key: cord-0956981-v2bcmwq8 authors: Kan Tsui, Wai Hong; Balli, Hatice Ozer; Gilbey, Andrew; Gow, Hamish title: Operational efficiency of Asia–Pacific airports date: 2014-08-31 journal: Journal of Air Transport Management DOI: 10.1016/j.jairtraman.2014.05.003 sha: 9a2a7e6f75b78246353836ae231beb349f9ceacc doc_id: 956981 cord_uid: v2bcmwq8 Abstract Airports are important drivers of economic development and thus under tremendous pressure from emerging competitors. However, few studies have analysed the operational efficiency of Asia–Pacific airports. This study therefore evaluated the operational efficiency of 21 Asia–Pacific airports between 2002 and 2011. A two-stage method was used: Data Envelopment Analysis (DEA) to assess airport efficiency, followed by the second-stage regression analysis to identify the key determinants of airport efficiency. The first-stage DEA results indicated that Adelaide, Beijing, Brisbane, Hong Kong, Melbourne, and Shenzhen are the efficient airports. The second-stage regression analysis suggested that percentage of international passengers handled by an airport, airport hinterland population size, dominant airline(s) of an airport when entering global airline strategic alliance, and an increase in GDP per capita are significant in explaining variations in airport efficiency. Several factors have stimulated the growth in air transport demand and airport development, such as rapid economic development, privatisation of the airport industry, and the liberalisation of aviation policy in the AsiaePacific region (e.g. Oum and Yu, 2000; Park, 2003; Williams, 2006; Yang et al., 2008; Zhang, 2003) . The growth is reflected by the increasing air traffic volumes handled by AsiaePacific airports. The Airport Council International (ACI) reported that several major AsiaePacific airports have been frequently ranked inside the world's top 30 busiest airports between . Moreover, ACI also projects that the announced growth rates for air cargo volumes and aircraft movements in the AsiaePacific region will reach 6.3% and 4.5%, respectively, by 2025 (ACI, 2007) . The International Civil Aviation Organisation (ICAO) also estimates that the AsiaePacific region will become the busiest and fastest growing air transportation market for international passenger traffic by 2025 (ICAO, 2008) . Governments in the AsiaePacific region have therefore invested heavily and constructed airport infrastructure and facilities to meet projected future air transport demand (O'Connor, 1995) . However, airports are also under pressure from emerging competitors competing for air traffic demand. To respond to this pressure, airport efficiency has been identified as a critical issue facing airport management (Chin and Siong, 2001; Forsyth, 2003 , Talley, 1983 . To investigate airport efficiency, Data Envelopment Analysis (DEA) has become the recognised method for efficiency evaluation due to its simplicity in constructing an efficiency frontier for identifying efficient or inefficient airports (Gillen and Lall, 1997) . Also, the DEA model requires no assumptions for specifying production functions between airport inputs and outputs. The DEA model can also compute multiple airport inputs and outputs within a single analysis without any difficulties of aggregation, and can assess an airport's relative efficiency in a single period or in a sequence of periods as well as requiring less information for analysis (e.g. Cooper et al., 2006; Pels et al., 2001 Pels et al., , 2003 . Therefore, we first applied the DEA model to assess the operational efficiencies of AsiaePacific airports, and then the SimareWilson bootstrapping regression analysis to identify which factors significantly explain variations in airport efficiency. There are three primary reasons why this study is meaningful: (i) airports operating in the AsiaePacific region seem to be less researched compared with their counterparts in the US, Europe, and South America; (ii) this study contributes to the existing literature by analysing the efficiency of a large group of AsiaePacific airports (21 airports) À the size of sampled airports in this study is a good reflection and representation of the airport industry in the AsiaePacific region due to their roles as the international or regional hub airports in their countries; and (iii) this study extends the work of Ha et al. (2010) , Lam et al. (2009) , and Yang (2010a,b) in assessing the operational efficiency of AsiaePacific airports and seeking to identify the causes of variations in airport efficiency. The format of this study is structured as follows. Section 2 presents the literature review with regard to airport efficiency evaluations. Section 3 outlines the DEA methodology and the SimareWilson bootstrapping regression analysis. Section 4 presents the dataset of sampled airports, and airport input and output variables for the DEA analysis as well as the key determinants for the second-stage regression analysis. Section 5 presents the results and discussion of the first-stage DEA analysis and the second-stage regression analysis. Section 6 concludes what are the key findings of this study. DEA has become a popular method of investigating airport efficiency. Prior DEA studies showed considerable differences in the airport input and output variables used for the efficiency analysis. Three specific forms of DEA analysis were identified from the literature: (i) DEA analysis with operational variables; (ii) DEA analysis with financial variables; and (iii) DEA analysis with second-stage analysis. Airport efficiency studies that have used DEA analysis with operational variables include Fernandes and Pacheco (2002) , Fung et al. (2008) , Ha et al. (2010) , Lam et al. (2009) , Lin and Hong (2006) , Lozano and Gutierrez (2009) , Roghanian and Foroughi (2010) , and Yoshida and Fujimoto (2004) . The reasons why DEA studies employ operational variables for benchmarking airport efficiency but then do not incorporate any financial variables are complicated and an in depth explanation is beyond the scope of the current study. However, one of the reasons may be lack of available financial data related to airport operations or because it is extremely difficult to gather relevant financial data for each airport analysed. Most airports are currently operated as commercial organisations to maximise the profitability from aeronautical and nonaeronautical activities (Graham, 2008) . Therefore the financial variables or indicators have been used in the prior studies as airport input and/or output variables in DEA analyses in order to achieve a fair evaluation of airport efficiency. DEA analysis with financial variables has been applied in such studies such as Barros and Dieke (2007) , Martin and Roman (2001) , Murillo-Melchor (1999) , Pacheco and Fernandes (2003) , Parker (1999) , Sarkis (2000) , Sarkis and Talluri (2004) , and Yang (2010a,b) . One potential problem is that the key determinants causing variations in airport efficiency may not be clearly understood using the operational and/or financial variables in the DEA analysis, although DEA studies of airport efficiency evaluations showed the ability to evaluate airport efficiency (Gillen and Lall, 1997) . A clear understanding of which factors affect airport efficiency would provide insight to airport managers and policy makers for improving airport efficiency through benchmarking; that is, it would help to compare an airport's performance with its peers in the same region and improve its operations. The approach combining a first-stage DEA analysis and a second-stage Tobit model has become a popular method to identify those significant determinants. A number of studies have used this two-stage approach to investigate airports, for example, Abbott and Wu (2002) , Barros and Sampaio (2004) , Gillen and Lall (1997) , Malighetti et al. (2007) , Pathomsiri et al. (2006) , Pels et al. (2001 Pels et al. ( , 2003 , Perelman and Serebrisky (2010) , and Yuen and Zhang (2009) . Although adopting Tobit models in the second-stage analysis has been popular, it is considered as an invalid approach to determine the factors for explaining variations in airport efficiency, due to the presence of inherent dependence among the DEA efficiency indexes from the first-stage DEA analysis (Casu and Molyneux, 2003; Xue and Harker, 1999) . Importantly, one basic assumption of regression analysis is violated e the independence within the sample. To solve this problem, Wilson (2007, 2008) introduced the bootstrapping methodology to solve this problem. Recently, studies have begun to apply the SimareWilson bootstrapping approach for estimating the significant determinants of airport efficiency. For example, Barros and Dieke (2008) used the truncated bootstrapped regression to estimate the efficiency and identify the determinants of 31 Italian airports between 2001 and 2003. They found that the method to bootstrap the DEA efficiency scores with a truncated regression analysis can better explain DEA efficiency levels. Similarly, Barros (2008) employed the truncated bootstrapped regression analysis to analyse the efficiency of Argentinian airports during the period of intense economic crisis. Curi et al. (2011) also used the bootstrapping methodology to investigate 18 Italian airports. During the same year, Tsekeris (2011) used the truncated bootstrapped regression to assess the relative technical efficiency of Greek airports and investigate factors that determine airport efficiency. Merkert and Mangia (2012) also applied the bootstrapping two-stage DEA model to analyse 46 Norwegian airports' efficiency. employed the input-oriented DEA model and the SimareWilson bootstrapping approach to analyse the efficiency of regional airports worldwide, and suggested that the more sophisticated two-stage model can deliver powerful insights into the performance of regional airports. Tsui et al. (2014b) also utilised the slack-based measure (SBM) model, the Malmquist Productivity Index (MPI), and the SimareWilson bootstrapping methods to investigate the efficiency and productivity changes of 11 New Zealand airports for the period of 2010e2012. The DEA methodology evaluates the relative efficiency of a decision making unit (DMU) by building a ratio which consists of the maximum weighted outputs to maximum weighted inputs for each DMU subject to a set of conditions (Charnes et al., 1978) . Considering a group of airports, where y rk and x ik are the known airport outputs and inputs of airport k. The DEA efficiency index of an airport is denoted as B o , which represents the inputs x io ði ¼ 1; 2; 3; …; nÞ that produce the outputs y ro ðr ¼ 1; 2; 3; …; mÞ; u r and v i are the weights of aggregation (virtual multipliers), that are non-negative which are chosen to maximise the value of B o . Thus, the fractional programming model is written as shown in Eq. (1): Later, Banker et al. (1984) developed the DEA-BCC model, which allows airports operating with lower airport inputs to have an increasing return to scale under the principle of Variable Return to Scale (VRS), and those operating with higher airport inputs to have a decreasing return to scale. The DEA-BCC model is written as shown in Eq. (2): where q ¼ airport efficiency index; 3 ¼ a constant (greater than 0) 4 ; s þ ro and s À io ¼ airport output and input slacks; l k ¼ the dual variable or the scalar vector associated with each airport. An airport is considered as a BCC-efficient airport when q is equivalent to 1 and has zero output and input slacks (s þ ro ¼ 0; s À io ¼ 0Þ. Otherwise, the airport is called a BCC-inefficient airport (Cooper et al., 2006) . The DEA efficiency indexes obtained from the first-stage DEA analysis will be used to regress on the factors (e.g. the specific operating characteristics, management/ownerships, and regional locations) related to the sampled AsiaePacific airports and identify the significant factors to explain variations in airport efficiency using the second-stage SimareWilson bootstrapping regression analysis 5 (see Simar and Wilson, 2007) . The initial estimation specification can be written as shown in Eq. (3): Eq. (3) is the first-order approximation of the unknown true relationship. Where q k is the DEA efficiency index of airport k. a is the constant, z k is a vector of observation-specific variables that is expected to associate with airport k's DEA efficiency index, b is a vector of parameters, and 3 k is the error term. Applying the SimareWilson bootstrapping approach, the distribution of 3 k is limited to the condition 3 k ! 1 À a À z k b: Thus, the distribution of 3 k becomes 3 ke iidNð0; s 2 3 Þ. Moreover, the true and unobserved dependent variable q k in Eq. (3) to be replaced by q * k (the DEA efficiency index of airport k after applying the SimareWilson bootstrapping approach), and the model specification can be written as shown in Eq. (4): A rigid DEA convention was followed to determine the total number of airport observations in association with the total number of airport input and output variables; the minimum number of airports observed should be greater than or equal to three times the sum of airport input and output variables to ensure that satisfactory discriminating power is possible (Banker et al., 1989; Cooper et al., 2006; Raab and Lichty, 2002) . The current study achieved this requirement with a sample size of 21 AsiaePacific airports, and a total of seven airport input and output variables for the first-stage DEA analysis. Table 1 shows the list of 21 major AsiaePacific airports for analysis between 2002 and 2011 (Data beyond 2011 was not yet available at the time of submitting this article). The data was collected from the following sources: International Civil Aviation Organisation (ICAO), Airport Council International (ACI), Air Transport Research Society (ATRS)-Airport Benchmarking Reports, civil aviation authority of the respective countries, airports' annual reports and websites. Individual airports were also contacted to obtain additional information. To select airport input and output variables for the first-stage analysis, we considered data availability, referred to extant Remarks: The classification of an airport's status is based on the airport's strategic role and flight connectivity network. For example, an international hub airport connects to at least 25 international destinations; a regional hub or non-hub airport flies to no more than 25 international destinations (Matthiessen, 2004) . literature (e.g. Doganis, 1992) , and sought the professional opinion from airport managers. At a result, we selected four airport input variables (i.e. number of employees, number of runaways, total runway length, and passenger terminal area) and three airport output variables (i.e. air passenger numbers, air cargo volumes, and aircraft movements) for the first-stage DEA analysis. Three tasks were performed in this study to identify the key determinants in explaining variations in airport efficiency. First, the airport input and output variables used in the first-stage DEA analysis will not be reused as the explanatory variables in the second-stage regression analysis, avoiding the problem of double-counting and possibly obtaining misleading or biased results (Lin, 2008) . Second, prior studies relating to airport efficiency were examined to identify the potential explanatory variables for the second-stage regression analysis. Lastly, an attempt was made to look at other principles applying the two-stage regression analysis that may assist in developing other relevant explanatory variables for this study (e.g. Boame, 2004; Fethi et al., 2000; Oum and Yu, 1994; Zheng et al., 1998) . Taking the literature and data availability into account, seven explanatory variables were developed for the second-stage regression analysis, which represents AsiaePacific airports' operating characteristics, management/ownerships, and regional locations (see Table 3 ). Data related to the selected explanatory variables was obtained from National Yearbooks, National Statistical Departments, World Bank Data, United Nation Data, and airports' annual reports and websites. The DEA Output-Oriented 6 and VRS framework was selected for the first-stage DEA analysis. Table 2 shows the DEA estimation results categorising in three groups of airports with reference to changes in airport efficiency including the DEA efficiency indexes for each airport over the years and the percentage of efficient airports during each study year. Table 2 shows that at least 52% of AsiaePacific airports are considered as 'efficient' between 2002 and 2011. Six airports were found to be best performers over the entire study periods having consistently full DEA efficiency indexes (i.e. Adelaide, Beijing, Brisbane, Hong Kong, Melbourne, and Shenzhen). Of these, three were international hub airports (i.e. Beijing, Hong Kong, and Melbourne). This might be consistent with the concept that the international hub or gateway airports are able to attract and handle more air transport demand than the regional or non-hub airports, leading to higher efficiency. Also, their strategic roles and extensive flight connectivity networks reflect their ability to attract more international and domestic passenger traffic (i.e. originedestination traffic and connecting traffic). The full efficiency of Beijing and Hong Kong for all ten years may be explained by their respective air traffic volumes being consistently ranked inside the world's top 30 busiest passenger airports for the period of 2002e2011. The full efficiency levels of Brisbane throughout the study period may be due to its prime location for holiday travel to the principal Australian tourist attractionethe Gold Coast. For Shenzhen, its remarkable record may have been largely due to the rapid economic growth of the Pearl River Delta (PRD) region in Mainland China (particularly experiencing 103.4e187.8% growth for three airport outputs between 2002 and 2011). Twelve airports were considered to be moderate performers since they were efficient in at least one of the ten years during the study periods. Overall, these airports either showed improvements (eight airports) or deteriorations (four airports) in their efficiency levels across the analysis periods, although there was no regular trend with respect to their respective efficiency levels. For the improving airports, in particular, Guangzhou deserves to be explored why its efficiency improved and it became efficient after 2007. Its rapid expansion improved the airport's flight connectivity network, covering more than 200 routes, which translated into an increase in airport traffic. In addition, Sydney was ranked as one of the world's top 30 busiest passenger airports in 2003, and its growth after 2003 could be attributable to its strategic role served as the main international gateway hub airport to and from Australasia and Oceania. In addition, Gimpo's inefficiency before 2010 was likely due to the opening of Incheon in 2000, which adversely affected its operations by attracting away international passenger and cargo traffic. 7 However, its three airport outputs achieved a 5.4e15.0% increase between 2010 and 2011, leading to its full efficiency level in 2011. Likewise, the decline in Jakarta's efficiency was likely related to the Bali bombings that occurred in 2002 and 2005 e these disruptive events had significant negative impacts on international visitors visiting Indonesia (Hitchcock and Putra, 2005) . In particular, Jakarta's positive air traffic growth after 2009 led to its full efficiency levels, with an average annual growth of air passenger numbers (10.6%), air cargo volumes (20.5%), and aircraft movements (8.9%), respectively. The airports never achieved full efficiency levels (i.e. DEA efficiency index ¼ 1) during the study periods (i.e. Incheon, Kuala Lumpur, and Singapore). Interestingly, these three major international hub and gateway airports were considered to be the worst performers. One explanation might be largely related to the consequences of under-utilisation or over-investment in airport resources or high capacity airports handling lower amounts of air traffic. Indeed, further investigation revealed that Incheon and Kuala Lumpur's inefficiencies across the years did not result from recent expansions but from ongoing overcapacity. Likewise, part of the explanation of Singapore's under-utilisation is the result of its passenger terminal expansion in 2007, while its international passenger traffic only increased by less than 3% between 2007 and 2008 as well as between 2009 and 2010, respectively, leaving Singapore with significant excess capacity. Regarding the deteriorating airports, Bangkok's inefficiency after 2008 was primarily the consequence of Thailand's political unrest, which triggered negative airport traffic growth (Yin and Walsh, 2011) . Moreover, Kansai became inefficient after 2006 as an additional runway came into operation in 2007, 8 but its air traffic volumes did not respond with a significant increase accordingly. Passenger terminal expansion might contribute to the deteriorations in efficiency of Manila. Furthermore, Narita became inefficient between 2005 and 2006 as annual air passenger numbers and annual aircraft movements increased by less than 3%, 6 Wober (2007) indicated that the output-oriented model required a given level of inputs to achieve the maximum output levels. In this study, the DEA Output-Oriented model means that airports focus on maximising three categories of air traffic outputs (i.e. air passenger numbers, air cargo volumes, and aircraft movements), holding all of the airport inputs constant. It should be noted that the DEA efficiency indexes of AsiaePacific airports reported above were generally consistent with those reported in the extant literature. In particular, Hong Kong was claimed to be the most efficient airport during the study periods, and Incheon was also claimed to have the worst efficiency (Ha et al., 2010; Lam et al., 2009) . Kuala Lumpur and Singapore were also identified as inefficient airports (Yang, 2010b) , which was largely due to on-going overcapacity and the poor scale efficiency. Overall, the dissimilarity of DEA efficiency indexes (or efficiency ranking) of AsiaePacific airports can be also seen in prior literature as the DEA efficiency indexes computed by the DEA methodology are highly dependent upon the sample size of airports and number of airport input and output variables used during the efficiency evaluation. The average performance of AsiaePacific airports during one particular year compared to other years is very important, as this would indicate whether any year was the best performing year with respect to overall airport efficiency. This is in line with the study of Sengupta (1995) , which stated that industrial competitiveness or efficiency can be evaluated through the analysis of average efficiencies. . This situation indicated that the majority of AsiaePacific airports did not achieve their maximum output levels throughout the study periods. It also corresponds to the fact that the smallest and largest number of efficient airports appeared during 2002 and 2007, respectively. Furthermore, the smallest average DEA efficiency index (in 2002) can be interpreted as, on average, AsiaePacific airports were only 89.1% efficient in that year, or, on average, the airports could almost increase by an additional 10.9% of outputs to attain their maximum outputs using the same amount of inputs. Fewer efficient airports were found during 2002, 2004, and 2010, which could largely be attributable to the impact of the September 11 terrorist attacks in 2001, the outbreak of Severe Acute Respiratory Syndrome (SARS) in late 2002 and mid-2003, and higher aviation fuel prices in 2010. 9 These unfavourable incidents for the global aviation industry may have led to the relatively poor performance of AsiaePacific airports, handling fewer air passenger traffic and air cargo volumes during these periods. That said, air cargo traffic was not as seriously affected as air passenger traffic during the SARS outbreak (Williams, 2006) . The average airport efficiency seemed to remain stable for the periods of 2005e2006 and 2010e2011. It could be said that AsiaePacific airports enjoyed a more favourable operating environment in these four years. More importantly, the best performing year was 2007, when the airport industry in the AsiaePacific region seemed to benefit from a more favourable economic atmosphere for their operations. The declines in average airport efficiency that appeared in 2008 and 2009 might largely be due to aviation fuel price surges alongside the global Remarks: Bold typefaces indicate the efficient airports. a Indicates an airport achieved consistently full efficiency levels. b Indicates an airport showed an improvement in efficiency levels. c Indicates an airport showed a deterioration in efficiency levels. d Indicates an airport never achieved full efficiency levels. economic downturn. These unfavourable economic factors had negative impacts for the worldwide air transport industry and, as a consequence, led to the slump in air passenger travelling and air cargo volumes across the AsiaePacific region. 10 To evaluate the determinants of efficiency of AsiaePacific airports, we adopted the approach of Simar and Wilson (2007) . After obtaining the DEA efficiency indexes in the first-stage, we calculated the following (truncated) regression equation through the bootstrapped procedure in the second stage (with DEA efficiency indexes bounded at both ends of the 0e1 distribution). For further details, see Simar and Wilson, 2007) . where q represents the estimated DEA efficiency score in the firststage. 'Trend' is a yearly trend. 'GDP' represents the logarithm of GDP per capita of the country or city in which an airport is located (in logarithm). 'PIP' represents the percentage of international passengers handled by an airport. The dummy value of airport hub status denoted by 'Hub' is 1 if an airport is an international hub airport, 0 otherwise. The dummy value of airport management denoted by 'Man' is 1 if an airport is government-controlled or owned, 0 otherwise. 'OH' represents airport's daily operating hours. 'Pop' represents the dummy variable which takes 1 if an airport's hinterland population is more than 4 million people, 0 otherwise. 'Alliance' represents alliance membership of dominant airline, and it is a dummy variable which takes the value 1 if the dominant airline of an airport becomes a member of a major global airline strategic alliance, 0 otherwise. First, Im et al.'s (2003) panel unit root test was employed to check the problem of unit roots of all relevant variables. The second-stage estimation results showed the factors for explaining airport efficiency were reported in Table 3 . 11 Four explanatory variables were found to be significant factors for explaining variations in airport efficiency: percentage of international passengers; airport's hinterland population; alliance membership of dominant airline; and the logarithm of GDP per capita. For 'percentage of international passengers' the coefficient was negative; for every percentage increase in international passengers handled by an airport, its efficiency reduced by 0.001 units. Importantly, this finding appears to be consistent with Pathomsiri et al. (2006) , who claimed that the handling of international passenger traffic has a negative impact on an airport's efficiency as larger airport infrastructure and facilities (e.g. check-in counters and baggage handling areas) need to be built to serve international travellers comparing with domestic passengers. We expected the sign of the coefficient estimation for the variable of 'airport hinterland population' to be positive, as a larger hinterland population may generate more airport demand, thus leading to higher airport efficiency. Surprisingly, this variable had a negative impact on airport efficiency. This may suggest that an airport that serves a larger hinterland population is less efficient than an airport that serves a smaller hinterland population; it also suggests that larger airport infrastructure or capacity need to be constructed to accommodate a larger hinterland population and the forecasted growth of air traffic demand across the AsiaePacific region. However, air transport demand and airport operations were inevitably affected by unwanted adverse incidents or difficult operating conditions that led to lower airport efficiency (Grais et al., Remarks: *, **, and *** indicate that the explanatory variable is significant at the 0.10, 0.05, and 0.01 significance level, respectively. The truncated regression analysis with bootstrapping (Simar and Wilson, 2007) results above was derived from 5000 bootstrapped iterations. 10 The total air cargo volumes of the sampled AsiaePacific airports showed negative growth of 3.2% between 2007 and 2008. 11 To check the robustness of the SimareWilson bootstrapping regression analysis, the random-effects (RE) Tobit model was also computed in this study (see Merkert and Hensher, 2011) . Table 3 shows both methodologies presented almost similar estimation results, therefore this study mostly discussed the estimation results of the SimareWilson bootstrapping regression analysis. 2003; Kozak et al., 2007; Siu and Wong, 2004) . Also, it should be acknowledged that it is extremely difficult to define the exact size of an airport's hinterland size due to improvements in aircraft technology that allow longer distance to be flown, the formation of strategic alliances between airlines, the establishment of hub-andspoke networks by many airlines, and airport overlap or congestion in multi-airport region (MAR) in which an airport competes air traffic volumes with its neighbouring airports. 12 (e.g. Graham, 1999; Graham and Guyer, 2000; Williams, 2006) . The coefficient suggests that if an airport serves a larger hinterland population, its efficiency would drop by 0.126 units. 'Alliance membership of dominant airline' variable was also reported as significant in both estimations, and suggests that if an airport's dominant airline enters a global airline strategic alliance, this might positively influence its home-based airport's efficiency; when the dominant airline(s) of an airport enters a global airline strategic alliance, the airport's efficiency will increase by 0.080 units as allied airlines could share airport facilities to handle more connecting traffic. More importantly, this finding provides evidence to support the argument of Gillen and Lall (1997) , who claimed that common use of airport facilities can improve efficiency by allocating passenger terminal facilities for airlines of a particular alliance so they have exclusive use of the passenger terminals. This gives airlines an incentive to use the designated passenger terminals more efficiently. Also, the current situation shows that an increasing number of large or legacy airlines have joined or intend to enter three major global airline strategic alliances (i.e. oneworld, Star Alliance, and SkyTeam) or formed their own partnerships (e.g. Qantas Airways and Emirates Airline). Importantly, allied activities between partner airlines are seen to affect airport operations in different ways such as a specific passenger terminal (e.g. Narita's Terminal One) being designated for a group of airlines associated with a particular alliance (in Narita's case, Star Alliance) (Cento, 2009) . As expected, "ln GDP per capita" has a positive and significant impact on airport efficiency when we used the Random Effect Tobit regression. This implies that there is a positive relationship between GDP per capita of a country or city with an airport's traffic demand (Abed et al., 2001; Tsui et al., 2014a) , and an airport's efficiency would be improved. The remaining variables were not statistically significant. For example, 'airport hub status' has no significant impact on the efficiency of an airport but its coefficient may imply that if an airport that serves as an international hub airport could be more efficient than those serve as regional airports or non-hub airports in the AsiaePacific region. Prior studies (e.g. Fung et al., 2008; Gillen and Lall, 1997; Lin and Hong, 2006; Perelman and Serebrisky, 2010; Tsui et al., 2014b) also claimed that international hub airports possess size and location advantages for transporting more airport traffic and, as a consequence, improve airport efficiency. Also, the insignificant variable of 'airport management' might imply that government-controlled/owned airports might perform better than privately-controlled/owned airports among the sampled AsiaePacific airports. It is worthwhile to note that this finding is not consistent with the literature relating to the effect of airport management/ownership upon airport efficiency (e.g. Barros and Dieke, 2007; Muller et al., 2009; Oum et al., 2006 Oum et al., , 2008 . As many key Asian international hub airports (e.g. Beijing, Hong Kong, and Singapore) are still under government ownership and control, since the governments consider an airport to be the strategic asset and/or an engine to contribute economic development of the country and city (Doganis, 1992) . Indeed, these airports now tend to operate on a more commercial basis, rather than being guided by non-economic political objectives while facing the growth in air transport demand and other emerging competitors in the region (Hooper, 2002) . Moreover, many AsiaePacific airports have been fully or partially privatised as the benefits of airport efficiency improvement and finance support for future development (Oum et al., 2006) . In addition, the insignificant positive coefficient of 'airport operating hours' might imply that longer operating hours of an airport might positively influence its operations and increase efficiency. This finding is in line with the perspective argued by Humphreys and Francis (2000) , and demonstrates that the duration of airport operating hours is a significant factor that positively affects airport operations and efficiency. However, this presumably cannot apply to Adelaide, Narita, and Sydney due to their curfew policies. The main purpose of our paper was to investigate the operational efficiency of 21 major airports in the AsiaePacific region, and identify the key factors to explain variations in airport efficiency. The empirical results suggested that six airports (i.e. Adelaide, Beijing, Brisbane, Hong Kong, Melbourne, and Shenzhen) are the 'efficient' airports which operated at the efficiency frontier during the period of this study. In addition, the average DEA efficiency indexes of AsiaePacific airports suggested a varying trend throughout the study periods, and that most airports operated below their optimal output levels. Four significant factors were found to account for the identified variations in airport efficiency among AsiaePacific airports: (i) more international passengers handled by an airport that may reduce its efficiency level; (ii) when an airport caters to a larger hinterland population, it will become less efficient than an airport that serves a smaller hinterland population; (iii) if the dominant airline(s) of an airport enters a global airline strategic alliance, this may improve its home-based airport's efficiency; and (iv) having an increase in GDP per capita of a country or city might increase an airport's efficiency. Airport management should also seriously pay attention to other controllable factorsefactors under managerial control (e.g. outsourcing activities and concession revenues) affecting airport efficiency. Nowadays, many airports worldwide have outsourced some operational functions and services to the third parties for saving operating costs, and also made efforts to generate nonaeronautical revenues (e.g. concession revenues). Unfortunately, such important airport efficiency measurements could not be included in this study because of lack of available financial data related to most of the sampled AsiaePacific airports. As an extension of this study, it may be meaningful to include such data (when available) that allows this study to take account of the effects of airports' strategy with regard to outsourcing activities and concession revenues on AsiaePacific airports' efficiency. Furthermore, it is important to consider the actual and likely impact of the global airline strategic alliance or other forms of airline partnerships on airport efficiency. For example, the recent partnership between Qantas Airways and Emirates Airline, aims to deliver the best in their respective flight networks and frequencies, lounges, loyalty programs, and customer experiences. Under this agreement, Qantas Airways will move its hub at Singapore Changi Airport to Dubai International Airport, which may reduce the amount of transit traffic to Europe via Singapore Changi Airport 12 For example, the Pearl River Delta (PRD) region in Mainland China is considered as one of the busiest multi-airport regions, in which Hong Kong competes with four other neighbouring airports, namely Guangzhou, Macau, Shenzhen, and Xiamen Airports. transported by Qantas Airways, thus affecting its traffic volume and efficiency. Total factor productivity and efficiency of Australian airports An econometric analysis of international air travel demand in Saudi Arabia Airport annual traffic statistics Some models for estimating technical and scale inefficiencies in data envelopment analysis An introduction to data envelopment analysis with some of its model and their uses Airports in Argentina: technical efficiency in the context of an economic crisis Performance evaluation of Italian airports: a data envelopment analysis Measuring the economic efficiency of airports: a Simar-Wilson methodology analysis Technical and allocative efficiency in airports The technical efficiency of Canadian urban transit systems A comparative study of efficiency in European banking The airline Industry: challenges in the 21st Century Measuring the efficiency of decision making units Airport performance: a comparative study between Changi airport and airports in the New York-Jersey Metropolitan Area Introduction to Data Envelopment Analysis and its Uses: With DEA-Solver Software and references New evidence on the efficiency of Italian airports: a bootstrapped DEA analysis Efficient use of airport capacity Measuring the efficiency of European airlines: an application of DEA and Tobit analysis Regulation under stress: development in Australian airport policy Productivity changes in Chinese airports 1995e2004 Developing measures of airport productivity and performance: an application of data envelopment analysis Managing Airports: an International Perspectives Airport-specific traffic forecasts: a critical perspective The role of regional airports and air services in the United Kingdom Assessing the impact of airline travel on the geographic spread of pandemic influenza Comparative analysis of efficiency for major Northeast Asia airports The Bali Bombings: tourism crisis management and conflict avoidance Traditional airport performance indicators: a critical perspective Asia-Pacific area traffic forecasts 2008e2025. International Civil Aviation Organisation, DOC 9915. Paper presented at the Fourteen Meeting of the Asia-Pacific Area Traffic Forecasting Group Testing for unit roots in heterogeneous panels The impact of the perception of risk on International Travellers Operational efficiency across Asia-Pacific airports Route-based performance evaluation of Taiwanese domestic airlines using data envelopment analysis: a comment Operational performance evaluation of international major airports: an application of data envelopment analysis Efficiency analysis and target setting of Spanish airports An empirical investigation on the efficiency, capacity and ownership of Italian airports An application of DEA to measure the efficiency of Spanish airports prior to privatisation International air traffic in the Baltic Sea Area: Huggateway status and prospects. Copenhagen in focus The impact of strategic management and fleet planning on airline efficiency e a random effects Tobit model based on DEA efficiency scores Management of airports in extreme winter conditions e some lessons from analysing the efficiency of Norwegian airports A review of different benchmarking methods in the context of regional airports Privatisation, restructuring and its effects on performance: a comparison between German and British airports An analysis of technical efficiency and productivity changes in Spanish airports using Malmquist index Airport development in Southeast Asia Privatization, corporation, ownership forms and their effects on the performance of the world's major airports Ownership forms matter for airport efficiency: a stochastic frontier investigation of worldwide airports Economic efficiency of railways and implications for public policy: a comparative study of the OECD countries railways Managerial efficiency of Brazilian airports An analysis for the competitive strength of Asian major airports The performance of BAA before and after privatization: a DEA study Measuring and determination of airport productivity in competitive markets Relative efficiency of European airports Inefficiencies and scale economies of European airport operations Measuring Technical Efficiency of Airports in Latin America. The World Bank Identifying sub-areas that comprise a greater metropolitan area: the criterion of country relative efficiency An empirical study of Iranian regional airports using robust data envelopment analysis An analysis of the operational efficiency of major airports in the United States Performance based clustering for benchmarking of US airports Dynamics of Data Envelopment Analysis: Theory of Systems Efficiency Estimation and inference in two-stage, semi-parametric models of productive efficiency Statistical inference in non-parametric frontier models: recent developments and perspectives Economic impact of SARS: the case of Hong Kong Introduction to Transportation Greek airports: efficiency measurement and analysis of determinants Forecasting of Hong Kong airport's passenger throughput Estimating airport efficiency of New Zealand airports Developing Strategies for the Modern International Airport: East Asia and Beyond Data envelopment analysis Overcoming the inherent dependence of DEA efficiency scores: a bootstrap approach. Unpublished Working Paper Measuring the efficiency of Asia-Pacific international airports-Parametric and non-parametric evidence Efficiency and productivity evidence from international airports in the Asia-Pacific region The privatization and commercialization of China's airports Restoring the image of a tourism destination in crisis: the Tourism Authority of Thailand's (TAT) Strategic Crisis Management Strategies and Destination Marketing Approaches in responding to political crises Japanese-airport benchmarking with the DEA and endogenous-weight TFP methods: testing the criticism of overinvestment in Japanese regional airports Effects of competition and policy changes on Chinese airport productivity: an empirical investigation Analysis of an international air-cargo hub: the case of Hong Kong Ownership structure and determinants of technical efficiency: an application of Data Envelopment Analysis to Chinese Enterprises (1986e1990)