was tested with Pearson analysis for input and output variables. As seen in Table 1, cargo has relatively small coefficients with respect to input variables, and some coefficients were negative. The variable was therefore dispensed with while retaining the other two output variables (number of passengers and movements). With respect to the input variables, the number of check-in counters, boarding gates, and terminal area had relatively small coefficients. Therefore, this research dispensed with these three input variables while retaining the other five

ARTICLE IN PRESS L.C. Lin, C.H. Hong / Journal of Air Transport Management 12 (2006) 342–351

(number of employees, number of runways, number of parking spaces, number of baggage collection belts, and number of aprons).

r's

co

al

on

pe

rs

Comparisons of efficiency values, ranking, and outlier values for the two basic models and three supplementary models are in Table 2. The indications are that nine airports had an efficiency value of 1 in both the CCR and BCC models, nine airports had scale inefficiency, and two airports had technical inefficiency. The airports with markedly higher efficiency values in every model were Atlanta Hartsfield and McCarran, whereas those with markedly lower efficiency values were Kennedy, New Tokyo and Pearson. However, under the FDH model, every airport was found to be efficient, with no significant difference being apparent among the DMUs. The FDH model thus can not distinguish among relatively efficient DMUs. An analysis of the CCR efficiency value shows that most airports had an outlier value between 1 and 10 except Amsterdam. Compared with Doyle and Green (1994), who proposed an outlier value of 23, the outlier values here are not high, and it was appropriate to carry out relative efficiency comparisons among the 20 airports. The DEA model also provides an index of return to scale for the various airports. The nine airports with an efficiency value of 1 in both CCR and BCC models were in an appropriate production scale with a CRS, u0 ¼ 0. In contrast, Frankfurt, Denver, Madrid Barajas, Newark, Kennedy, Bangkok, New Tokyo, and Pearson airports all have u0 values of less than 0, that signifies increasing return to scale and a stage of growth in which production scale is slightly

less than optimum. Chicago O’Hare, Los Angeles, and Dallas Fort Worth airports all had u0 values of greater than 0, that represents a diminishing return to scale. For the slack variable analysis, the BCC model is mainly applicable to an analysis of technical efficiency, and its slack variable analysis could be used to assess short-term improvement. The CCR model includes aggregate efficiency and technical efficiency, and its slack variable analysis can be used to assess long-term improvement (Lin and Chen, 2004). For the short-term improvement, there are 14 airports at the efficiency frontier with input and output slack variables of 0. Among the inefficient airports, Los Angeles airport has the greatest excess in the input variable ‘number of employees’. Frankfurt airport had the greatest excess in the input variables ‘number of parking spaces’ and ‘number of baggage collection belts’. Kennedy airport had the greatest excess in ‘number of aprons’. For the long-term improvement, nine airports are shown to have input and output slack variables of zero. Among the inefficient airports, Newark requires the greatest reduction 43.2% in ‘number of baggage collection belts’, whereas Dallas had the lowest excess 3.75% in ‘number of aprons’. With respect to the output slack variables, Chicago, Dallas, Denver, Newark, and Pearson need to increase the ‘number of passengers’. Kennedy, Bangkok and New Tokyo need to increase output standards for ‘movements’. In particular, New Tokyo needs an increase of 55% in ‘movements’. Overall, there were 17 airports with slack variables of 0 in terms of ‘number of employees’ and ‘number of runways’. This shows that the utilization of employees and runways is satisfactory in short-term and long-term operations. However, in general, utilization is poor for ‘number of baggage collection belts’ and ‘number of aprons’. In terms

py

4.2. DEA analyses

347

Table 2 Efficiency values, ranking and outlier values of different DEA models CCR model 1 (1) 0.916 1 (1) 0.976 0.932 0.825 1 (1) 1 (1) 0.746 1 (1) 0.844 0.797 0.616 0.947 1 (1) 1 (1) 0.772 0.701 1 (1) 1 (1)

Au

Atlanta Hartsfield Chicago O’Hare London Heathrow Los Angeles Dallas fort worth Frankfurt Paris Charles-de-Gaulle Amsterdam Denver McCarran Madrid Barajas Newark Kennedy Bangkok Miami Hong Kong New Tokyo Pearson Sydney (Kingsford Smith) Changi

BCC model

th o

Airport

(13) (10) (12) (15)

(18) (14) (16) (20) (11)

(17) (19)

1 (1) 1 (1) 1 (1) 0.993 1 (1) 0.911 1 (1) 1 (1) 0.850 1 (1) 0.920 1 (1) 0.665 1 (1) 1 (1) 1 (1) 1 (1) 0.891 1 (1) 1 (1)

(15) (17)

(19) (16) (20)

(18)

Scale efficiency

Simple cross efficiency model

A&P model

FDH model

Outlier value

1 0.916 1 0.983 0.932 0.905 1 1 0.878 1 0.917 0.797 0.925 0.947 1 1 0.772 0.787 1 1

0.943 0.737 0.827 0.754 0.701 0.552 0.701 0.528 0.574 0.889 0.594 0.557 0.409 0.489 0.633 0.805 0.517 0.535 0.757 0.769

1.923 0.916 1.174 0.976 0.932 0.825 1.378 1.095 0.746 1.805 0.844 0.797 0.616 0.947 1.129 1.327 0.772 0.701 1.656 1.793

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

3.147 9.635 9.589 3.869 8.084 4.708 7.514 20.098 8.467 7.509 4.490 5.139 1.191 7.612 7.092 8.088 6.022 10.966 8.993 8.593

(1) (8) (3) (7) (10) (15) (9) (17) (13) (2) (12) (14) (20) (19) (11) (4) (18) (16) (6) (5)

(1) (13) (7) (10) (12) (15) (5) (9) (18) (2) (14) (16) (20) (11) (8) (6) (17) (19) (4) (3)

ARTICLE IN PRESS L.C. Lin, C.H. Hong / Journal of Air Transport Management 12 (2006) 342–351

the number of flights in operation. This is consistent with the international trend of placing greater emphasis on passenger transport in the aviation industry.

4.3. Testing hypotheses

al

co

py

As seen in Table 3, first the various DEA models show significance levels greater than 0.05 for H10 and H20. For H30, two models show significance levels greater than 0.05, and two show significance levels less than 0.05. These results suggest that H10, H20 and H30 cannot be rejected, i.e., no significant differences among private airports, public airports, and mixed ownership airports as shown in Fig. 3. The aviation industry has historically been owned and operated by governments or local authorities. Although the modern trend is towards the privatization of airports, governments still remain protective, and some private airports might lack the required experience or expertise to be efficient. This suggests that ‘privatization’ might not be the appropriate solution in all cases. Secondly, large airports and small airports display no significant differences in operational performance, with the various DEA models showing significance levels greater than 0.05. Therefore, H40 cannot be rejected. Rather than relying on an expansion of airport facilities, outstanding performance can be obtained through rapid customs procedures, reasonable fees, fewer delayed flights, convenient transportation, and a robust domestic economy.

pe

rs

of short-term and long-term improvements, Atlanta, London, Paris, Amsterdam, McCarran, Miami, Hong Kong, Sydney, and Changi perform well with both input and output slack variables of zero. For the sensitivity analysis, to take Chicago airport as an example, the removal of ‘number of runways’ would drop the efficiency value from 0.92 to 0.85, and the removal of ‘movements’ would drop the efficiency value from 0.92 to 0.81. This shows that ‘number of runways’ and ‘movements’ were the most sensitive variables for the operational performance of this airport. The efficiency values of Atlanta and McCarran airports are maintained at 1 throughout implying that these two airports were quite efficient in the utilization of facilities. Overall, the ‘number of runways’ and ‘number of parking spaces’ are more sensitive than other input in terms of operational performance for most airports. With regard to output variables, ‘number of passengers’ and ‘movements’ are the sensitive variables in most cases. Analyzing of variable weight, among five input variables, the ‘number of baggage collection belts’ and ‘number of aprons’ has a higher frequency of zero weight, which implies that the other three input variables contributed more to the operational performance of the airports. Among the three input variables, the ‘number of runways’ has more weight than the others. With regard to output variables, the weight of ‘number of passengers’ was greater than ‘movements’, which indicates that the number of aviation passengers transported was more important than

on

348

Table 3 Summary of hypotheses testing P-value of CCR model

P-value of BCC model

P-value of SCE model

P-value of A&P model

Testing results

H10 H20 H30 H40 H50 H60 H70 H80 H90 H10 0 H11 0 H12 0

0.916 0.348 0.164 0.2 0.014 0.282 0.042 0.014 0.042 0.032 0.351 0.009

0.246 0.14 0.016 0.586 0.044 0.55 0.039 0.019 0.042 0.019 1 0.013

0.92 1 0.032 0.111 0.037 0.624 0.014 0.082 0.032 0.089 0.77 0.027

0.841 0.355 0.101 0.277 0.044 0.27 0.032 0.032 0.142 0.271 0.38 0.012

Accepted Accepted Accepted Accepted Rejected Accepted Rejected Rejected Rejected Rejected Accepted Rejected

Au

th o

r's

Hypothesis

The italic values mean that they are significant at 0.05 level. *The significance level is set at 0.05.

Not significantly different

Private airports

Mixed ownership airports

Not significantly different

Public airports

Not significantly different

Fig. 3. Hypotheses testing for ownership of airport.

ARTICLE IN PRESS L.C. Lin, C.H. Hong / Journal of Air Transport Management 12 (2006) 342–351

on

al

co

py

benchmarking purpose. The two sets of efficiency values are positively related with a Pearson coefficient of 0.51. The chart assists in assessment of the efficiency groupings of international airports and the implementation of appropriate actions for improving efficiency. The BCC efficiency value was classified into three groups on the y-axis. The ‘less favorable self-appraisal’ group includes the inefficient units. Efficient units are then distinguished on the basis of the frequency of reference by other airports in the efficiency analysis. The ‘favorable self-appraisal’ group contains those referred by others 0–2 times. The ‘more favorable self-appraisal’ group contains those referred by others more than twice. The simple crossefficiency value was also classified into three groups at 0.6 and 0.8 on the x-axis into a ‘less favorable peer-appraisal’ group, a ‘favorable peer-appraisal’ group, and ‘more favorable peer-appraisal’ group. The 20 airports are classified into four groups (Fig. 5). The A group (top right of Fig. 5) is constituted by airports that can act as benchmarks. These airports possess advantages in a range of indices and include Atlanta Heartsfield, McCarran, Hong Kong, and London Heathrow airports. The B group (upper center) includes airports that are outstanding in certain indices. Airports belonging to this group include Changi, Sydney, Chicago O’Hare, Paris Charles-de-Gaulle, Dallas Fort Worth, Miami, and Los Angeles. Ongoing improvement in areas of weakness should ensure that these airports reach benchmarking status in the future. The C group (top left) is constituted by airports that require significant effort to improve their operational performance. This group includes Newark, Amsterdam, New Tokyo, and Bangkok. The D group (bottom left) includes airports that have no advantages in various evaluation indices and poor results in terms of peer appraisal by other airports. This group includes Madrid Barajas, Denver, Frankfurt, Pearson, and Kennedy. These airports need to investigate their weaknesses, and more efforts should be placed on those factors with greater weights if operational efficiency is to be rapidly improved.

pe

rs

Thirdly, hub airports and non-hub airports displayed significant differences in operational performance, with a significance level less than 0.05 being found between the various DEA. H50 is therefore rejected. Cargos and passengers from the spokes can be centralized in a hub airport, enabling the airport to reach an economic scale and reduce average unit cost. Hub airports also enable increased frequency of flights, more direct flights, and a greater chance of round-trip flights on the same day (Button et al., 1999). The better operational performance can thus be achieved from hub airports. 7 8 9 Fourthly, H60 and H11 0 cannot be rejected, but H0, H0, H0 10 and H0 would be rejected. The results are summarized in Fig. 4. It can be concluded that the airports in North America and Europe had better operational performances than the airports in Asia and Australia. The explanation may be that North America and Europe have higher GDP, that more people travel by air in these regions, that more intense domestic and international trading activities exist, and that more advanced airport facilities are built to provide better services. Fifthly, countries with higher economic growth rate showed a significant difference in airport operational performance from that of countries with lower economic growth rate. The significance level was less than 0.05 among various DEA models. H12 0 is thus rejected. In other words, the operational performance efficiency of an airport is related to the national economic growth rate of the country in which the airport is located. Greater business trading activities improve airport efficiencies. Although management strives to improve the utilization of airport facilities, a general strengthening of international business trading activities is also required to increase overall operational performance.

r's

4.4. Efficiency grouping of airports

Au

th o

On the basis of self-appraisal from BCC efficiency values and peer-appraisal from simple cross-efficiency values, a distribution chart of airports can be established for

349

America Significantly different

Not significantly different Significantly different

Asia

Europe Significantly different

Not significantly different

Significantly different

Australia

Fig. 4. Hypotheses testing for location of airport.

ARTICLE IN PRESS More favorable peer-appraisal 10

4

1

3

17 8

15

Less favorable self-appraisal

6

0

3

5 7 2 19 4

18

1

20

co

14

py

16

12

2

A

B

Favorable self-appraisal

C

11 9

0.7

al 0.5

0.6

on

13

0.4

0.9 0.8

D

0.3

1.0

0.7

0.8

BCC efficiency value

Favorable peer-appraisal

More favorable self-appraisal

Less favorable peer-appraisal

Frequency of referred by others

L.C. Lin, C.H. Hong / Journal of Air Transport Management 12 (2006) 342–351

350

0.6 0.5

0.9

1.0

rs

Simple cross efficiency value

pe

Fig. 5. Chart of efficiency grouping for airports. Note: 1. Atlanta Hartsfield, 2. Chicago O’Hare, 3. London Heathrow, 4. Los Angeles, 5. Dallas Fort Worth, 6. Frankfurt, 7. Paris Charles-de-Gaulle 8. Amsterdam, 9. Denver, 10. McCarran, 11. Madrid Barajas, 12. Newark, 13. Kennedy, 14. Bangkok, 15. Miami, 16. Hong Kong, 17. New Tokyo, 18. Pearson, 19. Sydney, 20. Changi.

5. Conclusion

Au

th o

r's

Evaluation of the operational performance of airports is important for ensuring efficiency in the wider aviation industry. By adopting various DEA models, international airports can accurately assess aspects of their own performance that require improvement, and can gain an enhanced understanding of the current status and future developments in other countries. Here DEA analysis has provided several useful insights. First, efficiency value analysis has established that while some major airports are of optimal scale, others are not. Secondly, slack variable analysis is seen as potentially providing an understanding of how airports can improve their operational performance by enabling managers can focus on a limited number of variables for short-term and long-term improvement. Thirdly, sensitivity analysis shows that the ‘number of runways’ and the ‘number of parking spaces’ are two input variables that have higher sensitivity with respect to airport efficiency. Fourthly, an analysis of variable weight shows that the number of passengers contributes more in terms of airport operational performance than movements. The study also finds that the operational performance of airports is related to the existence of a hub airport and economic growth in the country in which it is located, implies that, before efficiency can be increased, a higher

frequency of flights and enhanced international business trade activities is needed. With respect to location, airports in North America and Europe have higher operational efficiencies than airports in Asia and Australia. The form of ownership of an airport and its size are not significantly related to operational performance, whereas the existence of a hub airport and the location of the airport are significantly related to operational performance.

References Airport Council International, 2003. Airport Capacity/Demand Profiles. /http://www.airports.org/S. Andersen, P., Petersen, N.C., 1993. A procedure for ranking efficient units in data envelopment analysis. Management Science 39, 1261–1264. Ashford, N., 1994. Airport management in a changing economic climate. Transportation Planning and Technology 18, 57–63. Banker, R.D., Charnes, A., Cooper, W.W., 1984. Some model for estimating technical and scale inefficiencies in data envelopment analysis. Management Science 30, 1078–1092. Boeing, 2002. World air Cargo Forecast 2002/2003. /http://www.boeing. com/commercial/cargo/pdf/WACFbrochure02-03.pdfS. Button, K., Lall, S., Stough, R., Trice, M., 1999. High-technology employment and hub airports. Journal of Air Transport Management 5, 53–59. Charnes, A., Cooper, W.W., Rhodes, E., 1978. Measuring the efficiency of decision making unit. European Journal of Operational Research 2, 429–444.

ARTICLE IN PRESS L.C. Lin, C.H. Hong / Journal of Air Transport Management 12 (2006) 342–351

on

al

co

py

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 7, 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 8, 233–238. Massoud, B., Bijan, V., 2003. Size versus efficiency: a case study of US commercial airports. Journal of Air Transport Management 9, 187–193. Morrel, P., Lu, C.H.-Y., 2000. Aircraft noise social cost and charge mechanisms—a case study of Amsterdam airport Schiphol. Transportation Research Part D: Transport and Environment 5, 305–320. Oum, T.H., Yu, C., Fu, X., 2003. A comparative analysis of productivity performance of the World’s major airports. Journal of Air Transport Management 9, 285–297. Pels, E., Nijkamp, P., Rietveld, P., 2001. Relative efficiency of European airports. Transport Policy 8, 183–192. Pels, E., Nijkamp, P., Rietveld, P., 2003. Inefficiencies and scale economies of European airport operations. Transport Research Part E: Logistics and Transportation Review 39, 341–361. Sarkis, J., 2000. An analysis of the operational efficiency of major airport in the United States. Journal of Operations Management 18, 335–351. Sarkis, J., Talluri, S., 2004. Performance based clustering for Benchmarking of US airports. Transportation Research Part A: Policy and Practice 38, 329–346. Sexton, T.R., Silkman, R.H., Hogan, A.J., 1986. Data development analysis: critique and extension. In: Silkman, R.H. (Ed.), Measuring Efficiency: An Assessment of Data Development Analysis. JosseyBass, San Francisco. Thomson, I.B., 2002. Air transport liberalization and the development of third level airports in France. Journal of Transport Geography 10, 273–285. Tulkens, H., 1993. On FDH efficiency analysis: some methodological issues and applications to retail banking, courts and urban transit. Journal of Productivity Analysis 4, 183–210. Wang, C.T., 1995. Discussed C.K.S market potential and development tactics by the development trend of freight transportation. Transportation 28, 59–74 (Chinese Edition). Yu, M.M., 2001. A Study on Aircraft Noise Regulation and Input Congestion Impact on the Performance of the Airports in Taiwan. Department of Civil Engineering, National Taiwan University (Chinese Edition). 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. Zhang, A., 2003. Analysis of international air-cargo hub: the case of Hong Kong. Journal of Air Transport Management 9, 123–138.

Au

th o

r's

pe

rs

Chin, A.T.H., Siong, L.E., 2001. Airport performance: a comparative study between Changi airport and airports in the New York-New Jersey Metropolitan Area. Paper Presented at World Conference on Transportation Research, Seoul. David, P., 1999. The performance of BAA before and after privatization. Journal of Transport Economics and Policy 33, 133–146. Deprins, D., Simar, L., Tulkens, H., 1984. Measuring labor-efficiency in post offices. In: Marchand, M., Pestieau, P., Tulkens, H. (Eds.), The Performance of Public Enterprises: Concepts and Measurement. North-Holland, Amsterdam. Doyle, J., Green, R., 1994. Efficiency and cross-efficiency in DEA: derivations, meanings and uses. Journal of the Operational Research Society 45, 567–578. Farrell, M.J., 1957. The measurement of productive efficiency. Journal of the Royal Statistical Society A 120, 253–281. Fernandes, E., Pacheco, R.R., 2002. Efficient use of airport capacity. Transportation Research Part A: Policy and Practice 36, 225–238. Fernandes, E., Pacheco, R.R., 2003. Managerial efficiency of Brazilian airports. Transportation Research Part A: Policy and Practice 37, 667–680. Forsyth, P., 2003. Regulation under stress: development in Australian airport policy. Journal of Transport Management 9, 25–35. Gerber, P., 2002. Success factors for the privatization of airports-an airline perspective. Journal of Transport Management 8, 29–36. Gillen, D., Lall, A., 1997. Developing measures of airport productivity and performance: an application of data envelopment analysis. Transportation Research Part E: Logistics and Transportation Review 33, 261–273. Gillen, D., Waters, W.G., 1997. Airport performance measurement and airport pricing. Transportation Research Part E: Logistics and Transportation Review 33, 245–247. Humphreys, I., 1999. Privatization and commercialization changes in UK airport ownership patterns. Journal of Transport Geography 7, 121–134. Humphreys, I., Francis, G., Fry, J., 2002. The benchmarking of airport performance. Journal of Air Transport Management 8, 239–247. Inamete, U.B., 1993. Key elements in managing airport—the policy environment and increasing efficiency. International Journal of Public Sector Management 6 (5), 12–23. International Civil Aviation Organization (ICAO), 2003. /http://www. icao.int/S. International Monetary Fund, 2003. The Directorate General of Budget, Accounting and Statistics, Executive Yuan. (Chinese Edition). /http:// www.dgbas.gov.tw/mp.asp?mp=1S. Lin, L.C., Chen, Y.J., 2004. Evaluation of operational efficiency of major airports in the Asia region: an application of data envelopment analysis. Journal of Chinese Institute of Transportation 16, 115–144 (Chinese Edition).

351