1
UNIVERSITY OF AMSTERDAM
Spatial Competition between Airports for Passengers:
An Empirical Analysis of the German Market
Melissa Newham
Supervisor: Prof. Dr. Jo Seldeslachts
A thesis submitted in partial fulfillment for the degree of Master in Science in Economics
in the
Faculty of Economics and Business University of Amsterdam
2
Spatial Competition between Airports for Passengers:
An Empirical Analysis of the German Market
Melissa Newham
Abstract
This paper quantifies the degree to which airports in the same region “steal” or cannibalize each other’s passengers using a unique dataset covering 21 German airports over an eight-year time period (2005-2012). Spatial competition in the airport industry is a relevant topic given the liberalisation of the air transport industry and the significant entry of secondary and regional airports in the last two decades in Europe, combined with the locational interdependence among airports. However, quantitative research concerning the extent to which airports compete for passengers is limited. Applying instrumental variable methods, my main empirical results indicate that airport entry significantly reduces passengers at nearby airports by a magnitude of 9% on average for the German market. This “passenger stealing” effect has repercussions for economic efficiency; as airports have high fixed costs and benefit from economies of scale as passenger numbers increase, it may be more efficient to have fewer airports in the market with each airport serving more passengers.
3
Acknowledgements
This study would not have been possible without the support of many people. I would like thank my supervisor, Prof. Dr. Jo Seldeslachts, for helping me to improve on previous versions of this study. Many thanks also to Carina Lampe and Jochem Meeuwissen at PwC Amsterdam for their support and advice as well as providing me with data. I would like to thank the Infrastructure Advisory team at PwC Frankfurt including Georg Teichmann, Benjamin Schöne, Johannes Single and Theresa Brandt for providing valuable insights into the German airport industry. Many thanks also to Donagh Cagney at ACI EUROPE for granting me access to ACI publications and data.
4
1 Introduction
Traditionally, airports have been viewed as natural monopolies and objects of public infrastructure. However, liberalisation of the European air transport market starting in the 1980s and a trend towards airport privatisation introduced new dynamics into the industry and increased competition between airports (Starkie 2002). Arguably one of the most consequential events for the airport industry has been the emergence and growth of low-cost carriers (LCCs). The entry of low-cost carriers, operating point-to-point between airports, dismantled the established airport hierarchy and encouraged the development of small and medium sized secondary and regional airports in the European market (Dmitry 2012).
Regional airports entering the market for commercial aviation services are often prior military or general aviation airports. They are located at some distance from major cities and are less convenient than primary airports. Regional airports have been able to attract traffic by offering low airport charges to LCCs who are eager to cut costs wherever possible in order to keep their flight prices low. This strategy is effective as the passengers of LCCs are more willing to put up with an inconvenient airport location if it means saving money on flight fares. Europe now has a network of over 440 airports, the majority (60%) of which are small regional airports serving less than 1 million passengers per year. European passengers have significant airport choice with 63% of EU citizens living within a two-hour drive from at least two airports (European Commission 2014a).
The resulting density of airport infrastructure in Europe has increased competition in the market. In academic literature, competition between airports is widely discussed (notably Starkie 2002; Forsyth et al. 2010). Various lines along which airports may compete have been put forward, for example; large hub airports may compete for connecting traffic, airports may compete as cargo hubs, airports may compete for passengers with other modes of transport (high-speed rail, ferries etc.) and airports may compete for departure and/or arrival passengers (Tretheway and Kincaid 2005). An airport’s catchment area refers to the geographical zone containing potential passengers of the airport. The greatest scope for airport competition is where airports have overlapping catchment areas and hence compete spatially for passengers (Forsyth 2010). However, this spatial aspect of airport competition is also one of the most under-researched topics in scientific literature concerning airports (Dmitry 2012).
5
This study aims to quantify the extent to which market structure affects passenger figures at airports. This is achieved by conducting an empirical analysis of the determinants of passengers at an airport using a time series dataset covering 21 German airports over an eight-year time period (2005-2012).1 Germany provides an ideal case study given that the market houses numerous regional airports and there is data available on smaller airports. Specifically, the empirical model developed in this study seeks to answer the question; what is the impact of nearby airports on airport passenger traffic? And conversely; what would be the effect of airport exit on the passenger traffic of nearby airports? The answer to this question is not clear a priori. It depends on the substitutability between the airports, and whether the size of the market increases with the number of airports or if more airports simply split the “pie”. In theory, airport entry could increase the size of the market (passengers willing to travel by air) by taking passengers from other modes of transport or encouraging passengers to travel that otherwise would not have. On the other hand, airport entry could “steal” potential passengers from other airports.
My key finding is that neighbouring airports do exert a negative externality on each other in terms of cannibalizing each other’s passenger markets. Applying instrumental variable methods, I find that in the German market an additional nearby airport will reduce passengers at an airport by approximately 9% on average. The results of this study have implications for economic efficiency and overall welfare as given that airports have high fixed costs and decreasing average costs with passenger numbers, passengers play a crucial role in the viability and profitability of airports. In the context of Germany, where the majority of regional airports have excess capacity and rely on State aid to remain in the market, my results suggest that overall efficiency could be increased by reducing the number of airports in the market such that each airport would serve a greater number of passengers.
The remainder of the paper is structured as follows. Section 2 provides background information and a review of the relevant literature on spatial competition and airport competition in order to properly frame my contribution. Section 3 emphasizes relevant dynamics in the airport industry that a model of passenger traffic at airports will need to take into consideration, and introduces my methodology. Section 4 describes the dataset of German airports. Section 5
1
In addition, discussions were conducted with industry experts to better understand how the sector works in real life and to cross check the plausibility of the results.
6
describes the estimation strategy employed. Section 6 presents empirical results. Section 7 outlines the efficiency and policy implications of my research for the German market. Section 8 concludes.
2 Background and Related Literature
Spatial competition deals with dispersed markets in which each firm faces only a few rivals in its immediate neighbourhood, further away there are more competitors however their influence is reduced by transportation costs (Gabszewicz et al. 2001). In a dispersed market “the market is commonly subdivided into regions within each of which one seller is in a quasi-monopolistic position” (Hotelling 1990, p. 41). Recently models of spatial competition have been applied to gas stations, movie theatres, hospitals and retail stores, but the airport industry is still weakly covered (Dmitry 2012).
Spatial competition is relevant to the airport industry given the dispersed market structure of the sector and the immobile nature of airports. Once an airport is established in a certain location it cannot move and the airport has to rely on the flexibility of its customers. Alternative airports that are able to serve the same potential passenger market are relevant competitors. In a spatial sense, airports will compete with rivals both as a departure point for local populations and as a destination point for leisure and business travelers.
In the literature, an airport’s catchment area is the term used to describe the airport’s market area, that is, the geographical region from which the airport gets the bulk of its passenger traffic. Catchment areas can be defined in a number of different ways; by geographical distance, by travel time or by travel cost. One view of competition for passengers is that airports compete most intensely for the proportion of passengers located on the outskirts of their catchment areas; in the regions where catchment areas overlap. As passengers incur travel costs to reach the airport, an airport is most attractive to the passengers located closest to the airport. Competition analysis based on overlapping catchment areas was initially put forward by Starkie (2002) and then applied in subsequent studies (see Starkie 2009).
However, viewing the relative attractiveness of an airport as based solely on travel time or travel costs neglects the fact that airport choice also depends on other airport characteristics such as flight prices, the variety of airlines operating at the airport, the destinations offered, flight
7
frequencies and the convenience of airport facilities e.g. parking lots, shops etc. (see Hess and Polak 2005; Zhang and Xie 2005; Pantazis and Liefner 2006; Martínez-Garcia, Ferrer-Rosell and Coenders 2012). Taking into consideration multiple airport characteristics, previous studies have attempted to assess the substitutability of airports in a region and the level of competitive pressure by constructing weighted indexes and substitution coefficients (see Strobach 2006; Malina 2006).2
The influence of airlines on a passenger’s choice of airport has also been recognized by recent literature that considers airports within a two-sided market framework (see Ivaldi et al. 2012; Gillen and Mantin 2013). Two-sided markets are broadly defined as markets in which an intermediary or platform enables interactions between two (or more) different types of end-users, and tries to get both sides of the market “on board” by deciding on an appropriate pricing scheme (Rochet and Tirole 2006). Standard examples of two-sided markets include magazines, newspapers, credit cards, shopping malls and radio stations.
Airports can be viewed as platforms bringing passengers and airlines together as they internalize the network externalities arising from two demands; passengers are better off if there are more airlines at the airport, and airlines are better off if there are more passengers that are willing to use the airport. As airports earn revenues from both sides of the market, they can cross-subsidize airport charges with non-aviation revenues in order to attract more airlines, which in turn brings in more passengers.3 An empirical study by Ivaldi et al. (2012), using data on US airports finds that flight frequencies at an airport significantly influence passengers’ demand for the airport. However, their model does not account for airport competition as it assumes that the airport is a monopoly platform in which airlines and passengers join to interact.
The effect of airport competition on airport prices, as opposed to passenger figures, has also be considered in the literature. There is mixed evidence on whether nearby airports reduce
2 Malina (2006) suggests a substitution coefficient which is defined as “the proportion of inhabitants within the
relevant regional market of an airport that consider another airport, which has been identified as meeting the demands of the airlines, to be a good substitute from their perspective as well” (Malina 2006, p. 2).
3 Modern airports generate revenues from both aviation and non-aviation (or commercial) activities. Non-aviation
activities include catering, retail, hotels, car rental, sale of advertising space, conference facilities, car parking, and activities related to commercial property. These activities benefit from increased traffic to and from the airport.
8
the market power of airports in setting airport charges. 4 In an empirical study of US airports, Van Dender (2007) finds that aviation revenues per passenger are negatively related to the number of nearby airports. In a European context, Bel and Fageda (2010) and Bilotkach et al. (2012) undertake empirical studies of airport charges. Whereas Bel and Fageda (2010) find that airport charges are negatively related to the number of nearby airports, Bilotkach et al. (2012) find that the presence of nearby airports does not significantly affect airport charges.
Another popular approach to estimating the intensity of competition between airports is interviews with experts and airport management (see Borins and Advani 2002; Gillen and Niemeier 2006). This approach is useful for an initial analysis of competition in the market, but faces problems of subjectivity as well as quantitative measurement. Hence, while the literature provides a rich discussion of which variables are likely to affect airport competition and market power, there remains limited empirical analysis on the topic. To the best of my knowledge, no studies have sought to quantify the extent to which spatial competition between airports affects passenger numbers. In this paper I aim to address this gap in the literature as well as to apply my results to provide policy recommendations for airport planning and management.
3 Methodology
The methodology of this study is to develop an empirical model for passenger traffic at airports in order to measure the extent to which airport competition affects passengers. The primary explanatory variable of interest is thus airport competition. However, an important additional determinant of passengers, as indicated by the previous section, is the airlines and flights available at the airport. Furthermore, the model will need to control for other variables that could affect passenger traffic such as the wealth and size of the population in the airport’s catchment area. It is necessary to consider from the outset where issues of endogeneity may arise in such a model, in order to deal with these concerns appropriately when implementing the econometric estimation.
4 Airport charges refers to a group of aviation-related sub-charges including landing fee, passenger fee, security fee,
infrastructure fee, terminal navigation fee, noise fee, parking fee and airbridge fee. Airport charges are typically subject to regulation. Regulation may be cost-based or incentive-based.
9
A typical assumption of empirical models that deal with spatial competition is that resources tend to cluster in the most lucrative markets. Where this assumption holds true, the number of firms in a market provides an indication of the profitability of the market; with larger markets being able to accommodate more firms (Berry and Waldfogel 1999; Davis 2006). This reasoning implies that the spatial dispersion of firms cannot be treated as exogenous. Hence modelling the entry decision of the firm is typically a necessary first step in the analysis. However, while this logic may apply to other dispersed markets such as retail stores, it cannot be easily applied to the case of the German airport industry.
Secondary research as well as interviews conducted with experts in the German airport industry indicate that airport entry is driven primarily by political will, and not profitability prospects. In Germany, the states or Bundesländer are responsible for the planning of airport projects. Regional authorities base decisions about the expansion of existing airports and the commencement of commercial operations on beliefs about the airport’s potential to generate tourism, employment and economic growth in the region (Heymann et al. 2005; European Commission 2014b). Moreover, the idea of having a local airport appeals to politicians partly for prestige reasons. Local authorities also tend to operate independently of one another in terms of airport management and planning, with state lines serving as political walls to cooperation.
As a result of the political, rather than profit, motives for airport development the majority of regional airports in Germany are loss-making and rely on State aid. Whereas in a typical market where loss-making firms would exit, exit from the airport industry can be avoided by sizeable subsidies funded by local taxpayers. Avoiding airport closure is appealing to politicians who are hesitant to admit poor decision making (Heymann et al. 2005). Thus specificities of the German airport market imply that the spatial dispersion of airports is not related to the profitability of the markets in which they are located. Hence, I argue that airport competition can be considered as an exogenous variable in an empirical model of passenger traffic. I will later bolster this claim with statistical evidence showing low correlations between airport density in Germany and market characteristics such as GDP per capita and population density.
An issue of endogeneity does however arise with regards to controlling for airlines at the airport. As explained previously, passengers are attracted to airports based on characteristics associated with airline services such as flight fares and flight frequencies. Similarly, airlines are
10
attracted to an airport based on the number of passengers that are willing to use the airport. This dynamic means that causality flows both from airlines to passengers, and from passengers to airlines. In this situation, there is reverse causality and the regression coefficients will be biased unless the problem of endogeneity is appropriately addressed. In order to deal with this problem of endogeneity I will apply instrumental variable methods. The statistical details of the estimation strategy employed are provided in section 5.
4 Data
The full dataset consists of an unbalanced panel of 21 airports in Germany with observations running from 2005 to 2012. Table 5 in Appendix 1 lists the airports in my sample and their corresponding passenger figures for 2012. There is substantial variation in the size of airports ranging from Frankfurt with over 50 million passengers per year to Erfurt with less than 200 000 passengers per year. The three airports in Berlin have been treated as one airport given that they are under common ownership and hence are not competing with one another. The data has been collected and compiled from three main sources: the German Airport Association (ADV), Eurostat and the airports’ annual financial reports. The dataset contains information on each airport’s annual operating results including passenger figures, total aircraft movements, cargo volumes, operating revenues, operating costs and profits. Additionally there is information on the airport’s catchment area, nearby airports, the market share of low-cost carriers at the airport and airport ownership.
In my dataset, the variable Nearby serves as the measure of competition faced by an airport. Nearby provides a count the number of alternative airports offering commercial services that exist in the airport’s catchment area. In this study an airport’s catchment area is defined as a circle of 100km radius around the airport. Airports are considered nearby if they are within a 100km straight-line distance from the airport and not more than 1 hour and 30 minutes travel time by car or train. The implicit assumption here is that passengers do not perceive airports that are more than a 1 hour and 30 minute drive away from each other as being substitutable.
In the context of spatial competition, defining the airport’s catchment area appropriately is crucial. As mentioned previously, catchment areas can be defined in a number of different ways; by geographical distance, by travel time or by travel cost. Applying a catchment area of 100km
11
radius is standard practice in the literature and is the distance applied by the European Commission in their definition of an airport’s catchment area (European Commission 2014b). However, this is a necessary simplification as in practice defining the catchment area should take into consideration the specificities of the airport, and catchment areas will vary in size and shape from airport to airport.
Figure 1 provides a map of the airports in Germany with circles of 100km radius indicated around the airports that are in the sample. Germany has one of the densest airport networks in Europe and the map indicates that there is substantial overlap of catchment areas in the German market. There has been one airport entry and one airport exit in the market affecting the airports in the sample during the time period considered. In 2007 Memmingen Airport in the state of Bayern entered the market and in 2010 Leipzig-Altenburg Airport in the state of Sachsen exited the market.
FIGURE 1 Map of Airport Catchment Areas
12
For each airport, data on passengers, cargo (in tonnes), total aircraft movements and the market share attributable to low-cost carriers at the airport are taken from the German Airport Association (ADV) website. Low-cost carriers refer to airlines with low fares, general availability and direct sales via the internet. However, as airlines broadly defined as low-cost can vary significantly in their business models, there is some discretion involved when assigning an airline to the low-cost segment. A list of which airlines are considered as low-cost by ADV is thus provided in Appendix 2. Low-cost carriers have established themselves as important players in the German airline industry after expanding significantly in 2009. Analysis of the data reveals that the average share of LCCs at any given airport for 2005 was 25%. In 2010 this figure was up to 57% where it has stabilized.
Data on the economic, infrastructural and demographic characteristics of an airport’s catchment area are obtained from Eurostat’s Regional Statistics Database. Variables include Tourism, Population density, Railways and GDP per capita. Ideally these characteristics should correspond to the 100km radius catchment area. However as data at this level is not readily obtainable, I use available data at the regional level. The “catchment area characteristics” associated with an airport thus correspond to those of the NUTS1 region (as defined by Eurostat) in which the airport is located. In the case of Germany, the NUTS1 regions are the states or Bundesländer. The variable Tourism is the number of nights spent by non-residents in tourist accommodation per year, divided by the area of the NUTS1 region. Railways is the total railways in kms per 1000km2 in the NUTS1 region in which the airport is located.
In the dataset, the variable Public-Private is a dummy variable equal to 1 for airports that are public-private partnerships, and zero for airports that are 100% publically owned. In 1982, the Federal Government first announced a program to privatize airports given budget restrictions, however since then few German airports have been privatized. The majority of the airports in the sample are 100% publicly owned. Ownership is typically shared between the Bundesland (state), Kreis (county) and Stadt (city). Only six airports in the sample are currently public-private partnerships.5
Finally, information on the airport’s annual financial performance in terms of operating revenues and operating costs is obtained from the airports’ annual financial reports. The reporting style is
13
standardized across German airports such that the results are comparable. Both financial variables are in thousands of euros.
TABLE 1 Descriptive Statistics
Table 1 above illustrates the pertinent descriptive statistics for the variables employed in my empirical analysis. It is evident that competition from nearby airports varies substantially across the sample with a maximum of six nearby airports and a minimum of zero. Large standard deviations relative to the mean for passengers and aircraft movements indicate that there is substantial variability in the traffic received by the airports in the sample. Table 1 indicates that some airports accommodate only low-cost carriers, with an LCC market share of 100%, while other airports have had years in which there have been no low-cost movements. There is also notable variability in cargo figures. Certain airports, for example Hahn and Leipzig-Halle, have pursued a strategy of specializing in cargo services. This implies that the largest airports in passengers terms are not necessarily the largest airports in cargo terms.
Graphical analysis of passenger, aircraft movement, cargo and revenue trends over time, conducted for each airport, indicate that there is also substantial variability in the variables across time for each airport. This volatility is typical of the airport industry where a new contract with
Variable Mean Std. Dev. Min Max
Passengers 9 059 363 13 382 900 175 864 57 274 099 Nearby 2.1 1.6 0 6 Cargo 194 413.8 480 040.9 0 2 275 106 Aircraft movements 111 405.8 129 895.7 6 419 485 915 LCC share 43.7 28.9 0 100 GDP per capita 30 866.2 6 738 17 899 52 555 Population density 662 883.2 137 3 951.5 Tourism 688.8 1 800.4 31.4 11 865.1 Railways 182 141 84 708 Public-Private 0.3 0.5 0 1 Operating revenues 293 228.4 534 353.2 2 253.3 2 473 300 Operating costs 210 271.4 374 521.3 2 248.5 1 772 299
14
an airline or the launch of a new route can have a dramatic effect on passengers, and congestion at hub airports will divert cargo to regional airports temporarily. For the entire German market, passenger figures have followed an upward trend over the years 2005 to 2012, with a dip occurring in 2009 at the time of financial crisis.
5 Estimation
In this section I provide the econometric specification of my primary empirical model and motivate the estimation strategy employed. The aim of the model is to determine the effect of airport competition on passenger traffic at airports. The dependent variable is the log of passengers. The key explanatory variable of interest is Nearby. Additional variables are included in the model as controls. The primary estimating equation for the passengers at airport i in year t is as follows:
( )
( ) ( ) ( )
( )
where are year dummies and is the error term.
Nearby gives the number of the alternative airports in the airport’s catchment area, hence the coefficient on Nearby measures the impact of one additional competing airport on passengers at the airport. If neighbouring airports do exert a negative externality on each other in terms of cannibalizing each other’s passenger markets then the coefficient on Nearby is expected to be negative and significant. As mentioned previously, I argue that Nearby can be treated as an exogenous variable as airport entry in Germany is linked to political motives rather than the size and profitability of the market. Moreover, the lack of relationship between airport density and the size and profitability of the passenger market is evidenced statistically in Table 2 below which shows weak correlations between Nearby and GDP per capita, Population density and Tourism.
15
TABLE 2 Matrix of Correlations for Catchment Area Characteristics
Given that passengers at the airport will be influenced by airlines at the airport and vice versa, an important additional explanatory variable is the services offered by airlines at the airport. To control for airline offering at the airport I use the log of low-cost carrier movements and the log of non-low-cost carrier movements at the airport. Low-cost carrier movements (and correspondingly non-low-cost carrier movements) are calculated by multiplying total annual aircraft movements by the low-cost carrier market share at the airport (as provided by ADV). Subject to data availability, I argue that together these variables provide a good measure of airline offering at the airport in terms of flight frequency and ticket prices.
A statistical problem that arises when using the log of low-cost movements and the log of non-low-cost movements is that there are a few observations where low-cost or non-low-cost movements equal zero. Logarithmic transformations of these zero observations is not possible, and omitting them is problematic as it would be non-random. Hu (1972) examines two approaches to handling the problem; first, a positive constant can be added to all sample values such that all observations become positive and logarithmic transformation is possible, or second, only the zero values are replaced by a small positive constant before taking logarithms. Hu (1972) shows that the second approach is preferable and yields parameter estimates with smaller deviations from their true value than the first approach. Hence in all specifications I apply the second approach of adding a small positive constant to the zero sample values for low-cost and non-low-cost movements before applying the logarithmic transformation.6
6 Given that the minimum value of aircraft movements in the sample is 6419, I use 500 as the value of the small
positive constant.
Nearby GDP per capita Population density Tourism Railways
Nearby 1
GDP per capita 0.022 1
Population density -0.140 0.342 1
Tourism -0.269 0.111 0.907 1
16
As discussed in section 3, a central econometric concern to the estimation of passengers is the simultaneous determination of airline services and passengers. In order to avoid the problem of endogeneity I use instrumental variable (IV) methods and instrument the log of low-cost movements and the log non-low-cost movements. Suitable instruments need to meet two conditions. Firstly, they should be correlated with airline services (instrument relevance). Secondly, they should only affect passengers through their effect on airline offering (instrument exogeneity). Hence, I identify variables that may affect contract negotiations with airlines and that would attract airlines, but not passengers directly, to the airport.
Potential instruments identified include cargo, airport ownership, airport operating costs per passenger and airport operating revenues per passenger. Cargo figures provide an indication of the cargo services available at the airport which would attract airlines but not passengers directly. Whether or not the airport is publically owned or a public-private partnership may affect the pricing and management strategy of the airport which could attract airlines without changing the underlying passenger demand. Previous research indicates that private airports are on average more efficient than publically owned airports (Oum, Yan and Yu 2008; Camargo 2013). Additionally, I include the airport’s operating costs per passenger in the instrument set as a further measure of airport efficiency. In my dataset, operating revenues per passenger provides the closest proxy measure of airport charges and incentive schemes offered by the airport.7 As passengers are primarily attracted to an airport by the price of the flight, which is determined in connection with airline competition and numerous other variables, airport charges are not likely to impact passengers directly. However, they do play an important role in attracting airlines to the airport. In particular, low-cost carriers are attracted to airports that are willing to offer low airport charges and discount schemes (Lei and Papatheodorou 2010; Graham 2013).
In preliminary regressions I apply Two-Stage Least Squares (2SLS) regression and instrument low-cost and non-low-cost movements with all potential instruments: cargo, ownership, operating costs per passenger and operating revenues per passenger. However, I find that the ownership dummy Public-Private is not significant in the first stage regressions and post-estimation statistical tests indicate that the inclusion of this variable leads to an invalid instrument
7
I use operating revenues per passenger as while airports levy charges both at the airline and passenger level, the majority of charges are at the passenger level. As ACI reports, “[In 2011] passenger-related charges accounted for 67% of total aeronautical revenues, while airline-related charges represented only 33%” (ACI 2012, p. 11).
17
set. Hence, in all final specifications I employ Two-Stage Least Squares regression and instrument for the log of low-cost and the log of non-low-cost movements with cargo, operating revenues per passenger and operating costs per passenger.
In addition to airline offering I need to control for the economic, demographic and infrastructural characteristics of the airports’ catchment area. Table 2 indicates that there are extremely high correlations between the variables Population density, Tourism and Railways. Population density has a 0.9 correlation with Tourism and a 0.98 correlation with Railways. If all three catchment area variables are included the model will suffer from multicollinearity. Whereas Population density and Tourism are expected to have a positive impact on passengers, the effect of Railways is less clear. Travel by train provides a substitute for air travel particularly in the case of short-haul inter-Europe and inter-Germany flights which is primarily what regional airports offer. This reasoning would suggest that the presence of more rail connections could have a negative impact on passengers. On the other hand, good railway connections could also increase an airport’s catchment area if it increases the accessibility of the airport for passengers residing further away from the airport (Starkie 2002).
In the primary regression (specification 1) I use the log of GDP per capita and the log of Population density as controls for the economic and demographic characteristics of the catchment area that are likely to influence passenger traffic. To test the robustness of the model, in specification 2 I use the log Tourism instead of log of Population density and in specification 3 I use log of Railways instead of log of Population density. As an additional robustness check, specification 4 includes a linear time trend as opposed to year dummies.
The final estimation of passengers for all specifications is run on the sample of airports excluding the two major hub airports Frankfurt and Munich. Hub airports are markedly different from regional airports in that they have vastly higher passenger numbers, they are often congested, they offer many intercontinental flights and a large proportion of their traffic is derived from transfer passengers.8 Moreover Frankfurt and Munich face significantly less competition for airlines and passengers as they serve as bases for Lufthansa, the largest airline in Europe. Hence, while regional airports near to hub airports are likely to compete with the hub,
8
In 2011, 54% and 39% of traffic at Frankfurt and Munich airport respectively was transfer traffic (www.frankfurt-airport.com 2011; www.munich-airport.de 2011).
18
hub airports themselves are unlikely to be in direct competition with smaller airports, and Nearby is unlikely to explain their passenger figures. I include Düsseldorf, the smallest of Lufthansa’s three bases, in the final sample, adding the indicator Lufthansa base to account for the potentially differing competitive dynamics at this airport. In all estimations the standard errors are corrected for the problem of heteroscedasticity arising from a non-constant variance across observations.
6 Results
Table 3 reports the estimation results for the four model specifications. Before discussing the variables of primary interest, I will discuss the adequacy of the model. Table 4 reports the relevant post-estimation test statistics for all specifications. Firstly, I test the endogeneity of low-cost and non-low-low-cost aircraft movements. In the case of IV-2SLS robust estimation the test for endogeneity gives the robust score chi2 statistic. For each specification I can reject the null hypothesis that low-cost and non-low-cost aircraft movements are exogenous at the 0.01 level. The endogeneity of these variables confirms that instrumental variable methods is an appropriate estimation strategy for a model of passenger traffic at airports.
Secondly, I test the appropriateness of the instruments. In order to be valid, instruments need to be both exogenous and relevant. Instrument exogeneity cannot be directly tested for however it is standard practice to confirm that instruments do not contradict one another by applying a test of overidentifying restrictions. In the case of IV-2SLS robust estimation the test for overidentifying restrictions gives Wooldridge’s robust score and a corresponding p-value. The null hypothesis of the test is that the instruments used are valid. A non-significant score implies that this hypothesis cannot be rejected. In the case of all four specifications, Wooldridge’s robust score is non-significant and hence the instruments used are non-contradicting and valid. The criterion of instrument relevance is assessed using the first-stage F-statistic, R2 and Shea’s partial R2.9 For all specifications the instruments appear to be highly relevant with significant F-statistics comfortably above 10.10 The R2 and Shea’s partial R2 from the first-stage regressions also provide
9 Shea’s partial R2 measures the correlation between the endogenous variables and the additional instruments after
partialling out the effect of other included variables. Unlike the R2 statistic, Shea’s partial R2 will not be inflated because of correlations between aircraft movements and catchment area characteristics.
10
Stock, Wright, and Yogo (2002) suggest that the F statistic should exceed 10 for inference based on the 2SLS estimator to be reliable.
19
confirmation that the instruments used are relevant. In sum, the regression model passes the required diagnostics and is well specified.
TABLE 3 Estimation Results
Dep var: ln(Passengers) 1 2 3 4
IV-2SLS IV-2SLS IV-2SLS IV-2SLS
ln(LCC movements) 0.890*** 0.898*** 0.888*** 0.890*** (0.050) (0.058) (0.049) (0.051) ln(Non-LCC movements) 0.142* 0.145 0.142* 0.147* (0.081) (0.088) (0.083) (0.087) Nearby -0.094** -0.095** -0.093** -0.092** (0.040) (0.040) (0.039) (0.041) Lufthansa base 0.605** 0.589** 0.607** 0.595** (0.247) (0.269) (0.254) (0.260) ln(GDP per capita) 0.110 0.171 0.096 0.060 (0.198) (0.215) (0.202) (0.209) ln(Population density) 0.007 0.010 (0.069) (0.072) ln(Tourism) -0.018 (0.043) ln(Railways) 0.029 (0.100) Time -0.131*** (0.023) Constant 3.515* 2.914 3.577* 4.854** (1.894) (1.970) (1.849) (1.961) R2 0.886 0.884 0.887 0.876 N 126 126 126 126
Notes: Standard errors are contained in parentheses, significance levels: *0.1 **0.05 ***0.01, year dummies included in specifications 1-3 but coefficients are not reported, sample excludes Frankfurt and Munich airport
20
The most important result for the purposes of this study relates to the estimated coefficient on the Nearby variable. Across all specifications the effect of a nearby airport on passengers is negative and significant at the 0.05 level. The magnitude of the effect is also very consistent across specifications. The log-linear structure of the model with regards to Nearby implies that the coefficient on Nearby in specification 1 should be interpreted as follows; an additional airport within an airport’s catchment area is predicted to reduce passengers at the airport by a magnitude of 9.4% on average, holding all other factors constant. Conversely, airport exit is predicted to increase the passengers of nearby airports by 9.4% on average. In short, the empirical results support the hypothesis that airports exert a negative externality on each other in terms of cannibalizing each other’s passenger markets.
The estimated coefficients for the included control variables all have the expected signs. Increased airline services at the airport in terms of low-cost and non-low-cost movements has a positive and significant effect on passenger traffic. Interestingly, the results indicate that low-cost airlines, as opposed to non-low-cost airlines, are the key drivers of passenger traffic for the airports in the sample. A 10% increase in low-cost carrier movements increases passengers by approximately 8.9% on average. While the effect non-low-cost airlines on passengers is also positive it is much smaller and less significant. This result is in line with qualitative research and case studies that find that low-cost airlines expand the passenger markets of regional airports; drawing in passengers from further away and encouraging passengers to fly that otherwise would not have (Pantazis and Liefner 2006; Graham 2013).
In the primary model the control variables for the economic and demographic characteristics of the airport’s catchment area, GDP per capita and Population density, are positive but not significant. These variables do however have significant and positive effect in the first stage regressions for low-cost and non-low-cost aircraft movements. In the first stage regression for non-low-cost aircraft movements a 1% increase in GDP per capita in the airport’s catchment area is expected to increase aircraft movements by 1.15% and a 1% increase in population density is expected to increase aircraft movements by 0.8%. These effects are both significant at the 1% level. Similarly in the first stage regression of specification 2, Tourism has a positive and significant effect on aircraft movements. This suggests that the effect of economic and demographic catchment area characteristics on passengers traffic is primarily an indirect
21
effect which operates through airlines. Wealth, population density and tourism in a region attract airlines, which in turn attract passengers to the airport. Hence these variables are not found to be significant after controlling for airlines at the airport.
The inclusion of Railways in place of Population density in specification 3 has little effect on the size and significance of the coefficient on Nearby. The effect of a denser railway network is positive but also not significant, providing no clear evidence with regards to whether or not railways compete with airports for passengers or expand the airport’s potential market. In specification 4, where a linear time trend is included instead of year dummies, there is also little change in the magnitude and significance of the estimated coefficients. In sum, the empirical model appears to be valid and robust, and the results clearly indicate that neighbouring airports in Germany do “steal” each other’s passengers to a moderate extent.
TABLE 4 Post-Estimation Tests
1 2 3 4
Endogeneity:
Robust score chi2 28.8404*** 28.3898*** 29.0978*** 28.5006***
Overindetifying restrictions:
Wooldridge’s robust score test 2.16302 2.27088 2.10895 1.5282
P-value 0.1414 0.1318 0.1464 0.2164
Relevance:
First stage F-stat
ln(Low-cost movements) 21.06*** 25.29*** 20.76*** 37.67*** ln(Non-low-cost movements) 25.07*** 21.50*** 23.58*** 50.27*** First stage R2 ln(Low-cost movements) 0.7133 0.7363 0.7040 0.7030 ln(Non-low-cost movements) 0.4780 0.4056 0.4671 0.4666 Shea's partial R2 ln(Low-cost movements) 0.5740 0.5061 0.5745 0.5655 ln(Non-low-cost movements) 0.1004 0.0808 0.0958 0.0952
22
7 Policy Implications
This section outlines an implication of my research with regards to economic efficiency, and accordingly provides policy suggestions for airport management and planning in Germany. Whereas economics typically dictates that greater competition leads to lower consumer prices and higher overall efficiency, this is not necessarily the case if there is a duplication of fixed costs (Suzumura and Kiyono 1987). Excessive entry into a market can result when two conditions hold; 1) average costs are decreasing in output and 2) entrants offer a substitute good such that entry “steals business” from competitors (Berry and Waldfogel 1999). In such a market it may be more efficient to have fewer firms with each firm producing are greater level of output.
Airports face high fixed costs and economies of scale. Fixed costs comprise of capital costs as well as costs driven by regulatory requirements such as safety and security operations which must be in place regardless of the number of passengers e.g. firefighting services. It is commonly held that airports need to reach a breakeven or threshold number of passengers in order to be viable (Doganis and Thompson 1974; ACI 2012; European Commission 2014b). Estimates of the breakeven point for airports range from 1 million passengers per annum (Doganis and Thompson 1974) to 0.5-2 million passengers per annum (Heymann et al. 2005). Analysis of the German dataset used in this study indicates that airports typically reach operating breakeven (where operating revenues cover operating expenses) at 0.2 million passengers per annum and full breakeven (where revenues also cover capital expenses) at approximately 4 million passengers per annum.11 Hence reaching a critical mass of passengers is essential to the profitability and viability of an airport.
The results of the present study bring about concerns of excessive entry in the German airport industry as the empirical analysis undertaken indicates that neighbouring airports do “steal business” from each other. As airports benefit from spreading high fixed costs over a larger passenger base, airport competition has the potential to result in inefficiencies where existing airports already have excess capacity. Of the airports in the sample 13 German airports have less
11 Given available data on airport profits and costs in the German dataset, the breakeven point for the German market
is estimated by regressing earnings before income, tax, depreciation and amortization (EBITDA) on passenger figures, and regressing capital costs on passenger figures. The operating breakeven point is taken to be the number of passengers at which the estimated EBITDA curve equals zero. Overall breakeven is taken to be point at which the estimated EBITDA and capital cost curves intersect.
23
than 4 million passengers per annum. Erfurt Airport has less than 0.2 million passengers per year. The majority of these smaller airports have excess capacity and are loss-making.
One clear example of inefficiency in the market is the case of Saarbrücken Airport and Zweibrücken Airport. Saarbrücken Airport is a small international airport in the state of Saarland and typically serves around 400 000 passenger per year. At a distance of only 39km from Saarbrücken Airport is Zweibrücken Airport which serves around 200 000 passengers per year. Both airports currently operate at a loss and are competing for a similar passenger market. In 2001, the state governments of Saarland and Rhineland-Palatinate issued a study to assess the possibilities for cooperation between the airports. However, political difficulties have thus far prevented any concrete efforts to cooperate. Another example of inefficiency is Germany’s newest regional airport, Kassel-Calden, which officially opened on 4 April 2013. Kassel-Calden Airport is the pride of the Hesse state government and cost approximately 270 million euros to build, however the airport has struggled immensely to attract passengers and planned flights have had to be cancelled due to low demand (Deutsche Welle 2013). The airport faces competition from Paderborn Airport which can be reached in 50 minutes by car and Dortmund Airport which can be reached in 1 hour and 25 minutes. Eurocontrol (2013) forecasts limited passenger growth for mature markets such as Germany in the future, hence excess capacity in the market can only be reduced by airport exits.
Due to sizeable subsidies awarded at the regional level, loss-making airports do not exit the market and excess capacity is sustained. Under the current structure of German air transport policy, responsibilities such the approval, planning, construction and operation of airports are delegated to regional authorities. Regional authorities support loss-making airports with subsidies as airports are believed to boost growth and development in the area, and for other political reasons. However, while it is possible that the region enjoys economic gains as a consequence of the airport, the county as a whole is likely to lose, as economic activity is simply shifted and not necessarily increased overall (Forsyth 2006). Overall welfare is reduced through the duplication of fixed costs as well as inefficiencies stemming from subsidies in and of themselves including managerial slack, competition distortions between airports and airlines and subsidy wars to attract low-cost airlines which direct taxpayers money away from more productive uses (Forsyth 2006; European Commission 2014b).
24
Overall economic welfare could be increased if excess capacity in the market were reduced by airport exits, however this will require an optimization of airport infrastructure at the federal level. The current system of airport management increases the influence of local interest groups and gives rise to political particularism where politicians further their own interests at the expense of the overall economic welfare. Whereas in other regards competitive federalism may be beneficial, this is not the case with regards to airport management. Management at the regional level gives politicians leeway to open and fund airports that are unlikely to reach the critical mass of passengers required to become profitable, and concurrently cannibalize the passenger markets of nearby airports, preventing surrounding airports from fully exploiting economies of scale. If airport management were to take place at the federal level, the airport market structure could be optimized in order to ensure that regional and secondary airports are able to operate at a higher capacity and can be maintained without excessive subsidies. This would free up taxpayers money for more productive uses and would allow airports to benefit from economies of scale.
A potential downside associated with airport exit is that airport charges may rise above the optimal level if there is reduced competition in the market. However, empirical research is unresolved on whether nearby airports do actually exert pricing pressure on each other (see Bel and Fageda 2010; Bilotkach et al. 2012). Price competition between airports is complicated by the regulation of airport charges (Forsyth et al. 2010). All German airports are subject to regulation. The majority of airports are subject to traditional cost-based regulation with a single-till approach (ICAO, 2013).12 If airport charges continue to be subject to cost-based regulation, the downward pressure on prices stemming from greater efficiency due to economies of scale is likely to outweigh any upward pressure from less competition.
8 Conclusion
The significant entry of regional and secondary airports in the last two decades, together with the immovable nature of airports and overlapping catchment areas provide a good context for an analysis of spatial competition in the industry. However, despite numerous studies on airport competition and market power, quantitative research concerning spatial competition
12 Airports can be operated on a single-till or dual-till basis; under the single-till approach both aviation and
non-aviation revenues are considered when setting airport charges, under the dual-till approach charges are determined solely on the basis of aviation activities.
25
between airport for passengers is limited. Hence the present paper provides a valuable addition to the existing literature by quantifying the degree to which neighbouring airports cannibalize each other’s passengers markets. The key finding of this paper is that neighbouring airports do exert a negative externality on each other in terms of “stealing” passengers from one another. In the German market, an additional nearby airport will reduce passengers at an airport by approximately 9% on average. Conversely, airport exit is predicted to increase the passengers of nearby airports by 9% on average.
As airports face high fixed costs and decreasing unit costs with passengers, competition may not be welfare improving when existing airports have excess capacity. In particular, competition is unlikely to be efficient when loss-making airports are supported by subsidies. In Germany, which has one of the densest airport networks in Europe, the majority of airports with less than 4 million passengers per annum are loss-making and have excess capacity. This paper suggests that overall welfare could be increased if there were fewer airports in the market with each airport serving a greater number of passengers. However in order to optimize the number airports in the market, German air transport policy will need to shift the responsibility for airport planning from the regional level to the federal level.
26
Airport IATA code Passengers (2012)
Frankfurt FRA 57 274 099
München MUC 38 217 181
Berlin airports BER, TXL, SXF 25 236 664
Düsseldorf DUS 20 808 472 Hamburg HAM 13 677 609 Stuttgart STR 9 683 309 Köln/Bonn CGN 9 258 861 Hannover HAJ 5 263 952 Nürnberg NUE 3 570 748 Hahn HHN 2 649 585 Bremen BRE 2 441 769 Weeze/Niederrhein NRN 2 206 898 Leipzig/Halle LEJ 2 089 530 Dortmund DTM 1 896 885 Dresden DRS 1 871 113 Karlsruhe/Baden-Baden FKB 1 278 481 Münster/Osnabrück FMO 1 013 430 Paderborn/Lippstadt PAD 864 139 Friedrichshafen FDH 539 930 Saarbrücken SCN 387 302 Erfurt ERF 175 864
Appendix 1
27
Appendix 2
List airlines defined as of low-cost carriers by ADV
Aer Lingus Air Arabia Maroc Air Baltic Air Berlin Blue Air Bmibaby Corendon Easyjet flybe Germanwings Iceland Express Intersky Jet 2 Niki Norwegian Ryanair Transavia Windjet Wizz Wizz Ukraine
28
References
ACI (2012). Economics Report 2012.
Bel, G. and Fageda, X. (2010). Privatization, regulation and airport pricing: An empirical analysis for Europe. Journal of Regulatory Economics, 37 (2): 142-161.
Berry, S. and Waldfogel, J. (1999). Free Entry and social inefficiency in radio broadcasting. Rand Journal of Economics. 30(3): 397-420.
Bilotkach, V., Clougherty, J., Mueller, J. and Zhang, A. (2012). Regulation, privatization, and airport charges: panel data evidence from European airports. Journal of Regulatory Economics, 42: 73-94.
Borins, S. and Advani, A. (2002). Managing airports: a test of the New Public Management. International Public Management Journal, 4(1): 91–107.
Camargo, L. (2013). Airport Privatization Movement in the 21st Century. Working Papers, Paper 9. [Online] Available:
http://opensiuc.lib.siu.edu/cgi/viewcontent.cgi?article=1010&context=ps_wp [Accessed 2014, June 25].
Davis, P. (2006). Spatial Competition in retail markets: movie theaters. Rand Journal of Economics, 37(4): 964-982.
Deutsche Welle (2013). Germany questions use of regional airports. 6 April 2013. [Online] Available: http://www.dw.de/germany-questions-use-of-regional-airports/a-16725429
[Accessed 2014, July 30].
Dmitry, P. (2012). Airport Benchmarking and Spatial Competition: A Critical Review. Transport
and Telecommunication, 13(2), 123-137.
Doganis, R. and Thompson, G. (1974). Establishing Airport Cost and Revenue Functions. The Aeronautical Journal,78: 285-304.
Eurocontrol (2013) Challenges in Growth Task 4: European Air Traffic in 2035. [Online] Available: http://www.eurocontrol.int/sites/default/files/content/documents/official-documents/reports/201306-challenges-of-growth-2013-task-4.pdf [Accessed 2014, June 25].
European Commission (2014a). New State aid rules for a competitive aviation industry. Competition policy Brief, Issue 2 February 2014. [Online] Available:
http://ec.europa.eu/competition/publications/cpb/2014/002_en.pdf [Accessed 2014, June 25].
29
European Commission (2014b). Guidelines on State Aid to Airports and Airlines.
Forsyth, P. (2006). Estimating the Costs and Benefits of Regional Airports Subsidies: A Computable General Equilibrium Approach, paper given at German Aviation Research Society Workshop, Amsterdam, June-July 2006.
Forsyth, P., Gillen, D., Müller, J. and Niemeier, H-M. (2010). Airport Competition: The European Experience. England: Farnham: Ashgate Publishing Limited.
Forsyth, P. (2010). Competition between Major and Secondary Airports – Implications for Pricing Regulation and Welfare. In: Airport Competition: The European Experience. Forsyth, P., Gillen, D., Müller, J., and Niemeier, H-M. Eds. Surrey: Ashgate Publishing Ltd. 77 – 88.
Gabszewicz, J., Thisse, J., Fujita, M. and Schweizer, U. (2001). Location theory. Great Britain: Harwood Academic Publishers GmbH.
Gillen, D. and Mantin, B. (2013). Transportation Infrastructure Management One-and Two-sided Market Approaches. Journal of Transport Economics and Policy (JTEP), 47(2):207-227. Gillen, D. and Niemeier, H-M. (2006). Airport Economics, Policy and Management: The
European Union. In the Fundación Rafael del Pino Workshop on Infrastructure Economics: a Comparative Analyses of the Main Worldwide Airports, Madrid, Spain (p. 53). Madrid, Spain: Rafael del Pino Foundation.
Graham, A. (2013). Understanding the low cost carrier and airport relationship: A critical analysis of the salient issues. Tourism Management, 36: 66-76.
Hess, S. and Polak, J. (2005). Mixed logit modelling of airport choice in multi-airport regions. Journal of Air Transport Management, 11:59-68.
Heymann, E., Vollenkemper, J., Frank, H. J. and Walter, N. (2005). Expansion of regional airports: Misallocation of resources. Deutsche Bank Research, Germany.
Hotelling, H. (1990). Stability in competition. Springer New York. 50-63.
Hu, T. (1972). The Fitting of Log-Regression Equation When Some Observations in the Regressand are Zero or Negative. Metroeconomica, 24(1): 86-90.
ICAO (2013). Case Study on Commercialization, Privatization and Economic Oversight of Airports and Air Navigation Services Providers Germany. [Online] Available
http://www.icao.int/sustainability/CaseStudies/Germany.pdf [Accessed 2014, June 25].
Ivaldi, M., Sokullu, S. and Toru, T. (2012). Chapter 10 Are Airports Two-Sided Platforms?: A Methodological Approach, in James Peoples (ed.) Pricing Behavior and Non-Price
30
Characteristics in the Airline Industry (Advances in Airline Economics, Volume 3). Emerald Group Publishing Limited, 213-232.
Lei, Z. and Papatheodorou, A.(2010) . Measuring the effect of low-cost carriers on regional airports' commercial revenue. Research in Transportation Economics, 26(1): 37-43.
Malina, R. (2006). Competition and regulation in the German airport market. Discussion Paper 10. Germany: Institute of Transport Economics.
Martínez-Garcia, E., Berta Ferrer-Rosell, B. and Coenders, G. (2012). Profile of business and leisure travelers on low cost carriers in Europe. Journal of Air Transport Management, 20: 12-14.
Oum, T., Yan, J. and Yu, C. (2008). Ownership forms matter for airport efficiency: A stochastic frontier investigation of worldwide airports. Journal of Urban Economics, 64: 422-435.
Pantazis, N. and Liefner, I. (2006). The impact of low-cost carriers on catchment areas of established international airports: the case of Hanover airport, Germany. Journal of Transport Geography, 14(4): 265 -272.
Rochet, J. and Tirole, J. Two‐sided markets: a progress report. The RAND Journal of Economics, 37(3): 645-667.
Ryan, C. and Birks, S. (2005). Passengers and low cost flights: evidence from the Trans-Tasman routes. Journal of Travel and Tourism Marketing, 19(1): 15-27.
Starkie, D. (2002). Airport regulation and competition. Journal of Air Transport Management, 8: 63–72.
Starkie, D. (2009). The airport industry in a competitive environment: A United Kingdom Perspective. In Competitive Interaction between Airports, Airlines and High-Speed Rail London: OECD Publishing. 67-9.
Stock, J., Wright, J. and Yogo, M.. (2002). A survey of weak instruments and weak identification in generalized method of moments. Journal of Business and Economic Statistics, 20: 518– 529.
Strobach, D. (2006). Competition between airports with an application to the state of Baden-Württemberg. Working Paper 272/2006. Germany: University of Hohenheim.
Suzumura, K. and Kiyono, K. (1987). Entry Barriers and Economic Welfare. Review of Economic Studies, 54:157-167
31
Tretheway, M., and Kincaid, I. (2005). Competition between airports in the new Millennium: what works, what doesn’t work and why. In The 8th Hamburg Aviation Conference, 16– 18 February 2005 (p. 18). Hamburg, Germany: InterVISTAS Consulting Inc.
Van Dender, K. (2007). Determinants of fares and operating revenues at US airports. Journal of Urban Economics, 62:317–336.
Zhang, Y. and Xie, Y. (2005). Small community airport choice behavior analysis: a case study of GTR. Journal of Air Transport Management, 11: 442–447.
