Impacts of low cost carrier services on efficiency of the major U.S. airports

Impacts of low cost carrier services on efficiency of the major U.S. airports

Journal of Air Transport Management 33 (2013) 60e67 Contents lists available at SciVerse ScienceDirect Journal of Air Transport Management journal h...

278KB Sizes 13 Downloads 43 Views

Journal of Air Transport Management 33 (2013) 60e67

Contents lists available at SciVerse ScienceDirect

Journal of Air Transport Management journal homepage: www.elsevier.com/locate/jairtraman

Impacts of low cost carrier services on efficiency of the major U.S. airports Yap Yin Choo*, Tae Hoon Oum Sauder School of Business, The University of British Columbia, Vancouver V6T 1Z2, Canada

a b s t r a c t Keywords: Low cost carrier services Effects on airport efficiency Major US airports

While traditionally many LCCs use secondary airports because of the availability of slots, spare capacity and low aeronautical cost, in recent years more and more LCCs are shifting their operations to major airports. This has made necessary for some major hub airports to attempt to attract LCC business. Since it is not clear whether such a move would improve or harm efficiency of the hub airports, this paper investigates this issue using a panel data of 63 major U.S. airports for the 2007e2010 period. After controlling the effects of other airport characteristics on efficiency, we found non-monotonic relationships between the level of LCC presence and airport efficiency. Efficiency of airport decreases as LCC presence increases from very low level, reaches the lowest efficiency point, and then increases as LCC services become dominant. Our empirical findings suggest existence of economies of airport specialization in either FSCs or LCCs. The even mix between FSC and LCC services at an airport appears to be inefficient. This result is somewhat contradictory to the findings of recent studies on LCC effects on European airports. Crown Copyright Ó 2013 Published by Elsevier Ltd. All rights reserved.

1. Introduction The emergence and growth of low cost carriers (LCCs) in the last few decades is spectacular and unprecedented following deregulation and a numbers of market liberalization. The LCC model which was pioneered by Southwest Airlines in 1970 has recorded a tremendous success and brought unprecedented benefits to travelers and economy to the extent that many of us now call “Southwest effect”. Following that, LCC services have been expanding vigorously in Europe, Oceania, South America and Asia, and thus, exert enormous pressure to full service carriers (FSC). In 2012, LCCs’ share of seats accounted for 26% of the world market, which have been doubled in the last 10 years (OAG Aviation, 2012). In the United States, the Airports Council International, North America ACI (2012) reported that the share of LCC enplanements has increased at a rate of 3.5% per year while FSC enplanements recorded a compounded annual rate of reduction of 1.9% during the 2007e 2011 period.

* Corresponding author. E-mail address: [email protected] (Y.Y. Choo).

While numerous studies have been conducted on the effects of LCCs on air fares, passenger traffic and competition, as pointed out by Graham (2013) 1 there is very limited amount of literature on the effects of LCC on airport operations and performance, and also, the geographic coverage of such studies are limited mostly to Europe. Furthermore, the empirical findings on the effects of LCCs on airport efficiency are inconclusive. Hence, the purpose of this paper is to fill this apparent gap in literature by analyzing the impact of LCCs on airport efficiency in the United States. This paper is organized as follows. The next section provides a literature review on the effects of LCCs on airport performance. Section 3 reviews the methodologies used to measure airport productivity and efficiency, followed by a discussion on model formulation in Section 4. The data used for our empirical analysis is described in Section 5 while the results of econometric analyses on the impacts of LCCs presence on efficiency of major US airports is given in Section 6. Conclusions are drawn in Section 7.

1 Graham (2013) conducted an extensive literature review on the relationship between low cost carriers and airports, and categorized them into airport demand issues (such as factors affecting airport choice) and supply issues (such as LCC impacts on airport performance and design). Only 8% of the literature reviewed is about North America.

0969-6997/$ e see front matter Crown Copyright Ó 2013 Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jairtraman.2013.06.010

Y.Y. Choo, T.H. Oum / Journal of Air Transport Management 33 (2013) 60e67

2. Literature and discussion on potential effects of LCC on airport performance During the past decade or so, there has been much increased effort on measuring, comparing and benchmarking airport performance. For example, the Air Transport Research Society (ATRS, 2002-2012) has been publishing their airport benchmarking task force report on productivity and efficiency of major airports in Asia, Europe, North America and Oceania, every year since 2002. As summarized and reviewed comprehensively in Liebert and Niemeier (2013), a large number of academic papers have been published concerning methodology and empirical measurement of airport productivity and efficiency.2 However, only a very limited number of studies investigated the relationship between LCCs and airports. To our knowledge, Bottasso et al. (2012) are the only study which addressed the issue of LCC effects on airport efficiency. Using a panel data of large UK airports for the 2002e 2005 period, Bottasso et al. (2012) found that the intensity of LCC presence impacted positively on total factor productivity of the UK airports. Furthermore, they suggested two possible reasons for this finding: (i) traffic increase stimulated by LCCs will generate larger non-aeronautical revenue sources for the airports, and (ii) the airports adopt more efficient organizational model in order to meet LCCs operational requirements. Obviously, the extent of LCC presence affects both revenues (outputs) and costs (inputs) of the airports, and thus, productivity (efficiency) of the airports. However, literature on the directions of LCC effects on revenues and costs are not clear-cut and thus largely empirical. On the revenue (output) side, LCCs may have both positive and negative effects. Many LCCs’ airports are secondary airports, located at remote site, which may increase the demand for car parking and car hire business to compensate for their lower aeronautical revenue. Certainly, airports nowadays have to diversify their revenue source from duty free outlets, catering, food and beverage services, car-office-space rental services, car parking, etc. The academic literature is not lacking of the evidence on the LCCs’ roles in increasing passenger traffic and associated economic activities at the airports and regions. For instance, Graham and Dennis (2007) show that LCCs contributed to increased passenger load in selected airports in UK and Ireland. Dennis (2007) and Barrett (2004) also presented similar findings on a number of airports in Europe. However, whether the increase of LCCs traffic translates into higher revenue is uncertain. While LCCs bring forth business opportunities to airports via the increased traffic and thus generating more revenue from non-aeronautical sources,3 LCCs pay lower aeronautical fees for use of the airport, and provide basis even for the Full Service Carriers (FSC) to demand lower aeronautical fees (see Humphreys et al. (2006) for an expanded discussion on these). Further, the evidence supporting the higher nonaeronautical revenue from increased LCC traffic is also unclear (Graham, 2013). For example, Castillo-Manzano et al. (2010) found that LCC passengers at Spanish regional airports spent 7% less than FSC passengers, and Lei and Papatheodorou (2010) found that the average commercial revenue spent per LCC passenger at Greek airports was lower than FSC passengers.4 On the contrary, Gillen

2 It is interesting to note that after review of such extensive literature, the authors conclude that the effects of ownership forms, privatization, and scale on airport performance are inconclusive, and thus, need further research. 3 ATRS (2002-2012) found that airports in North America, Europe and Asia Pacific generated at least 50% of their revenue from non-aeronautical source in 2010. 4 Francis et al. (2003) revealed that despite experiencing growth in LCC passenger traffic, the commercial revenue at those two anonymous European airports remained low at only 21% of total operating revenue.

61

and Lall (2004) observed that the non-aeronautical revenue per enplaned passenger for Albany (NY) airport increased from US$9.70 in 1999 to US$10.55 in 2000 when Southwest airline began its operation. On the cost (input) side, there are also both sides of the arguments, i.e., positive and negative effects of LCC presence on airport costs (input use). In order to attract and support LCCs, airports need to offer simplified passenger and flight processes while being cost efficient to be able to afford low aeronautical charges. This would have a strong positive effect on cost (input use) efficiency of airports. However, LCCs not only requires more efficient operational processes, but also require airports to customize facilities and designs in order to support much needed short turnaround time for aircraft.5,6,7 However, restructuring and streamlining airport activities to align with LCCs’ “no frills” services is a great challenge, especially, to major airports that were designed to provide high quality services (to legacy carriers) for expensive user charges.8,9 In order to serve the rapidly expanding LCC clients, airports need to differentiate their services that often involve more than changes in operational processes but design of airport terminals.10 As the LCCs no longer limit their flights to secondary airports, LCC presences have been increasing at major airports such as Atlanta, Los Angles, San Francisco, New York’s JFK, La Guardia, Netwark, etc., and this poses a challenge for them to come up with flexible operational design to differentiate their services between LCC and FSC services. Therefore, an important question to ask is whether it is more efficient for the airports handling a mixture of both LCCs and FSCs than the airports specializing in handling FSCs or LCCs traffic only. Adding more LCCs traffic to an existing airport handling predominantly FSC traffic may not be efficient if there are economies of specialization.11 This is an interesting empirical question to explore in this study.12

5 The literature on LCC’s requirement for airport facilities agree on the importance of having quick turnaround facilities. For instance, Warnock-Smith and Potter (2005) found that high demand for LCC services within the airport catchment area, quick and efficient turnaround facilities, slot availability, and good aeronautical fees discount are the top four factors that LCCs consider in choosing airport to fly to. Similarly, Barrett (2004), through an interview with the CEO of Ryanair, found that low airport charges, quick 25 min turnaround time, single-storey airport terminals and quick check-in are among the critical requirements for LCC airport choice. 6 De Neufville (2008) provided specific design standards fitting LCCs business model such as higher densities of passengers per unit of area, fewer gate positions for a given number of daily flight, and recommended a flexible design to manage risk and uncertainty in the future. 7 LCCs seek quick turnaround time between flight arrival and departure, usually between 25 min and 20 min to maximize their aircraft utilization. As such, availability of slot and spare airport capacity is critical to support LCCs’ business model, which explains why, early on, many LCCs avoided primary airports and use secondary airports (Barbot, 2006). 8 For instance, AirAsia was upset with the new low cost terminal (KLIA2) built by Malaysia Airport Holding Berhad to include aerobridges and better infrastructure, which inflated the cost from US$ 629 million in 2009 to US$1.22 billion. 9 Ben Abda et al. (2012) found that most LCCs presence and market share were at the largest airports recently. Unlike the secondary airports, most of these major airports have slot constraint. 10 Bangkok learnt a painful experience as it had to delays the opening of Bangkok Suvarnabhumi airport, causing an increase of more than 25% capitalized cost due to inability to adjust itself to LCC services (De Neufville, 2008). 11 Recently, as a solution to avoid conflicting needs between FSC and LCC services, some airports have developed a separate low cost terminal (LCT) i.e. Marseille, Bordeaux, Lyon, Tampere Pirkkala, Turku, Budapest, Amsterdam and Copenhagen in Europe, Austin and New York JFK in North America, Singapore, Kuala Lumpur and Zhengzhou in Asia (Hanaoka and Saraswati, 2011). 12 Graham (2013) reports that insufficient literature on LCT made it difficult to assess the success and financial viability of LCT. Furthermore, the development of LCT that coexists with legacy terminal is a relatively new phenomenon in airport industry and its effect is hard to gauge at this juncture.

62

Y.Y. Choo, T.H. Oum / Journal of Air Transport Management 33 (2013) 60e67

3. Methodology for efficiency measurement There are three commonly used methods measuring and analyzing the productivity and efficiency among airports are: (1) Index Number Method, (2) Data Envelopment Analysis (DEA), and (3) Stochastic Frontier Analysis (SFA) Method (see, for example, Liebert and Niemeier, 2013; Lin et al., 2013; Oum et al., 1992). Lin et al. (2013) found that among these three methodologies, the efficiency rankings based on DEA method are considerably more different from the other two methods. In this paper, therefore, we use the Index Number Approach as the primary method for measuring airport efficiency since it is the method employed by the Air Transport Research Society (ATRS, 2002-2012) global airport performance task force reports, from which we drew most of the extensive data required for this study. In addition, SFA approach is used for robustness check on the results. As details of each of these methods are discussed elsewhere including in the three papers just cited, this section will describe the main properties of these two methodologies only briefly. As a non-parametric approach, the Index Number Method directly measures productivity as output index over input index. The total factor productivity of a firm is calculated as the ratio of aggregate output index over aggregate input index. However, as it is impossible to measure capital input quantities and/or costs consistently across all airports because investments and expenditures on capital equipment, buildings and other infrastructure inputs such as runways and terminals are often invested over a long time and riddled with “hidden” subsidies and taxes. Therefore, following the ATRS global airport benchmarking task force tradition (ATRS, 2002-2012) in this study the variable factor productivity (VFP) measures are used as airport operating efficiency indicator. This is equivalent to the assumption that airport capital inputs are quasi-fixed. Since airports use multiple inputs to produce multiple outputs, the multilateral index number method proposed by Caves et al. (1982) is used to aggregate the inputs and outputs, and to compute the VRP indices as follows:



Yi ¼ f ðxi ; bÞexpðVi  Ui Þ

  lnVFPk  lnVFPj ¼ lnYk  lnYj  lnXk  lnXj

(2)

where, Yi represents the output of the ith firm; f(xi;b) is the deterministic core function of an input vector xi, and an unknown parametric vector b; Vi is a normally distributed random variable that represents the effects of unobservable explanatory variables and random shocks. Ui is a non-negative random variable representing inefficiency which is assumed to follow either half-normal, exponential, or gamma distribution. Vi is the traditional symmetric noise term. If we adopt a translog specification for the deterministic part of production function, the Stochastic Frontier production function can be expressed as:

lnYi ¼ b0 þ

n X

bj lnXj þ

j¼1

n X n X

bjk lnXj lnXk þ ðVi  Ui Þ

(3)

j¼1 k¼1

where, Yi is aggregate output index for airport i; Xj is the jth input; Vi is assumed to follow the distribution N ð0; s2V Þ; Ui is assumed to follow N ðm; s2V Þ where m  0. One may compute the technical efficiency of firm i by evaluating the following expression:

TEi ¼



X Rij þ Ri Yij X R þ Ri Y ik ¼ ln wik  ln w 2 2 i i Yi Yi XW þ Wi X X Wij þ W i Xij ik  ln wik þ ln w 2 2 i i Xi Xi

employees directly paid for by airport operators, and the “soft cost inputs” which include catch-all other variable inputs (other than labor input) including outsourced services for the goods, services and materials purchased directly by an airport. Similar to nonaeronautical output index, the “soft cost input” is deflated by cost of living index of the metropolitan census division. As mentioned previously, a Stochastic Frontier Analysis (SFA) method is used for a robustness check of our empirical results. The basic empirical framework for SFA is to specify a production function and add an error term with two components: negative (truncated) error term to capture inefficiency inherent in the particular observation, and the traditional symmetric noise term. The general form of stochastic frontier production function can be written as follows:

    E Yi c Ui Xi EðYi jUi ¼ 0; XÞ

¼ expðUi Þ

(4)

The SFA production function in Equation (3) is estimated by using the input quantity indices (labor input and soft cost input) and the output quantity index (aggregated using the multilateral index procedure).

(1)

where VFPk is the productivity of kth firm; Yik and Xik represent the ith output and input of the kth firm respectively; Rik and Wik are the weights for the ith output and input of the kth firm, respectively using revenue and cost shares; A bar over weights represents sample arithmetic mean, while a tilde indicates geometric mean. We consider numbers of aircraft movements (ATMs), passenger volume and non-aeronautical revenue as outputs to be aggregated using the multi-lateral index procedure. To deal with the price differences between cities in the U.S., the city-based Cost of Living Index (COLI) is used to deflate non-aeronautical revenue to compute the quantity index of non-aeronautical revenue output.13 On the input side, two variable input categories are included: the labor input which is defined by number of full time equivalent

13 As air cargoes are handled directly by airlines and/or third-party logistics firms and constitute a small portion of most airport’ revenue, it is not included explicitly as an individual output for computing VFP. Therefore, we included the percentage of air cargo revenue in airport’s total revenue as an explanatory variable later in the second stage regression analysis on VFP.

4. Description of data As pointed out previously, the primary source of our data for this research is the Airport Benchmarking Database compiled and maintained by the Air Transport Research Society (ATRS, 20022012). The ATRS database contains revenues, costs and other financial data, traffic data, user charges and capacity data of the major airports and airport authorities around the world. Our sample for this study consists of unbalanced panel data of 63 major U.S. airports for the 2007e2010 period (Please see Appendix 1 for the list of U.S. airports included in this study). The ATRS data is compiled from various sources including airports’ annual reports, the US Federal Aviation Administration (FAA) and the data collected directly by contacting airports. However, the main variable of our interest, the percentage of LCC passengers in the total enplaned and deplaned passengers at each airport, is not available in the database, and thus, needed to be collected for this study. In order to compute the share of LCC passengers in a particular airport, we used the Air Carrier Statistics database (T-100 data bank compiled by Office of Airline Information, Bureau of Transportation Statistics, Research and Innovative Technology Administration and is available online).

Y.Y. Choo, T.H. Oum / Journal of Air Transport Management 33 (2013) 60e67

T-100 database contains origin-destination passengers carried between airports by domestic and international airlines. The share of LCCs is calculated from sum of enplaned and deplaned passengers carried by LCCs divided by total passengers at each airport.14 The ATRS benchmarking study has already identified key characteristics of airports which affect airport efficiency, in addition to the LCC passenger share we have decided to include those key airport characteristics variables: i.e., airport output scale, share of non-aeronautical revenue in airport’s total revenue, average size of aircraft used the airport, share of international traffic handled at the airport, share of connecting passengers in total passengers handled at the airport, air cargo tonnes handled at the airport. Table 1 provides the summary statistics for the variables to be used in our empirical analysis including the two alternative dependent variables (Variable Factor Productivity Index, VFP, and SFA Efficiency Index). Note that the share of LCC passengers in our sample airport data is, on average, 31.1%, ranging from the minimum of 0.43% for Houston Bush International Airport (IAH) to the maximum of 97% for Dallas Love Field airport (DAL), a base of Southwest Airline. Table 2 shows the growth of LCC shares in the U.S. airports by hub-type during this study’s sample period (2007e2010).15 Regardless of the hub categories, the growth of LCC passenger share is still prominent. Especially, the medium hub airports’ LCC passenger shares increased from 34% in 2007 to 44% in 2010. 5. Model estimation and empirical results Utilizing the results of the ATRS investigation, all of the regression models we estimate in this study are of log-linear regression on VFP efficiency scores. We estimate the Ordinary Least Squares (OLS) regression. The formal panel data regression procedures are also used in order to remove potential omitted variable bias as these procedures can capture the unobserved airport-specific heterogeneity and thereby remove omitted variable bias. To capture the unobserved airport-specific heterogeneity, both fixed and random estimators can be used. The crucial distinction between fixed and random effects lies in the assumption whether there is any correlation between the unobserved (omitted) variables that affect the dependent variable and the explanatory variables included in the model. The random effect model assumes the unobserved individual effects are random variables that are distributed independently of the explanatory variables included in the model. It also assumes that the unobserved, time-invariant individual effects are uncorrelated with all other independent variables in the model. On the other hand, the fixed effect model treats the unobserved individual effects as random variables that are potentially correlated with the explanatory variables. The fixed effect eliminates time-invariant elements; hence, the coefficients of time-invariant explanatory variables are not identified. Table 3 reports the log-linear models of Variable Factor Productivity (VFP) estimated by the Ordinary Least Squares (OLS), the Random Effect model, and by the Fixed Effect model, respectively, in Columns 1, 2 and 3. The share of LCCs passengers (LCCs) has a statistically significant negative coefficient in all of the three regressions regardless of the

14 In this research, we use the definition of LCCs set by Center for Asia Pacific Aviation (CAPA). The CAPA LCC definition includes AirTran, Allegiant Air, Frontier Airlines, Jetblue Airways, PEOPLExpress, Southwest Airlines, Spirit Airlines, Sun Country, Sunwing Airlines, USA3000 Airlines and Virgin America. 15 FAA defines airport categories in terms of the percentage of total passenger boardings within the United States. Most airports fall into Medium and Large hub, which are airports having at least 0.25% of national total passenger boardings.

63

Table 1 Summary statistics of the variables. Variable

Mean

Std. dev.

Min

Max

VFP 0.9708 0.2546 0.5233 1.9321 SFA 0.2937 0.0939 0.1180 0.5102 LCC passenger shares 0.3107 0.2289 0.0043 0.9691 Output index 1.3781 0.9900 0.2529 5.1888 % of non-aero revenue 0.4986 0.1313 0.1552 0.8572 % of international Pax 0.0755 0.1113 0.0000 0.5236 % connecting Pax 0.2007 0.2041 0.0005 0.7740 % cargo traffic 0.0142 0.0180 0.0001 0.0866 Runway utilization 78,132 41,950 14,380 223,126 Aircraft size 73.49 19.70 21.94 119.29 Total number of observations ¼ 229 VFP ¼ variable factor productivity index; SFA ¼ stochastic frontier analysis core. Table 2 Average share of LCCs passenger by hub type. Hub

Count

2007

2008

2009

2010

Large Medium Small

32 20e28 2e3

0.24 (0.23) 0.34 (0.15) 0.26 (0.14)

0.25 (0.24) 0.36 (0.15) 0.27 (0.13)

0.26 (0.25) 0.40 (0.20) 0.29 (0.14)

0.27 (0.26) 0.44 (0.22) 0.31 (0.12)

Standard deviation in parentheses. Hub type is differentiated by the percentage of annual passenger boardings within the United States with: Large ¼ 1% or more; Medium ¼ At least 0.25%, but less than 1%; Small ¼ At least 0.05%, but less than 0.25%.

estimation method used. This very robust result is contrary to the findings of Bottasso et al. (2012) who found a positive impact of LCCs’ presence on airport productivity based on UK airports data. At the first glance, this result appears to be counter-intuitive because adapting to LCCs operational needs such as short turnaround time may induce or force airports to be more efficient. However, there may be several reasons why we observe negative effects of LCC presence on airport efficiency in North America, differently from the positive effects of LCC presence on UK airport efficiency, the key findings of Bottasso et al. (2012). First, while most LCCs in UK avoid primary airports and use secondary airports, LCCs in United States including Southwest, US airways, Virgin America, etc. have penetrated into many primary airports in major ways. As such, many major airports have significant mix of FSA and LCC traffic, and may have difficulty serving both constituencies efficiently. Second, all of the US full service carriers do not serve free meals to economy passengers, and thus, both FSC passengers and LCC passengers need to buy meals at airports. Since FSC passengers tend to buy higher priced foods than LCC passengers, the airports with higher share of LCC passengers would generate less nonaeronautical revenue (including concession revenue) per passenger than similar airports with higher share of FSC passengers. This difference is non-trivial considering that non-aeronautical revenue per passenger in US airports is, on average, about $4 per passenger.16 The third potential reason why Bottasso et al. (2012) was able to obtain the positive sign on LCC presence variable may be because they did not control for airport characteristics variables. Considering that many of our airport characteristics variables (share of connecting passengers, share of international passengers, runway utilization factor, average size of aircraft using the airport, etc.) statistically significant in our regression models, it is possible that their regression model suffers from the omitted variable bias. The coefficients of other variables impacting airport efficiency by and

16 For example, during our sample period (2007e2010) Dallas-Ft Worth airport (AA hub) averaged non-aeronautical revenue of $4.89 per passenger while Dallas Love Field airport (SW focus city) averaged only $3.16 per passenger.

64

Y.Y. Choo, T.H. Oum / Journal of Air Transport Management 33 (2013) 60e67

Table 3 Regression parameter estimates (Dependent variable: VFP). VFP

LCCs Output Non-aero share International Connecting Cargo Runway utilization Aircraft size 2008 2009 2010 Constant R-squared

OLS

Random effect

Fixed effect

Coef.

Std. err.

Coef.

Std. err.

Coef.

Std. err.

0.1586** 0.0447 1.2139** 0.1934 0.3128** 1.0214 0.0668**

0.0646 0.0354 0.1232 0.1549 0.0933 0.9662 0.0369

0.1964** 0.0434 0.7605** 0.4250** 0.2179* 2.8781* 0.0779*

0.1075 0.0537 0.1633 0.1990 0.1220 1.6788 0.0478

0.5063** 0.4405** 0.1065 0.0108 0.2258 2.8791 0.1038

0.2165 0.1887 0.2703 0.2063 0.2417 2.4933 0.1080

0.2218** 0.0109 0.0315 0.0280 0.4672 0.4722

0.0591 0.0351 0.0378 0.0357 0.4833

0.1512** 0.0166 0.0064 0.0005 0.5706 0.4203

0.0703 0.0139 0.0184 0.0195 0.7430

0.0830 0.0217** 0.0150 0.0067 1.5046 0.3316

0.1237 0.0116 0.0245 0.0275 1.5777

*The coefficient is significant at the 10% level. **The coefficient is significant at the 5% level.

large are the same as reported in the ATRS airport benchmarking reports. For interpretation of these controlling variables, readers are referred to the ATRS reports. As discussed in the literature review section, there is a possibility of existence of economies specialization of airports either in FSC services or in LCC services. There is also some anecdotal evidence showing that airports, especially US airports, have not perfected their management and operational model for serving substantial mix of FSC and LCC traffic efficiently. In order to test for this hypothesis, we added a new variable (Square of LCC presence variable, LCC^2) in all three alternative forms of regressions. The new improved models are reported in Table 4. The coefficients for LCCs remain negative and statistically significant in all three models. On the other hand, the (LCC^2) variable has a positive coefficient in all models, and statistically significant in the OLS and fixed effect panel regressions. This provides some evidence of the non-monotonic relationship between the level of LCC presence and the airport efficiency. For the airports designed mainly to serve FSC passengers (legacy carrier hub airports), the increase in LCC traffic reduces the efficiency of the airport due to loss in economies of specialization. On the other hand, the efficiency increases for the airport with very high concentration of LCC traffic since their facilities and employees are geared up to serve LCC passengers. This findings suggest that an airport becomes more efficient in either by

concentrating on the FSC services or LCC services but not both. Fig. 1 depicts the marginal relationship between the share of LCC passengers and the variable factor productivity of US airports (after removing the effects of other control variables in the regression). Fig. 1 presents the marginal effects of share of LCC traffic on variable factor productivity of US airports. This figure confirms with our refined expectation that, ceteris paribus, the airports with high concentration of FSC passenger services or the airports with high concentration of LCC passenger services achieve higher productivity and thus more efficient than the airports mixing substantial FSC-LCC traffic together. Our result in Fig. 1 shows that about 60% LCC-40% FCC mix would be the worst case. Based on Fig. 1, we categorized our airport sample into three groups based on the share of LCC passengers: (a) Airports with LCC shares less than 40%; (b) Airports with 40e80% LCC shares; (c) Airports with more than 80% LCC shares. Then we ran new set of regressions by including the following dummy slope variables: (LCC share*dummy variable for category (a)) and (LCC share*dummy variable category(c)). The new regression results are reported in Table 5. The p-value of the F-test of joint hypothesis for LCCs and interaction variables between LCCs and FSC dominant (a) and LCC dominant (C) groups is 0.0499, which is significant at the 5% level. The LCCs has a

Table 4 Regression estimates (Dependent variable: VFP) e LCCs^2 variable included. VFP

LCCs LCCs^2 Output Non-aero share International Connecting Cargo Runway utilization Aircraft size 2008 2009 2010 Constant R-squared

OLS

Random effect

Fixed effect

Coef.

Std. err.

Coef.

Std. err.

Coef.

Std. err.

0.5148** 0.3982** 0.0439 1.2466**

0.1872 0.1758 0.0353 0.1205

0.2773 0.3043 0.0546 0.1647

1.4859** 1.4903** 0.4490** 0.1660

0.3814 0.5475 0.1871 0.2626

0.2370 0.2438** 1.0411 0.0682**

0.1704 0.0979 0.9680 0.0363

0.5464** 0.4119 0.0373 0.7912 ** 0.4586** 0.1798 2.9305** 0.0872**

0.2074 0.1223 1.7067 0.0486

0.0823 0.2144 2.4026 0.1495

0.2484 0.2411 2.5926 0.0985

0.2144** 0.0099 0.0324 0.0274 0.4613 0.4805

0.0605 0.0352 0.0378 0.0356 0.4688

0.1390** 0.0156 0.0089 0.0027 0.6862 0.4230

0.0723 0.0140 0.0184 0.0195 0.7564

0.1272 0.0227** 0.0172 0.0073 2.1472 0.3526

0.1228 0.0119 0.0245 0.0273 1.4571

*The coefficient is significant at 10% level. **The coefficient is significant at 5% level.

Y.Y. Choo, T.H. Oum / Journal of Air Transport Management 33 (2013) 60e67

65

.85

Variable Factor Productivity .9 1 .95

1.05

Table 6 Stochastic frontier analysis e Tobit regression.

0

.2

.4 .6 Share of LCC Passengers

.8

1

Fig. 1. Effects of share of LCC on variable factor productivity (VFP).

statistically significant negative coefficient in all of the three regressions. Consistent to a priori expectation, the coefficients for both the interaction variables between LCCs and category (a) (FSC dominant group), and category (c) (LCC dominant airports) are positive with the airport category (b) as the base, indicating that efficiency for airports that specialize in either LCCs or FSC are more efficient than airport category (b). However, the LCC coefficient is statistically significant only for the LCC dominant airports. 6. Robustness checks In addition to the VFP analysis (an index number approach) the SFA methodology discussed in the methodology section is used to generate the alternative efficiency scores as a means to check robustness of our empirical results. Our results indicate that the SFA efficiency scores have a high correlation with VFP scores (correlation coefficient of 0.97). Similarly as in the VFP regression analysis, a second stage regression was estimated on the SFA efficiency scores. However, since SFA efficiency score has the upper bound value of 1.0, we

VFP

Coef.

Std. err.

LCCs LCCs^2 Output Non-aero share International Connecting Cargo Runway utilization Aircraft size 2008 2009 2010 Constant Log-likehood value

0.2726** 0.2256* 0.4173** 0.6894** 0.1024 0.1659** 0.7685 0.0385** 0.1206** 0.0187 0.0222 0.0009 1.5342** 176.01

0.1235 0.1279 0.0221 0.0741 0.0987 0.0559 0.5432 0.0218 0.0385 0.0215 0.0217 0.0216 0.3187

*The coefficient is significant at 10% level. **The coefficient is significant at 5% level.

estimated Tobit regression model in order to avoid truncation bias. Table 6 presents the Tobit regression results on SFA scores. Table 6 shows that all explanatory variables have the same set of coefficient signs of the explanatory variables as in the VFP regressions. Even with the SFA efficiency estimates as the dependent variable, the coefficient of LCC presence variable has a statistically significant negative sign, confirming the negative effect of LCC presence on airport productivity. As in the VFP regressions, (LCC**2) variable has a positive sign indicating some evidence for the existence of economies of specialization of airport either in FSC services or in LCC services. Our sample includes six airports located in State of Texas. These six airports show great diversity in the share of LCC, ranging from 0.5% (IAH) to 97% (DAL). Therefore, it is interesting to compare the productivity measures of these six airports. Table 7 compares the productivity measures for Houston-Bush Intercontinental Airport (IAH), Dallas Fort-Worth International Airport (DFW), San Antonio International Airport (SAT), Austin Bergstrom Airport (AUS), William P. Hobby Airport in Houston (HOU), and Dallas Love Field (DAL) in 2010. The VFPs of the airports specializing in either FSCs (IAH and DFW) or LCCs (HOU and DAL) are relatively higher than the airports with almost equal FSC-LCC traffic mix (SAT and AUS). The same pattern is

Table 5 Regression estimates (Dependent variable: VFP) e inclusion of interaction variables-LCC*FSC/LCC dominant dummies. VFP

LCCs LCCs*FSC Dominant LCCS*LCC Dominant Output Non-aero share International Connecting Cargo Runway utilization Aircraft size 2008 2009 2010 Constant R-squared

OLS

Random effect

Fixed effect

Coef.

Std. err.

Coef.

Std. err.

Coef.

Std. err.

0.2979** 0.0498

0.0971 0.0960

0.2639** 0.0859

0.1192 0.0843

0.5783** 0.0139

0.2211 0.0897

0.2345**

0.0899

0.1510**

0.0793

0.2261**

0.0903

0.0435 1.2611** 0.2331 0.2549** 0.9047 0.0711**

0.0369 0.1240 0.1638 0.0914 0.9046 0.0366

0.0340 0.7898** 0.4328** 0.2086* 2.7853** 0.0966*

0.0592 0.1539 0.1808 0.1208 1.3842 0.0536

0.4673** 0.0764 0.0233 0.2312 2.1975 0.1446*

0.1491 0.2249 0.2433 0.1681 3.0410 0.0851

0.2066** 0.0085 0.0323 0.0292 0.5626 0.4888

0.0643 0.0360 0.0363 0.0361 0.5396

0.1392** 0.0148 0.0088 0.0053 0.8408 0.4301

0.0701 0.0147 0.0174 0.0185 0.8045

0.1180 0.0196 0.0196 0.0124 2.0976 0.3582

0.1090 0.0146 0.0203 0.0220 1.3254

*The coefficient is significant at 10% level. **The coefficient is significant at 5% level.

66

Y.Y. Choo, T.H. Oum / Journal of Air Transport Management 33 (2013) 60e67

Table 7 Comparison of productivity measures across airports in Texas, 2010 Airport

IAH

DFW

SAT

AUS

HOU

DAL

Share of LCC VFPa Pax/employee ATM/employee Non-aero/employee Output/employeea Pax/softcost ATM/softcost Non-aero/softcost Output/softcosta Revenue/expenses Non-aero rev/pax

0.005 0.892 35,168 456 90,814 0.74 0.377 0.005 0.972 0.901 1.747 2.582

0.020 0.909 30,899 354 151,118 0.822 0.285 0.003 1.395 0.863 1.394 4.891

0.406 0.774 17,935 256 80,463 0.489 0.321 0.005 1.438 0.993 1.164 4.411

0.447 0.816 25,199 322 119,114 0.684 0.265 0.003 1.251 0.818 1.369 4.724

0.941 0.838 24,051 369 80,514 0.581 0.36 0.006 1.204 0.988 1.412 3.348

0.969 1.16 44,936 649 142,045 1.018 0.436 0.006 1.379 1.124 1.548 3.161

IAH: Houston Bush International Airport; DFW: Dallas-Ft.Worth International Airport; SAT: San Antonio International Airport; AUS: Austin Bergstrom Airport. HOU: Houston William P. Hobby Airport; DAL: Dallas Love field Airport. a Indices are normalized at the respective US sample means.

also observed in all other partial productivity measures except nonaeronautical revenue per employee and non-aeronautical revenue per unit of soft cost input. In terms of efficiency in generating nonaeronautical revenue, the share of LCC traffic does not seem to be relevant for the case of these airports in Texas.

7. Conclusion Typically LCCs have utilized secondary airports in order to take advantage of slot availability and spare airport capacity, which is essential for quick aircraft turn-around as an important cost-saving strategy. However, over the last decade or so this business model has changed with more and more LCCs shifting their operations to major airport. For instance, JetBlue’s main base is in New York JFK Airport and Virgin America’s principal base of operations and sole hub is in San Francisco international Airport. Southwest presence at major airports such as Philadelphia, Pittsburgh, Denver, Atlanta and San Francisco has become more apparent. Although in the past these major airports were used largely by full service network carriers, some of these airports also started to attract the growth oriented LCCs. The surge of LCCs make it inevitable for these airport operators to adjust their business, facility design, and operational processes to satisfy the new requirement of LCC. Against this backdrop, this paper attempted to investigate the impact of LCC presence on the efficiency of major airports in U.S. Using an unbalanced panel data of 63 U.S. airports from 2007 to 2010, several alternative forms of OLS and panel regressions were estimated in order to discern the effects of varying degree of LCC presence in the airports on their operating efficiency using the Variable Factor Productivity (VFP) as the dependent variable. In

contrast to our intuition and the findings of previous research on this subject, the LCC presence (measured by the share of LCC passengers handled by an airport) is found to have a negative effect on operating efficiency of the major U.S. airports. Our further statistical investigation of the data and further regression analysis revealed the following intriguing nonmonotonic relationship between the percentage of LCC passengers in an airport and VFP scores, our operating efficiency indicator: (a) After removing the effects of all other factors, the airports specializing either in full service network carrier passengers or those specializing in serving LCC passengers achieve higher operating efficiency (net VFP) than the airports mixing both full service network carriers and LCCs passengers, in a substantial degree. (b) When a major airport previously served the full service network carrier passengers starts to attract LCC passengers, operating efficiency of the airport decreases with share of LCC passengers. This empirical finding suggests the existence of economies of specialization in airport management and operation. Certainly, mixing of full service network carriers and LCC services in an airport without careful examination of overall operating efficiency considerations does not appear to be a good idea. The airports around the world continue to evolve in response to the changes in market and business environment. Since the last decade, many primary airports have developed separate low cost or budget terminals in order to serve LCC passengers. Our research results calls for serious future investigation on the effects of such airport management practices on the operating efficiency of whole airport.

Appendix Table A1 List of United States airports included in sample. Airport code

Airport name

Airport code

Airport name

ABQ ALB ANC AUS BNA BOS BUR BWI CLE CLT CMH

Albuquerque International Sunport Albany International Airport Ted Stevens Anchorage International Airport Austin Bergstrom Airport Nashville International Airport Boston Logan International Airport Bob Hope Airport Baltimore Washington International Airport Cleveland-Hopkins International Airport Charlotte Douglas International Airport Port Columbus International Airport

MEM MIA MKE MSP MSY OAK OKC ONT ORD PBI PDX

Memphis International Airport Miami International Airport General Mitchell International Airport Minneapolis/St. Paul International Airport Louis Armstrong New Orleans International Airport Oakland International Airport Will Rogers World Airport Ontario International Airport Chicago O’Hare International Airport Palm Beach International Airport Portland International Airport

Y.Y. Choo, T.H. Oum / Journal of Air Transport Management 33 (2013) 60e67

67

Table A1 (continued ) Airport code

Airport name

Airport code

Airport name

CVG DAL DCA DEN DFW DTW EWR FLL HNL HOU IAD IAH IND JAX JFK LAS LAX LGA MCI MCO MDW

Cincinnati/Northern Kentucky International Airport Dallas Love Field Ronald Reagan Washington National Airport Denver International Airport Dallas Forth Worth International Airport Detroit Metropolitan Wayne County Airport Newark Liberty International Airport Fort Lauderdale Hollywood International Airport Honolulu International Airport William P. Hobby Airport Washington Dulles International Airport Houston-Bush Intercontinental Airport Indianapolis International Airport Jacksonville International Airport New York-John F. Kennedy International Airport Las Vegas McCarran International Airport Los Angeles International Airport LaGuardia International Airport Kansas City International Airport Orlando International Airport Chicago Midway Airport

PHL PHX PIT PVD RDU RIC RNO RSW SAN SAT SDF SEA SFO SJC SLC SMF SNA STL TPA TUS

Philadelphia International Airport Phoenix Sky Harbor International Airport Pittsburgh International Airport Theodore Francis Green State Airport Raleigh-Durham International Airport Richmond International Airport Reno/Tahoe International Airport Southwest Florida International Airport San Diego International Airport San Antonio International Airport Louisville International-Standiford Field Seattle-Tacoma International Airport San Francisco International Airport Norman Y. Mineta San José International Airport Salt Lake City International Airport Sacramento International Airport John Wayne Orange County Airport St. Louis-Lambert International Airport Tampa International Airport Tucson International Airport

References ACI, 2012. NA ACI Economic Bulletin March. Retrieved February 2, 2013, from. http://www.aci-na.org/sites/default/files/economicbulletin-march.pdf. ATRS (Air Transport Research Society), 2002-2012. The ATRS Global Airport Performance Benchmarking Report: Global Standards for Airport Excellence. Vancouver, British Columbia, Canada. www.atrsworld.org. Barbot, C., 2006. Low-cost airlines, secondary airports, and state aid: an economic assessment of the RyanaireCharleroi Airport agreement. Journal of Air Transport Management 12 (4), 197e203. Barrett, S., 2004. How do the demands for airport services differ between fullservice carriers and low-cost carriers? Journal of Air Transport Management 10 (1), 33e39. Ben Abda, M., Belobaba, P., Swelbar, W., 2012. Impacts of LCC growth on domestic traffic and fares at largest US airports. Journal of Air Transport Management 18, 21e25. Bottasso, A., Conti, M., Piga, C., 2012. Low-cost carriers and airports’ performance: empirical evidence from a panel of UK airports. Industrial and Corporate Change, 1e25. Castillo-Manzano, J., López-Valpuesta, L., González-Laxe, F., 2010. The effects of the LCC boom on the urban tourism fabric: the viewpoint of tourism managers. Tourism Management 32 (5), 1085e1095. Caves, D., Christensen, L., Diewert, E., 1982. Multilateral comparisons of output, input, and productivity using superlative index numbers. The Economic Journal 92 (365), 73e86. De Neufville, R., 2008. Low cost airports for low cost airlines: flexible design to manage the risks. Transportation Planning and Technology 31 (1), 35e68. Dennis, N., 2007. Stimulation or saturation? Perspectives on the European lowcoast airline market and prospects for growth. Journal of the Transportation Research Board, 52e59.

Francis, G., Fidato, A., Humphreys, I., 2003. Airport-airline interaction: the impact of low-cost carriers on two European airports. Journal of Air Transport Management 9 (4), 267e273. Gillen, D., Lall, A., 2004. Competitive advantage of low-cost carriers: some implications for airports. Journal of Air Transport Management 10 (1), 41e50. Graham, A., 2013. Understanding the low cost carrier and airport relationship: a critical analysis of the salient issues. Tourism Management 26, 66e76. Graham, A., Dennis, N., 2007. Airport traffic and financial performance: a UK and Ireland case study. Journal of Transport Geography 15 (3), 161e171. Hanaoka, S., Saraswati, B., 2011. Low cost airport terminal locations and configurations. Journal of Air Transport Management 17 (5), 314e319. Humphreys, I., Ison, S., Francis, G., 2006. A review of the airport-low cost airline relationship. Review of Network Economics 5 (4), 413e420. Lei, Z., Papatheodorou, A., 2010. Measuring the effect of low-cost carriers on regional airports` commercial revenue. Research in Transportation Economics 26 (1), 37e43. Liebert, V., Niemeier, H.M., 2013. A survey of empirical research on the productivity and efficiency measurement of airports. Journal of Transport Economics and Policy 47 (No. 2), 157e189. Lin, Z., Choo, Y.Y., Oum, T.H., 2013. Efficiency benchmarking of North American airports: comparative results of productivity index, data envelopment analysis and stochastic frontier analysis. Journal of the Transportation Research Forum 52 (1), 47e67. OAG Aviation, 2012. August Executive Summary. Retrieved January 28, 2013, from. http://www.oagaviation.com/OAG-FACTS/2012/August-Executive-Summary. Oum, T.H., Tretheway, M.W., Waters, W.G., 1992. Concepts, methods and purposes of productivity measurement in transportation. Transportation Research A 26A (6), 493e505. Warnock-Smith, D., Potter, A., 2005. An exploratory study into airport choice factors for European low-cost airlines. Journal of Air Transport Management 11, 388e 392.