WHAT HURTS THE DOMINANT AIRLINES OF HUB AIRPORTS



WHAT HURTS THE DOMINANT AIRLINES AT HUB AIRPORTS?IntroductionNetwork airlines have increasingly concentrated their flights in a small number of airports that they dominate and from which they operate their hub-and-spoke routes. By adopting this strategy they are able to reduce their costs, through the exploitation of density economies, and they can offer higher flight frequencies, which are highly valued by business and connecting passengers. While competition between network airlines operating at different hubs to attract connecting passengers may be intense, at their own hub airports the airlines have typically benefited from a rather weak competition with low-cost airlines. However, in Europe, network airlines are increasingly concerned by the expansion of the operations of low-cost carriers at their operating bases. For example, the current financial distress being faced by Iberia and Alitalia is, in part, attributable to competition from low-cost airlines operating in Madrid and Rome-FCO airports, respectively. KLM has been forced to operate with a low-cost subsidiary on many routes out of Amsterdam, while the bankruptcy of Malev can be explained in part by the success of low-cost airlines operating from Budapest. More generally, in the period 2002-2013, the network airlines’ share has fallen in 17 of 22 large European airports that have traditionally been dominated by former flag carriers (see details in Table 1). A loss in the competitiveness of the dominant network airlines may have a markedly negative impact on their respective hub airports. The dominance of the network airlines has benefitted the airports and their corresponding urban areas in a number of ways. The traffic is higher than that generated solely by local demand because a large proportion of passengers in hub airports are connecting passengers. Furthermore, the number and geographical scope of non-stop destinations is especially high at hub airports. In this regard, several studies have shown that air traffic services have a strong influence on firms’ location choices (Brueckner, 2003; Green, 2007; Bel and Fageda, 2008; Bilotkach, 2013), to the effect that the amount of traffic and the number and range of non-stop destinations associated with an airport are determinants of the attractiveness of an urban area. The influence of air traffic services on location choices is particularly relevant for those firms that specialize in knowledge-intensive activities in which face-to-face interaction between workers in different cities is crucial. However, airport dominance and high route concentration may lead to higher average air fares, as has been well documented in the literature. Without intending to be exhaustive, this is a conclusion reported in Borenstein (1989), Brander and Zhang (1990), Brueckner et al. (1992), Evans and Kessides (1993), Oum et al. (1993), Marín (1995), Berry et al. (1996), Fisher and Kamerschen (2003), Fageda (2006), Goolsbee and Syverson (2008) and Bilotkach and Lakew (2014). Equally well documented is the downward pricing pressure that low-cost airlines exert on the routes they operate. Studies that deal with the impact of low-cost airlines on price competition include Dresner et al. (1996), Windle and Dresner (1999), Morrison (2001), Hofer et al. (2008) and Oliveira and Huse (2009). A further, potentially negative, effect associated with hub airports concerns congestion. Airlines operating hub-and-spoke networks may have incentives to keep frequencies high even if congestion at their hub airports is aggravated. In a study of large US airports, Fageda and Flores-Fillol (2013) report that network airlines operating at hub airports, unlike airlines operating point-to-point routes, do not reduce frequencies when airport delays become longer.Thus, whatever the positive and negative impacts of hub airports, no one seems averse to having one in their vicinity. Given this state of affairs, our aim is to determine which dimensions of competition might undermine the competitive position of dominant airlines at hub airports. Specifically, we seek to address the question of whether competition takes place at the route, airport and/or city-pair levels. Furthermore, we seek to disentangle whether what actually matters is the overall degree of competition or the identity of the competitor (that is, network or low-cost airlines). We identify the competitive position of each airline by the flight frequencies they are able to provide on a given route. Flight frequency is typically considered the main attribute of air service quality as it determines the schedule delay cost, i.e., the difference between the desired and actual time of departure. We estimate an equation in which the dependent variable is the frequencies offered by European network airlines on routes departing from their hub airports, using data for the period 2002-2013. Controlling for several route attributes, we consider both route and airport concentration variables together with the share of non-network airlines at the route and airport levels. Additionally, we examine the impact that Ryanair may have when offering flights for the same city-pair market from nearby secondary airports. Finally, we also take into account the effect of mergers, and the EU-US open skies agreement. Previous studies examining the determinants of airline frequencies have focused primarily on route competition and on an airport’s hub status (Schipper et al., 2002; Richard, 2003; Pai, 2010; Bilotkach et al., 2010, 2013; Brueckner and Luo, forthcoming). Additionally, several studies have analyzed the impact of low-cost airlines on price competition but less evidence has been found regarding their impact on service levels in hub-and-spoke structures. The main contribution of this paper, therefore, is the examination it undertakes of whether low-cost airlines have an effect on network airline frequencies beyond those routes on which they directly compete. While previous studies have emphasized the determinants of pricing power, here we focus on the potentially harmful effects that low-cost airlines may have on the competitiveness of Europe’s hub airports.It should be stressed that most empirical studies of airline competition have been conducted for the US market, for which data availability is much better. In this paper, however, we provide evidence of competition between network and low-cost airlines for a large sample of European airports. Finally, we also provide evidence of the impact of mergers on the hubs of the smaller airline in Europe. In this regard, Bilotkach et al. (2013) show that the merger of Delta and Northwest led to a re-organization of the route structure in favor of the hubs of the larger airline. The rest of this paper is organized as follows. In the next section, we explain the data used in the empirical analysis and the criteria applied in building the sample and variables. Then, we specify the empirical model and state our expectations for each explanatory variable. The following section deals with various econometric issues and reports the regression results. The last section contains our concluding remarks. DataThe empirical analysis draws on route-level data from large airports in the European Union (as well as Norway and Switzerland) and covers a period that extends from 2002 through 2013. We include the large European airports at which the same airline was dominant throughout the period of study and at which that dominant airline was not a low-cost carrier. Following these criteria, our sample is based on the following airline-airport pairs: Air France (Paris-CDG, Paris-Orly), Air Lingus (Dublin), Alitalia (Rome-FCO), Austrian Airlines (Vienna), British Airways (London-LHR, London-LGW), Czech Airlines (Prague), Iberia (Madrid), Finnair (Helsinki), KLM (Amsterdam), LOT (Warsaw), Lufthansa (Frankfurt, Munich, Dusseldorf), SAS (Stockholm-ARN, Copenhagen, Oslo-OSL), SN Brussels (Brussels), Swiss (Zurich), TAP (Lisbon) and Tarom (Bucharest). A number of large European airports are not included in the analysis because we were unable to identify one dominant airline operating out of them for the whole period. For example, the bankruptcy of Malev in 2011 prevents us from including Budapest, while the de-hubbing of Alitalia from Milan-MXP has meant that Alitalia has operated very few flights at this airport since 2011. Likewise, it has proved impossible to determine whether Olympic Airlines or Aegean was the dominant airline in Athens, while Manchester has had a highly diversified pool of airlines offering flights with no single company accounting for a share of more than 10 per cent, and various airlines have been dominant in the period in Barcelona (Iberia, Clickair and Vueling). Other airports, such as Palma de Mallorca, Berlin-TXL and London (LTN, STN), are not included because they are dominated by low-cost carriers. However, it should be pointed out that not all the airports included in our sample can be unequivocally classified as hub airports throughout the whole period. Specifically, British Airways has been progressively reducing its traffic at London-LGW, but it continues to account for around 18 per cent of the total flights at this airport. According to data from the UK Civil Aviation Authority, the share of connecting passengers at LGW is still higher than 10 per cent. The exclusion of important European airports from our sample is a limitation but the aim of the analysis we conduct is to determine the influence of different attributes of competition on the hub operations of network airlines. In this regard, our restricted sample covers a very high proportion of all hub operations undertaken by network airlines at Europe’s airports. We have been able to collect complete data for 952 routes on which the airlines under consideration provided an air service in all the years of the period studied (2002-2013). As such, the analysis excludes information for thin routes. Overall, our sample contains 11,424 observations, although regressions are based on 10,472 observations because we use a one-year lag of some of the explanatory variables. Our data include intra-European routes as well as links to non-European destinations. However, the collection of population and per capita GDP data is more homogeneous in the case of the intra-European routes. For EU destinations, population and per capita GDP data refer to the NUTS 3 regions (the statistical unit used by Eurostat) and have been provided by Cambridge Econometrics (European Regional Database publication). For non-EU destinations, population data refer to metropolitan areas and the information has been drawn from various sources: the OECD, United Nations (World Urbanization Prospects), World Bank and national statistics agencies. To construct the per capita GDP variable for these non-European destinations, we use the country classification by income groups developed by the World Bank. Thus, we construct an index in which we distinguish between low income, lower middle income, upper middle income and high income countries. As such, the regressions that consider all routes use the country index for the per capita GDP variable, while the regressions that consider solely the intra-European routes use the continuous variable at the NUTS 3 level. Note that population and per capita GDP data (the latter when considered at the EU regional level) are only available up to 2011 and so we use the data from 2011 to compute figures for 2012 and 2013 to complete the series. Airline frequency data at the route level have been obtained from RDC aviation (Capstats statistics), while route distance data are taken from the Official Airline Guide (OAG). Note we use explanatory variables that distinguish between two types of airline: 1) airlines integrated in international alliances (Oneworld, Star Alliance and SkyTeam) and/or former flag carriers of the respective countries of the airports of origin in our sample, and 2) airlines that are not former flag carriers and which are not integrated in alliances. Thus, our approach distinguishes between airlines that exploit connecting traffic at European airports as an essential part of their business and airlines that focus their business on point-to-point routes. By drawing this distinction, we are able to avoid the complex task of having to draw up a list of low-cost carriers without comprehensive data regarding airline costs.In the case of the European Union, former flag carriers can be considered to be network airlines regardless of whether they are integrated in international alliances or not. Taking this into account, most airlines not integrated in such alliances and which provide services on intra-European routes concentrate their business on point-to-point routes. Some exceptions do exist, especially on certain intercontinental routes where airlines such as Emirates or Air Transat offer services, but their aggregated impact is modest. In any case, the criterion employed allows us to distinguish between airlines that operate hub-and-spoke routes and those that operate point-to-point routes, especially in the case of intra-European routes. We also consider whether the dominant airline has been involved in a merger with a larger company. In our sample, the airlines involved in such mergers are KLM (since 2005), Iberia (since 2012), Austrian Airlines (since 2010) and Swiss (since 2006). SN has signed a strategic partnership deal with Lufthansa but the latter does not have a majority stake. Finally, we include a variable that examines the influence of Ryanair operating from a nearby secondary airport. Hence, we construct a dummy variable that identifies whether Ryanair offers at least one daily flight from an airport that is less than 100 kilometers from the city center of the airport of origin in our sample. We compute the value of one when Ryanair offers daily flights in the same city-pair market as that of the dominant airline in the airports in our sample. For example, SN Brussels offers flights on the Brussels-Manchester route and Ryanair also provides a service to Manchester from Brussels-Charleroi. In our sample, the secondary airports at which Ryanair enjoys a considerable presence are Charleroi, Skavasta, Beauvais, Stansted, Luton, Ciampino, Weeze, Bratislava and Moss.Table 1 reports the data for the sample airports. In the period 2002-2013, the evolution in total traffic is quite diverse with some airports recording substantial growth (for example, Helsinki, Lisbon, Oslo and Bucharest), and others recording losses (for example, Stockholm, Brussels and Madrid). The mean traffic share of the dominant airline is in all circumstances higher than 30 per cent and in some cases as high as 60 per cent. Here, the dominance of the leading airlines at some airports has been strengthened while in other it has weakened. The network airlines’ share is generally well above 50 per cent, with the exceptions of Dublin, Oslo, London-LGW and Bucharest where low-cost airlines, such as Ryanair, Norwegian, Easyjet and Wizzair, have a sizeable presence. This being said, the network airlines’ share has fallen in most of the airports in the period of study – in fact, in 17 of the 22 airports making up our sample. Overall, the airports considered here present great variation in the evolution of their traffic and in the respective shares attributed to the different airlines operating out of them. Yet, what seems clear from the data in Table 1 is the trend towards an increase in the presence of airlines not integrated in an alliance in many airports that were previously controlled by former flag carriers. This is quite remarkable if we consider that our sample of airports excludes those that are the home base of a low-cost airline. Insert Table 1 about hereEmpirical modelIn this section, we implement a multivariate analysis to identify the determinants of the flight frequencies offered by the dominant airlines at Europe’s large airports. We use similar control variables to those employed in other empirical studies that estimate the determinants of frequencies on air routes (see, for example, Schipper et al., 2002; Richard, 2003; Pai, 2010; Bilotkach et al., 2010; Brueckner and Luo, forthcoming). Our specific contribution is to analyze the impact of several variables of competition at the route, city-pair and airport levels. To this end, we estimate the following equation using data for a large number of routes departing from our sample of European airports: Frequenciesdominant_airlinekt= α + β1Populationdestinationkt + β2Incomedestinationkt + β3Distancek + β4DEUk + β5DUS_openskieskt + β6Dmergerkt + β7Dcompetition_secondary_airportkt + β8HHIroutekt + β9Dcompetition_no-networkkt + β10HHIorigin_airportkt + β11Share_no-networkorigin_airportkt + α’Dorigin_airportk + μ'Dyeart + ε (1)In this equation, the dependent variable is the total number of annual flights offered by the dominant airline on route k in year t. As explanatory variables, we include variables that measure the population and per capita income of the destination in order to control for demand. We expect airlines to offer higher frequencies on routes that link richer and more populous cities. We also take into account the influence of the route’s distance, calculated as the number of kilometers flown to link the route’s endpoints. Airlines may be required to offer high-frequency services on shorter routes so as to compete with surface transportation modes. Note also that airlines may prefer to use smaller planes at higher frequencies on short-haul routes. Thus, we would expect a negative relationship between distance and frequency. Additionally, we include dummies for intra-European routes and routes to the United States for the period after the EU-US open skies agreement was signed. Controlling for other factors, demand on intra-European routes might be higher as a result of the greater degree of integration of EU members and the fact that the EU market is a liberalized market. By contrast, former European flag carriers may encounter more competition in the EU-US market following the open skies agreement, while the supply of flights may have diversified with the introduction of new airlines and airports. Thus, we expect a positive sign for the coefficient associated with the intra-EU route variable and a negative sign for the variable capturing US destinations after the introduction of the open skies agreement. Furthermore, we consider a dummy variable that takes a value of one for routes and periods in which the dominant airline at the airport was acquired by another larger airline. Following the merger, a reorganization of the route network might have been implemented in favor of the airports of the larger airline (Bilotkach et al., 2013). Hence, we expect a negative sign for the coefficient associated with this variable. The main focus of our analysis is on the competition variables. We include a dummy variable that takes a value of one for those routes on which Ryanair offers daily services in the same city-pair market from a nearby secondary airport. The coefficient associated with this variable should be negative if the services of the dominant airline at the large airport are substantially affected by competition from Ryanair operating at the nearby airport. However, it might also be the case that city-pair markets that are connected via different airports are particularly dense corridors whose demand is not fully controlled by our explanatory variables. We also include two variables that seek to capture competition at the route level. First, we consider the route concentration, measured using the Herfindahl-Hirschman index, in terms of flight frequencies. The expected sign of the coefficient associated with this variable is not a priori clear. There may be a market power effect so that the dominant airline at the airport might reduce frequencies if it is subject to less competition on the route. However, greater competition implies less demand from point-to-point passengers so that frequencies might be lower on less concentrated routes. In addition to route concentration, we consider a dummy variable that takes a value of one when non-network airlines have a share that is greater than 10 per cent on the route. Recall that we consider non-network airlines to be those airlines that are not former European flag carriers and which are not integrated in alliances. Controlling for route concentration, a relevant presence of non-network airlines on a given route might reduce demand from point-to-point passengers, so we would expect a negative sign for the coefficient associated with this variable. Competition between airlines integrated in the same alliance may be weak (in fact, they usually operate code-share agreements) and other airlines (other than the dominant one) operating hub-and-spoke networks frequently use the airport under consideration to feed their hubs. In such circumstances, the dominant airline may be especially harmed by competing airlines on the same route that operate point-to-point routes, because these airlines are more likely to be disputing the passengers with final destination at the airport under consideration by means of aggressive offers. It should come as no surprise then to find that the frequencies of the dominant airlines are reduced when they have a non-network competitor on a given route. Although the effects of such offers should be modest in the case of connecting passengers, their impact can be much greater in the case of point-to-point passengers.Less clear is the extent to which the degree of dominance enjoyed by an airline at an airport might affect the frequencies that this airline offers on a given route. Although network effects should be great at hub airports, since the percentage of transfer passengers is generally high, previous studies have tended to focus on the impact that route concentration has on frequencies. In this regard, we include two additional variables as explanatory factors: the Herfindahl-Hirschman index in terms of airline frequencies at the airport level, and the share that non-network airlines have at the airport. The weaker position of the leading airline at the airport may mean that this airline offers lower frequencies on the routes served (and fewer non-stop destinations). More competition at the airport level, measured using the concentration index, may mean less demand from connecting passengers, because the hubbing airline offers less competitive connections – indeed, the coordination of banks of arrivals and departures could be poorer with increased connecting times. Thus, we expect a positive sign for the coefficient associated with the airport concentration variable. Furthermore, a higher share of non-network airlines at the airport may have negative consequences for the dominant airline. Indeed, the dominant airline may receive less demand from point-to-point passengers on those routes that suffer the rivalry of low-cost airlines, which may reduce the flight frequency offered. Lower frequencies on some routes may have indirect effects on other routes because the demand from connecting passengers may be lower as a result of less competitive connections. Indeed, a low-cost airline with a considerable share at the hub airport may harm the dominant airline because the latter might lose traffic from its point-to-point passengers. Hence, the dominant airline may have lower demand and profitability on the spoke routes. However, low-cost airlines operate at low fares so they may be able to increase traffic on the routes that they operate. Thus, part of this additional traffic might comprise connecting passengers, from whom the dominant airline can receive some benefits. Overall, the expected sign of the coefficient associated with this variable is not entirely clear although the negative effect related to less demand from point-to-point passengers may override the positive effect related to more demand from connecting passengers, because low-cost airlines tend to focus on the former passenger type. Finally, we include dummies for the airports of origin and year. The airport dummies control for time-invariant airport-specific omitted variables, while the year dummies control for the common trend on all routes in the dataset. Specifically, airport and time fixed effects may help to control for relevant factors such as congestion for which, unfortunately, we have no available data. εit is a mean-zero random error. Estimation and resultsIn this section, we deal with a number of econometric issues and discuss the results of the regressions. The estimates may present non-stationarity and temporal autocorrelation problems. The Wooldridge test for autocorrelation in panel data shows that we may have a problem of serial autocorrelation, which must be addressed. We also apply the panel unit root test developed by Levin et al. (2002), which can be regarded as an augmented Dickey-Fuller (ADF) test when lags are included. This test indicates that there is no non-stationarity problem with our dependent variable.We perform the estimation using two different techniques that take advantage of the panel nature of our data: the fixed and random effects models. An advantage of the fixed effects model is that it allows us to control for any omitted variables that correlate with the variables of interest and which do not change over time. As such, the fixed effects model is more reliable than other estimation techniques. A shortcoming of the fixed effects model is that it may be less informative than other techniques because the effect of time-invariant variables cannot be identified. In our context, the Hausman test is not useful for choosing between the random and the fixed effects model because the former can include more variables, such as distance, the dummy for intra-EU routes and, especially, the dummies for the airports of origin. Hence, what we have opted to do is to present the results using both the random and fixed effects models assuming an AR(1) process in the error term and standard errors robust to heteroscedasticity. The results for the main variables in our analysis are not affected by the estimation technique used. An additional issue that must be addressed is the potential endogeneity of the concentration variables. To deal with this, we include a one-year lag of the concentration variables as explanatory variables. It is difficult to make a case for the correlation between lagged concentration and current unobserved shocks. We also experimented with additional lags of these variables and the results are not affected. In order to simplify the presentation of our results, we only report the results of regressions with a one-year lag of the concentration variables. We make the estimation using all the observations and for the different subsamples. Specifically, we distinguish between intra-EU routes and routes that link the airports of origin in our sample with non-EU destinations. Population and per capita income data are richer in the case of the intra-European routes and the distinction between network and non-network airlines operating on the route is clearer in the case of the intra-EU routes. Furthermore, the use of code-share agreements between the dominant airline and other airlines on some routes may generate some distortion in our results. This is particularly true on the non-EU routes. In the case of intra-EU routes, we also distinguish between short-haul and long-haul routes. Thus, we estimate our equation for routes shorter and longer than the mean distance for intra-EU routes, which is about 900 kilometers. For our purposes, it is important to make this distinction between short-haul and long-haul routes because the proportion of connecting passengers over total passengers may be increasing with distance. We do not have systematic data to support this assertion but it is evident that long-haul routes in Europe are strongly concentrated at hub airports characterized by a high-proportion of connecting passengers. Note also that competition from low-cost airlines should be stronger on short-haul routes.In short, we make the estimation using these samples: 1) all routes, 2) all intra-EU routes, 3) intra-EU routes of less than 900 kilometers, 4) intra-EU routes of more than 900 kilometers. Table 2 shows the descriptive statistics of the variables used in the empirical analysis, while Table 3 presents the correlation matrix of these variables. It can be seen that all the variables present sufficient variability, as the standard deviation is high in relation to the mean values. In the case of the correlation matrix, it is notable that the correlation between the concentration variables and those that reflect the presence of non-network competitors is sufficiently low for us to be able to identify the specific effects of each variable. Tables 4 and 5 show the results of the estimates when using the random and fixed effects model, respectively. As explained above, the fixed effects model is not able to capture the effect of time-invariant variables. This explains why the overall explanatory power of the model is considerably higher in the regressions that use random effects. In the regressions that use the random effects model, the control variables, in general, work as expected. The frequencies of the dominant airline are higher when the route links more populous and richer endpoints. The fixed effects model does not seem to capture the impact of these control variables as it concentrates on the within-variation of data. Furthermore, we find a negative relationship between frequencies and distance in the regressions that can be identified. The coefficient associated with the dummy variable for intra-EU routes is positive and statistically significant in the regressions that can be identified. As expected, frequencies on intra-EU routes are higher due to greater demand in a geographical area characterized by no regulatory restrictions and strong economic integration. Interestingly, the frequencies of the dominant airline have fallen after the open skies agreement was introduced on routes to US destinations, but the effect is not statistically significant. Overall, it is not clear from our results that former European flag-carriers have been exposed to more intense competition in the EU-US market after the open skies agreement came into force. We also find that the frequencies of dominant airlines are lower in the period following their merger with a larger airline. The coefficient associated with this variable is always negative (although it is not clearly significant in the regressions that focus on long-haul European routes). Overall, we find evidence for Europe that mergers may imply a re-organization of the route structure in favor of the hubs of the larger airline (see Bilotkach et al., 2013, for an analysis with similar results for the US airline market). We do not find any evidence that competition from Ryanair operating out of secondary airports harms the dominant airlines considered here. It could be the case that the city-pair markets in which Ryanair competes with former flag-carriers are particularly dense corridors, at least in our sample. It could also be that Ryanair and former flag carriers attract different types of passenger, so that the leading low-cost airline in Europe is in fact fighting for more price-sensitive passengers. Regardless of the reason for this result, we can conclude that competition from secondary facilities does not appear to harm the competitive position of the dominant airlines operating from Europe’s main airports.At the route level, the coefficient of the route concentration variable is only statistically significant (with a positive sign) in the regressions for the intra-EU routes of more than 900 kilometers. Recall that the result for this variable could be explained by a market power effect whereby the dominant airline reduces frequencies with less competition and with a lower demand from point-to-point passengers so that the dominant airline increases frequencies with less competition. The latter effect only seems to be apparent on longer intra-EU routes. The coefficient associated with the dummy variable for relevant non-network competitors is negative and statistically significant in all the regressions, with the exception of those that focus on intra-EU routes of more than 900 kilometers. This latter result indicates that low-cost airlines may be less competitive on longer routes. At the airport level, the coefficient of the variable of airport concentration is only statistically significant (with a positive sign) in the regressions that use the random effects model and that focus on intra-EU routes of more than 900 kilometers. A weaker position of the dominant airline at the airport may have an impact on frequencies particularly on longer routes where we can expect the proportion of connecting passengers to be higher. Notably, the coefficient associated with the variable of the share of non-network airlines at European airports is negative and statistically significant in all the regressions. It seems that the negative impact on the dominant airline associated with the decrease in demand from point-to-point passengers is higher than the potentially positive effect related to the increase in traffic generated by low-cost airlines. On the basis of our results, we clearly demonstrate (as expected) that non-network airlines mainly generate traffic from point-to-point passengers. This is quite reasonable since connecting passengers seem to prefer to fly with network airlines that provide a better coordination of schedules, easier baggage management and stronger guarantees of making the connection. If we examine the elasticities in the estimation using the random effects model (which allows us to control for more variables than is the case with the fixed effects estimation), it is interesting to note the differences between the short- and long-haul European routes. The elasticity obtained for the route concentration variable is almost 0 in the case of short-haul routes, while it is around 2% for long-haul routes. The elasticity obtained for the airport concentration variable is 1% on short-haul routes and 6% on long-haul routes. The equivalent figures are 16 and 10% when we consider the variable capturing the share of the non-network airlines. Interpreting the elasticities for the dummy of low-cost competitors on the route is not so straightforward, but the magnitude of the coefficient and the statistical significance clearly show that it is only relevant on short-haul routes. Overall, the picture is quite clear and the outcomes are similar in the random and fixed effects regressions. The regressions that do not distinguish by route distance show that the impact of the variables for non-network competitors is stronger than that of the concentration variables. When the routes are distinguished by distance in the European sample, we find that the influence of the variables for concentration seems to be higher on longer routes while the effect of the variables for non-network competitors seems to be higher on shorter routes. Indeed, low-cost airlines have been able to exploit their cost advantages at the expense of network carriers in short-haul and medium-haul markets, suggesting that their competitive pressure is decreasing with distance. This could account for the stronger impact of the variables for non-network competitors on shorter routes. Furthermore, a possible explanation for the stronger effect of the concentration variables on longer routes might be related to the fact that the proportion of connecting passengers over total passengers is higher on longer routes. Overall, our empirical analysis indicates that dominant airlines may be worried by the increased presence of non-network airlines at their hub airports. Critically, the negative effects suffered by dominant airlines as a result of the stronger presence of non-network airlines at their operating bases are not only felt on the routes on which they compete directly with each other, but on other routes that may suffer a reduction in demand from connecting passengers. Concluding remarksThe main contribution of this paper has been to show that network airlines are undermined to some extent by the greater presence of low-cost airlines at their hub airports. Moreover, the impact of low-cost airlines seems to extend beyond the routes on which these companies compete directly with the dominant network airline, because, as the proportion of connecting passengers on many routes with a hub airport as their endpoint is generally high, the effects of competition on one route seem to be transferred to other routes. However, the fact that Ryanair offers flights in the same city-pair market does not seem to be detrimental to hubbing airlines. Moreover, competition from airlines in alliances seems to be weak, although further analysis should disentangle the effect derived from airlines that are integrated in the same alliance as that of the hubbing airline from that derived from airlines in a different alliance. We also find evidence that mergers in Europe may result in a re-organization of the route structure in favor of the hubs of the larger airline. Various policy tools could be adopted to support hub operations at Europe’s airports, should it be deemed convenient (Bilotkach and Fageda, 2013), including investments, airport charges and allocation of slots, the latter currently being allocated through the application of administrative rules. Investments are an expensive option and, moreover, there may be urban and environmental restrictions that prevent hub airports from being expanded. If we take into account, therefore, the limitations associated with the capacity solution, charges represent an obvious alternative for airport authorities. Fu et al. (2006) show that an increase in airport charges would harm low-cost airlines significantly more than it would network airlines, while Allroggen et al. (2013) find that discounts on airport charges are more likely when airport managers seek to attract low-cost airlines, such as Ryanair. Given that low-cost airlines operating at hub airports are a serious threat to the competitiveness of network carriers, the setting of airport charges is a powerful tool in the hands of the airport authorities. In short, the positive impact of low-cost airlines in terms of promoting competition on the routes they operate is well documented. However, our analysis shows that at large airports these positive effects must be weighed against the weakening of the hub operations suffered by the dominant network airline. ReferencesAllroggen, F., Malina, R., Lenz, A.K. (2013). Which factors impact on the presence of incentives for route and traffic development? Econometric evidence from European airports. Transportation Research-E 60(1), 49-61.Bel, G., Fageda, X. (2008). Getting there fast: Globalization, intercontinental flights and location of headquarters. Journal of Economic Geography 8(4), 471-495.Berry, S., Carnall, M., Spiller, P.T. (1996). Airline Hubs: Costs, Markups and the Implications of Customer Heterogeneity. NBER Working Paper, 5561, 1-38.Bilotkach, V. (2013). Are Airports Engines of Economic Development? A Dynamic Panel Data Approach. ‘Unpublished results’.Bilotkach, V., Fageda, X. (2013). The public interest of the hub operation at Schiphol. Position paper for Airneth.Bilotkach, V., Fageda, X., Flores-Fillol, R. (2013). Airline Consolidation and the Distribution of Traffic between Primary and Secondary Hubs. Regional Science and Urban Economics 43 (6), 951-963.Bilotkach, V., Fageda, X., Flores-Fillol, R. (2010). Scheduled service versus personal transportation: The role of distance, Regional Science and Urban Economics 40 (1), 60-72.Bilotkach, V., Lakew, P.A (2014). On sources of market power in the airline industry: Panel data evidence from the US airports. Transportation Research-A, 59 (1), 288 – 305.Bilotkach, V., J. Mueller, Nemeth, A. (2013). Consumer welfare effects of dehubbing: Case of Malev bankruptcy. ‘Unpublished results’.Borenstein, S. (1989). Hubs and High Fares: Dominance and Market Power in the US Airline Industry. RAND Journal of Economics 20 (3), 344-65.Borenstein, S., Netz, J. (1999). Why do all the flights leave at 8.am?: Competition and departure time differentiation in airline markets. International Journal of Industrial Organization, 17 (5), 611-640.Brander, J.A., Zhang, A. (1990). A Market Conduct in the Airline Industry: An Empirical Investigation. The Rand Journal of Economics 21 (4), 567-583.Brueckner, J.K. (2003). Airline traffic and urban economic development. Urban Studies 40 (8), 1455-1469.Brueckner, J.K, Luo. D. Measuring Firm Strategic Interaction in Product-Quality Choices: The Case of Airline Flight Frequency. Economics of Transportation, in press.Brueckner, J.K., Spiller, P.T. (1994). Economies of traffic density in the deregulated airline industry. Journal of Law and Economics 37 (2), 379-415.Brueckner, J.K., Dyer, N.J, Spiller, P.T. (1992). Fare Determination in Airline Hub-and-Spoke Networks. The RAND Journal of Economics 23 (3), 309-333.Castillo-Manzano, J.I, López-Valpuesta, L., Pedregal, D.J. (2012a) What role will hubs play in the LCC point-to-point connections era? The Spanish experience. Journal of Transport Geography 24 (1), 262–270.Castillo-Manzano, J.I., López-Valpuesta, L., Pedregal, D.J. (2012b). How can the effects of the introduction of a new airline on a national airline network be measured? A time series approach for the Ryanair case in Spain. Journal of Transport Economics and Policy 46 (2), 263–279.Dresner, M., Lin, J.S. C., Windle, R. (1996). The impact of low-cost carriers on airport and route competition. Journal of Transport Economics and Policy 30 (2), 309-329.Fageda, X. (2006). Measuring conduct and cost parameters in the Spanish airline market. Review of Industrial Organization, 28 (4), 379-399.Fageda, X., Flores-Fillol, R. (2013). Airport congestion and airline network structure. ‘Unpublished results’. Fisher, T., Kamerschen, D.R. (2003). Price-Cost Margins in the U.S Airline Industry using a Conjectural Variation Approach. Journal of Transport Economics and Policy 37 (2), 227-259.Francis, G., Dennis, N., Ison, S., Humphreys, I. (2007). The transferability of the low-cost model to long-haul airline operations. Tourism Management 28(2), 391-398.Fu, X., Lijesen, M., Oum, T.H. (2006). An Analysis of Airport Pricing and Regulation in the Presence of Competition Between Full Service Airlines and Low Cost Carriers. Journal of Transport Economics and Policy 40 (3), 425–447Goldsbee, A., Syverson, C. (2008). How do incumbents respond to the threat of entry? Evidence from the major airlines. The Quarterly Journal of Economics 123(4), 1611-1633. Green, R.K. (2007). Airports and Economic Development, Real Estate Economics 35(1), 91–112.Hofer, C., Windle, R. J., Dresner, M. E. (2008). Price premiums and low cost carrier competition. Transportation Research -Part E 44 (5), 864-882Marín, P.L. (1995). Competition in European Aviation: Pricing Policy and Market Structure. Journal of Industrial Economics 16 (2), 141-159.Morrison, S.A (2001). Actual, Adjacent and potential competition: Estimating the full effect of Southwest airlines. Journal of Transport Economics and Policy 35 (2), 239-256.Oliveira, A.V.M, Huse, C. (2009). Localized competitive advantage and price reactions to entry: Full-service vs. low-cost airlines in recently liberalized emerging markets. Transportation Research Part-E 45 (2), 307–320.Oum, TH., Zhang, A., Zhang, Y. (1993). Inter-firm Rivalry and Firm-Specific Price Elasticities in Deregulated Airline Markets. Journal of Transports Economic and Policy 27 (2), 171-192.Pai, V., 2010. On the factors that affect airline flight frequency and aircraft size. Journal of Air Transport Management 16 (4), 169-177.Pels, E., Njegovan, N., Behrens, C. (2009). Low-cost airlines and airport competition. Transportation Research-E 45 (2), 335–344.Redondi, R., P. Malighetti, Paleari, S. (2012). De-hubbing of airports and their recovery patterns. Journal of Air Transport Management 18 (1), 1-4.Richard, O. (2003). Flight frequency and mergers in airline markets. International Journal of Industrial Organization 21 (6), 907–922Salvanes, K.G., Steen, F., Sorgard, L. (2005). Hotelling in the air? Flight departures in Norway. Regional Science and Urban Economics 35 (2), 193-213.Schipper, Y., P. Rietveld, Nijkamp, P. (2002). European airline reform: an empirical welfare analysis. Journal of Transport Economics and Policy 36 (2), 189-209.Urban Audit (2008). Connecting passengers at selected European airports 2007. London First report ‘Imagine a world class Heathrow’, Urban Audit. Wei, W., Hansen, M (2006). An aggregate demand model for air passenger traffic in the hub-and-spoke network. Transportation Research Part-A 40 (10), 841–851. Windle, R., Dresner, M. (1999). Competitive responses to low cost carrier entry. Transportation Research-E 35(1), 59-75.TABLESTable 1. Data for sample airports in 2002-2013Total frequenciesShare dominant airlineConcentration index (HHI)Share network airlines AirportMeanVariationMeanVariationMeanVariationMeanVariationAmsterdam (AMS)18982215.4%52.7%18.9%0.28839.1%77.6%3.7%Stockholm (ARN)100056-11.9%40.9%-8.2%0.205-13.9%62.0%-7.9%Brussels (BRU)97324-12.3%35.3%34.4%0.14747.1%74.5%5.8%Paris (CDG)239592-8.0%58.0%-0.4%0.344-1.0%83.9%-7.8%Copenhagen (CPH)118790-2.3%47.0%-7.1%0.241-9.0%64.6%-4.2%Dublin (DUB)799307.1%35.5%37.6%0.25162.5%51.0%7.6%Dusseldorf (DUS)9616719.7%40.9%-4.4%0.21320.4%65.7%-23.4%Rome (FCO)15103811.1%45.2%-6.5%0.223-13.6%68.0%-4.7%Frankfurt (FRA)2259612.9%62.2%11.6%0.39123.4%87.7%-0.6%Helsinki (HEL)69132110.9%54.1%14.7%0.32825.8%71.0%-5.9%London (LGW)11048422.9%33.0%-68.8%0.213-29.5%41.8%-71.1%London (LHR)245117-4.4%42.6%34.5%0.20161.5%75.9%13.4%Lisbon (LIS)6351840.1%56.3%53.9%0.35083.4%74.1%21.1%Madrid (MAD)200263-11.4%52.1%-14.1%0.297-22.7%85.6%-14.2%Munich (MUC)18297915.2%62.7%14.2%0.40227.8%80.2%5.5%Oslo (OSL)9418169.5%45.6%-30.5%0.276-22.5%45.6%-30.5%Paris (ORY)11092618.2%57.1%-17.5%0.344-33.6%68.2%-18.2%Bucharest (OTP)3021484.6%52.6%-36.9%0.301-53.6%52.6%-36.9%Prague (PRG)6022526.6%51.5%-41.8%0.285-61.6%80.3%-22.6%Viena (VIE)11292022.2%57.2%-18.3%0.341-30.9%80.4%-20.1%Warsaw (WAW)5543627.3%63.8%-17.0%0.418-29.8%88.4%-15.4%Zurich (ZRH)112232-5.7%56.2%-9.1%0.328-17.4%83.7%-9.7% Table 2. Descriptive statistics of the variables used in the empirical analysisVariableMeanStandard dev.Frequenciesdominant_airline (annual)1143.781019.32Populationdestination (thousands)4287.735927.12Incomedestination (index)3.7850.524Distance (kms)2272.612762.23DEU (dummy)0.7170.450DUS_openskies (dummy)0.0340.182Dmerger (dummy)0.1220.327Dcompetition_secondary_airport (dummy)0.0330.180HHIroute(Percentage over one)0.6730.250Dcompetition_no-network (dummy)0.3080.461HHIorigin_airport (Percentage over one)0.3060.080Share_no-network origin_airport (Percentage over one)0.2300.135Table 3. Correlation matrix of the variables used in the empirical analysisFreqPopInc.DistDmergeDEUDo.skyDsec.HHIrDnone.HHIa.ShareNNaFreq.1Pop.-0.221Income0.27-0.391Dist.-0.410.65-0.321Dmerger-0.040.03-0.020.0091DEU0.42-0.600.53-0.80-0.031Dopenskies-0.120.160.080.330.04-0.311Dsecondary0.11-0.010.07-0.08-0.060.12-0.041HHIrou.-0.03-0.12-0.010.030.040.04-0.02-0.091Dnonetwork0.08-0.05-0.01-0.050.020.01-0.070.02-0.431HHIairp.0.0020.020.04-0.14-0.06-0.060.009-0.200.12-0.141ShareNNa0.02-0.070.13-0.050.160.16-0.0060.11-0.0020.13-0.531Table 4. Results of estimates (Random effects –GLS regression with an AR 1 disturbance). Dependent variable: Frequencies of dominant airlineExplanatory variablesAll sampleIntra-EU routesIntra-EU routes < 900 kms Intra-EU routes > 900 kmsPopulationdestination0.027(0.005)***0.11(0.01)***0.12(0.02)***0.10(0.008)***Incomedestination250.91(60.46)***7.32(1.73)***11.08(2.83)***7.04(1.24)***Distance-0.14(0.017)***-0.91(0.06)***-1.09(0.26)**-0.47(0.06)***DEU398.68(112.79)***---DUS_openskies-6.66(19.53)---Dmerger-20.57(10.52)**-30.20(14.95)**-37.36(22.60)*-19.09(15.80)Dcompetition_secondary_airport28.15(25.41)26.41(29.59)36.26(50.49)25.88(26.74)HHIroute18.73(16.86)16.76(23.30)0.53(36.14)37.79(23.53)*Dcompetition_no-network-24.97(6.30)***-23.34(8.17)***-44.13(12.93)***-2.009(8.11)HHIorigin_airport100.73(74.98)158.38(102.45)76.46(159.31)206.11(104.13)**Share_no-networkorigin_airport-664.54(70.91)***-786.30(94.64)***-1112.60(152.74)***-401.09(92.22)***Intercept-670.76(280.57)**1074.41(265.28)***770.60(508.51)504.56(173.14)***Airport dummiesYESYESYESYESYear dummiesYESYESYESYESR2Test χ2 (joint sig.)Number observations0.33919.18***104720.35720.65***71610.29427.75***42790.58618.67***2805Note 1: Standard errors in parenthesis (robust to heterocedasticity)Note 2: Statistical significance at 1% (***), 5% (**), 10% (*)Note 3: We use one lag of concentration variables (HHI_route, HHI_airport) Table 5. Results of estimates (Fixed effects -Within regression with an AR 1 disturbance)Dependent variable: Frequencies of dominant airlineExplanatory variablesAll sampleIntra-EU routesIntra-EU routes < 900 kms Intra-EU routes > 900 kmsPopulationdestination0.015(0.012)0.18(0.16)0.0005(0.33)0.03(0.14)Incomedestination-3.16(3.58)1.94(5.42)4.12(3.76)Distance----DEU----DUS_openskies-0.56(21.26)---Dmerger-36.31(11.30)***-50.03(15.98)***-65.06(24.22)***-28.97(16.67)*Dcompetition_secondary_airport25.86(26.83)26.59(31.20)23.80(52.77)29.68(28.68)HHIroute-7.37(17.80)-10.86(24.54)-46.42(38.11)39.98(24.78)*Dcompetition_no-network-16.98(6.44)***-16.90(8.32)**-38.04(13.17)***9.45(8.19)HHIorigin_airport73.87(90.24)117.35(123.26)118.98(191.16)60.37(124.87)Share_no-networkorigin_airport-893.94(81.35)***-1104.83(109.59)***-1603.52(178.31)***-542.97(105.24)***Intercept1156.86(20.85)***983.17(113.66)***1865.60(176.73)***693.85(126.72)***Airport dummiesNONONONOYear dummiesYESYESYESYESR2Test F (joint sig.)Number observations0.0525.59***104720.0621.90***71610.0816.88***42790.198.04***2805Note 1: Standard errors in parenthesis (robust to heterocedasticity)Note 2: Statistical significance at 1% (***), 5% (**), 10% (*)Note 3: We use one lag of concentration variables (HHI_route, HHI_airport) ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download