Discounts and Public Service Obligations in the Airline. Market: Lessons from Spain

Transcription

1 Discounts and Public Service Obligations in the Airline Market: Lessons from Spain Joan Calzada y and Xavier Fageda z Universitat de Barcelona and IREA-GiM March 20, 2011 Abstract We study the impact of the universal service policy applied in the Spanish airline market during the period Our analysis shows that routes bene ting from price discounts given by the government to island residents present higher prices than the rest of domestic routes, but similar ight frequencies. This can be explained by the e ect of discounts on demand elasticity, the airlines di culties in acquiring new slots, and the high costs of increasing frequencies. Moreover, we nd that intra-island routes where airlines are regulated with price caps and frequency oors have lower prices and higher frequencies than unregulated routes of similar characteristics. Our results also suggest that island residents price discounts help to guarantee the pro tability of routes protected by public service obligations. Keywords: Air Transportation; Residents Discounts; Price Caps; Frequency Floors. We thank the editor, two anonymous referees, Jan Brueckner, Juan Luís Jiménez, Ricardo Flores-Fillol and seminar participants in the University of Las Palmas de Gran Canaria, IST (Lisboa), ATRS (Porto) and EARIE (Istambul) for useful suggestions. This study has received nancial support from the Spanish Ministry of Education (ECO ECO), the Government of Catalonia (SGR ) and the Fundación Ramón Areces. J. Calzada acknowledges the support of the Barcelona Graduate School of Economics. y Departament de Politica Econòmica. Av. Diagonal 690, Barcelona 08034, Spain, calzada@ub.edu. z Corresponding author: Departament de Politica Econòmica, Universitat de Barcelona, Av. Diagonal 690, Barcelona 08034, Spain, xfageda@ub.edu.

2 1 Introduction The number of competitors in European air travel has increased signi cantly in recent decades as a consequence of liberalization and the successful entry of low-cost carriers in many shorthaul routes. However, not all regions take bene t of competition. Thin and/or peripheral routes can only survive by reducing the frequency of ights and increasing the prices, and even in this case private airlines may decide to abandon them. The traditional way of dealing with this problem in the EU has been to subsidize the population living in peripheral communities and/or to establish public service obligations (PSOs) on the airlines exploiting protected routes. In recent years, however, these policies have been largely criticized for their inconsistency and lack of transparency. First, in the EU there are no clear criteria of which routes should be protected and for which reasons. And second, the instruments used to nance thin routes and citizens mobility vary signi cantly across countries. This paper analyzes the e ects of two types of universal service policies applied in the Spanish airline market during the period First, we assess the e ects of the 50 % price discounts granted by the government to island residents when they travel in one of the 23 domestic routes that have an island as endpoints. And second, we asses the e ects of PSOs (price caps and frequency oors) established in 14 routes that connect Spanish islands among themselves (intra-island routes). To our knowledge, this is the rst empirical paper to study how the universal service regulations a ect prices and frequencies at the route level in an European country. 1 We believe our analysis o ers some light on what is the e ectiveness and who are the main bene ciaries of this type of regulation. The empirical literature has been proli c in analyzing the in uence of market structure variables on airline prices at the route level. Borenstein (1989), Brander and Zhang (1990), Berry et. al (1996), Brueckner and Spiller (1994), Dresner et al. (1996), Evans and Kessides (1993), Fisher and Kamerschen (2003), Fageda (2006), Hofer et al. (2008), Marín (1995), Morrison (2001), and Oum et al. (1993) estimate how prices are in uenced by route competition, airport dominance, or the presence of low-cost carriers. However, only Starkie and Starrs 1 Santana (2009) analyzes the e ect of PSOs on the productive e ciency of European airlines for the period She estimates the cost functions of European and US airlines and assesses the e ect of PSOs in each of these regions. Our paper, however, focuses on the e ects of discounts and PSOs at the route level. 1

3 (1984) have analyzed the prices of thin routes in Australia, and Bitzan and Junkwood (2006) in the US. On the other hand, Bilotkach et al. (2010), Borenstein and Netz (1999), Brueckner and Pai (2009), Pai (2010), Salvanes et al. (2005), Schipper et al. (2002) and Wei and Hansen (2007) examine the e ect of issues such as route distance, competition or aircraft size on the frequencies o ered. Following the methodology of these papers, we estimate price and frequency equations at the route level, taking into account the e ects of island residents price discounts and PSOs. Our rst contribution is to show that airlines set higher prices in routes that connect the Spanish mainland and the Canary and Balearic islands, where the government gives to the island residents a 50 % discount on the airfares. The intuition is that discounts reduce the demand elasticity of island residents, thus allowing airlines to set higher prices. As a consequence, island residents enjoy of lower prices due to the discount, but non-residents pay higher prices than in unregulated routes that have similar characteristics. We also obtain that price discounts are not an e ective policy for increasing the frequencies o ered, possibly because the additional demand they generate does not justify the use of more ights, or because airlines have di culties in obtaining additional slots. This rigidity in the transportation capacity may also explain why routes with price discounts have higher prices. The second contribution of our analysis is that the price caps and frequency oors set in intra-island routes lead to lower prices and higher frequencies than those encountered in unregulated domestic routes of similar characteristics. This result suggest that the authorities are "over-protecting" the passengers of these routes, in the sense that they receive better price and frequency conditions than the passengers of unprotected routes. This situation might respond to some policy objectives such as regional development and social cohesion, which can be very important but that are not the focus of our study. The rest of the paper is organized as follows. Section 2 reviews the literature on universal service obligations in the airline industry and describes the Spanish regulatory framework. Section 3 explains the estimation strategy, Section 4 presents the data, and Section 5 shows the main results. Finally, section 6 summarizes our conclusions. 2

4 2 Literature review and Spanish regulation 2.1 Literature review Before the liberalization reform initiated in the nineties, universal services in network industries were provided by public or regulated monopolies and nanced through subsidies from the public budget and through cross-subsidies from pro table to unpro table consumers. For example, in the telecommunications sector, uniform prices involved a cross-subsidization from high to low-cost regions and from long distance to local calls. In the postal sector, loss-making public companies received direct transfers from the public budget. In air transportation, thick routes subsidized remote/unpro table routes. After the liberalization these nancing mechanisms became unsustainable and new universal service policies were developed to counterbalance the negative e ect of the reform on some consumers groups and regions. Governments de ne the basic services that must be guaranteed to the whole population, select the operators in charge of providing them (universal service provider), and the instruments necessary to nance these operators. Next, we brie y review the economic literature that studies these three regulatory problems. The de nition of the basic universal services that must be made available to all citizens is a controversial issue. Usually the services o ered depend on the social preferences and the economic situation of countries. Cremer (2009, p. 271) has pointed out that the main problem for de ning the universal services is to determine when the social bene ts generated by these services justify their costs and, more importantly, the restriction in competition they usually entail. Following Cremer et al. (2001) and Cremer (2009) we can identify some economic justi cations for facilitating to all the population the access to air transportation: (1) Redistribution of income: universal service policies are an alternative redistributive mechanism to taxes and direct transfers; (2) Network externalities: by subsidizing some passengers others might enjoy additional frequencies, and by subsidizing some routes others might receive more tra c; (3) The public good: a national network of air transportation enhances social cohesion and equity; and (4) Regional policy: prices and frequencies can be regulated to facilitate regional development. A number of papers have studied the method for determining the rms in charge of pro- 3

5 viding the PSOs and the subsidy they must receive. Most studies consider the use of auctions (Anton et al., 1998; Chone et al., 2000; Sorana, 2000). In the case of air transportation, Williams (2005) analyzes the tendering system implemented in Norway. It is interesting to say that this type of tenders are not used in Spain because there is free entry in the routes protected with PSOs. A third group of papers focus on the economic distortions generated by the PSO nancing mechanisms (Armstrong, 2001; Calzada, 2009; Mirabel el al., 2009; Valletti et al., 2002). For the particular case of air transportation, Nolan et al. (2005) examine the social welfare implications of di erent regulations: direct subsidies, protected route packages, and revenue guarantees. In general, these papers distinguish two main forms of nancing the rm in charge of providing the service: cross-subsidies from one group of consumers to another, and direct subsidies to the rm. The second option is generally considered more e cient and transparent. Subsidies can be nanced through the public budget or by all airlines through a universal service fund. Finally, very little attention has been paid to nd the optimal universal service nancing mechanism. Billette de Villemeur (2004) shows that optimal allocations of price and frequency in the airline industry can be reached by means of a price-cap constraint that depends on the frequency of the service. In Appendix A we present an extension of this model that is useful to analyse the e ects of price discounts and PSOs. 2.2 Universal service obligations in the Spanish airline market In the EU, universal service policies are used to promote the mobility of the population in remote and peripheral areas. 2 France, Italy, Portugal, Spain and the UK establish discount schemes that cover the residents of selected islands and regions. For example, residents of Madeira and Azores in Portugal and Sardinia in Italy bene t from price discounts. Some countries like France, Irish Republic, Spain and Norway complement this direct social aid with public service obligations imposed on airlines serving peripheral regions or thin routes. However, other countries like UK scarcely use them. In most cases, protected routes must o er a minimum daily service frequency and/or satisfy speci c timetable obligations. In addition, 2 For a detailed explanation see Santana (2009, Table 1) and Williams and Pagliari (2004, Tables 2 and 4). 4

6 usually governments also x a maximum fare. As a result, the amount of subsidy given per oneway journey varies widely both between and within countries. For this reason, some authors such as Williams and Pagliari (2004) and Williams (2005) claim that usually PSOs respond to local political pressure and bear little relation to issues a ecting the periphery, economic development, and the availability of other transportation models. In Spain, in recent decades the government has implemented several measures to promote the mobility of island residents and the residents of the cities of Ceuta and Melilla. Next we summarize those that a ect air transportation: Residents discount scheme.- Between 2001 and 2004 residents of Canary and Balearic Islands and the cities of Ceuta and Melilla enjoyed a 33 % discount, nanced by public funds. From 2004 to 2007 the discount increased progressively from 33 % to 50 %. 3 According to the sector authorities, this measure has implied an increase in the yearly subsidy o ered to island residents from 173 to 371 million euros. 4 Subsidies of airport fees.- Airport fees on domestic routes that link the mainland and the islands are about 40 % lower than on the rest of domestic routes. Moreover, airport fees on intra-island routes are almost ve times cheaper than on the rest of domestic routes. 5 Public Service Obligations.- 13 intra-island routes in Canary islands and 3 intra-island routes in Balearic islands are subject to PSOs that guarantee the continuity, frequency, capacity, quality and a ordability of the service. On November 2003, the Spanish government established the following conditions for operating these routes: 1. Timetable requirements. Airlines must provide the service from 7h to 9h in the morning. The return at night must be provided from 20h to 22h, depending on the route. 2. Frequency oors. Each route must operate several ights per day. For example, between Mallorca and Menorca at least 4 daily ights must be o ered in the winter and 5 daily 3 The discount only a ects citizens of the EU and the European Economic Area who can prove that are residents in the Spanish islands or in Ceuta and Melilla. 4 See the debate about these subsidies between the Ministry responsible of the sector, and the regional authorities of Balearic Islands in: 5 This information has been obtained from the Spanish airport operator (AENA). 5

7 ights in the summer. Between Gran Canaria and Tenerife, at least 12 daily ights must be o ered in the winter and at least 14 in the summer. 3. Seating capacity. Airlines must o er minimum levels of seating capacity. For example, on the Mallorca-Menorca route they must o er at least 63,000 seats in winter, and on the Gran Canaria-Tenerife route at least 295,000 seats in winter. Note that in comparison to other European countries, PSOs in Spain are applied to routes with relatively high demand (the population and the tourism in some routes is signi cant). 4. Price caps. Fares must not exceed 82 euros for each round of the trip between Mallorca and Ibiza and Mallorca and Menorca. On the Canary routes, fares in the 13 routes protected vary from 56 euros between Gran Canaria and Tenerife to 94 euros between La Palma and Lanzarote. 6 An important feature of the Spanish regime is that serving a route with PSOs does not imply any exclusivity right to the airlines. Moreover, airlines don t receive any public subsidy for o ering these PSOs. In spite of this, quite often protected routes are dominated by one operator. In the Balearics, Air Nostrum, a regional airline owned by Iberia, operates most of the ights. In the Canaries, the main operator is Binter, an airline that specializes in intraisland ights. In this paper we show that the pro tability of these routes might be due to the increase of tra c generated by the discounts given to island residents. 3 Estimation strategy This section analyzes the e ects of the universal service policies implemented in the Spanish airline market during the period Following the existing literature, we estimate price and frequency equations that include explanatory variables related to demand, distance and route competition. However, the novelty of our price and frequency equations is the inclusion of 6 Since 2003, these fares have been updated each year in accordance with the Retail Price Index and the adjustment in airport fees. In 2008, the Spanish government modi ed the pricing regime to promote entry. In the new regulation, for at most 50 % of the seats occupied the maximum fares are substituted by reference fares. This implies that airlines can x a price above 82 euros for some passengers if the average price is not higher than this gure. See Ministry of Fomento, FOM/1085/2008, April 7th. 6

8 two variables for island residents discounts and PSOs that measure the impact of the universal service policies. The equations we estimate are the following: Pricing equation. We call p kt the price of route k at period t. Taking this into account, the price equation we estimate is the following: p kt = b 0 + b 1 X kt + b 2 Dist k + b 3 HHI kt + b 4 D discount kt + b 5 D pso kt +b 6 D summer t + b 7 T imet rend t + e kt ; (1) where X kt is the Demand, Dist k is the Distance that separates the route s origin and destination airports, HHI kt is Route Concentration (Hirschman Her ndahl index), D discount kt is a dummy variable that interacts with the price discount granted by the government to island residents, D pso kt is a dummy variable that takes value 1 in routes with PSOs, D summer t is a dummy variable that takes value 1 in the summer session, and T imet rend t is a variable that accounts for changes over time not captured by the variables considered in the empirical model. Frequency equation. We call f kt the weekly frequency o ered by all airlines in route k at period t. The estimation of the frequency takes the following form: f kt = c 0 + c 1 P op kt + c 2 GDP c kt + c 3 T our kt + c 4 HHI kt + c 5 Dist k +c 6 D discount kt + c 7 D pso kt + c 8 D summer t + c 9 T imet rend t + e kt ; (2) where P op kt is the Population, GDP c kt is the Gross Domestic Product per capita, and T our kt is Tourism at both endpoints of the route. Next, we discuss the expected e ect of the explanatory variables in the price and frequency equations. Demand. The expected sign of the coe cient of this variable is ambiguous. Intense tra c in a route makes it possible to gain density economies, as airlines can use bigger planes at higher load factors and optimize the use of the crew. In a competitive environment this should lead to lower prices. However, more tra c might lead to higher mark-ups over costs if capacity constraints are present. 7 7 Since the seminal paper of Caves et al. (1984), it is frequent to distinguish between density and scale 7

9 Note that at the route level prices and demand can be determined simultaneously. In order to avoid any endogenous bias in the price equation we consider three instruments for demand: the mean population in the route s origin and destination provinces (Pop); Gross Domestic Product per capita (GDPc); and Tourism (Tour) at the region level. In the frequency equation we adopt a di erent strategy to capture the in uence of demand. An estimation that regress frequency against demand displays a R 2 above 0.90, which might re ect an over-identi cation of the model that distorts the individual interpretation of the other explanatory variables. In this situation it is unlikely that an instrument variable procedure can correct this over-identi cation. Thus, we prefer to include the variables Population, GDP per capita and Tourism directly as explanatory variables. We expect a positive sign of the coe cient of these three variables associated with demand in the frequency equation. 8 Distance. Route s length is a major determinant of airlines costs and we expect that its coe cient in the price equation will be positive and lower than one. This would imply that the costs increase less than proportionally with an increase in kilometers own. Long-haul routes involve higher average speeds, less intense consumption of fuel, and lower airport charges per kilometer. 9 On the contrary, we expect a negative relationship between frequency and route s distance. In longer routes airlines may prefer to reduce frequencies and use bigger planes whose e ciency increases with distance. In addition, since in long-haul routes intermodal competition with cars, trains and ships is soft, airlines may o er less frequencies. Route concentration. The Hirschman Her ndahl index at the route level re ects how competition a ects the price. We expect the coe cient associated to this variable to be positive in the pricing equation, since less competition should imply higher prices. In the case of the economies. The rst is related to the decrease in average costs from increasing tra c in the route, since companies can use larger aircrafts at higher load factors. By contrast, scale economies shows the decrease in average costs from both increasing tra c in the route and the number of routes served. 8 Note that the proportion of tourists is higher in routes with an island as endpoint. As a result, the correlation between the variables of discounts and tourism intensity is high (see data from the Spanish Institute of Tourism Studies on the number of international tourists per capita that arrive to Spanish regions). Unfortunately, we are unable to construct a variable of tourism intensity for the summer and the winter season, which would help to capture more appropriately the e ect of tourism. 9 We obtain similar results when we use price per kilometer as dependent variable. In this case, the coe cient associated with the variable distance is negative, and should be interpreted in the same line as a positive sign and coe cient lower than one in the regression that uses prices as dependent variable. 8

10 frequency equation, by contrast, we expect a negative sign, since airlines shall o er less ights the weaker is the competition in the route. Remark that route concentration can be determined simultaneously with prices and frequencies. To account for a possible bias due to the endogeneity of this variable we follow Fageda (2006) and instrument the Hirschman Her ndahl index (HHI) at the route level through the HHI at the airport level. This variable is constructed by calculating the concentration index of airlines departures at the airports (we consider the mean value of the HHI at both the origin and the destination airports). This formulation could carry an endogeneity bias if concentration levels depend on the pricing choices of rms. However, we consider that this e ect is greatly diluted because pricing and frequency choices refer to the route level while concentration at the airport level refers to all the routes departing from a given airport. Island residents price discounts. Our predictions about the e ect of discounts on prices and frequencies are based on the theoretical analysis presented in Appendix A. The expected e ect of the government s subsidy to island residents is an increase in the price set by airlines. The reason is that discounts make the demand of island residents less elastic and as a consequence airlines can set higher mark-ups. The discount should also increase the tra c in the route, although in our price equation this e ect is captured by the demand variable. The theoretical model also shows that the frequency should be higher in the routes where island residents are subsidized because discounts increases consumption. Notice, however, that this e ect could not take place if there are capacity restrictions such limited slots. Public services obligations. In the Spanish market, airlines operating the service in routes that connect two islands are subject to a price cap and a minimum frequency requirement. In Appendix A we analyze the e ects of these two types of public service obligations (PSOs). Notice, however, that we should expect the coe cient of this dummy variable to be nonsigni cant if the regulator xes the maximum price and the minimum frequency equal to those of non-regulated routes. 4 The data We have data for 86 domestic routes. Of these, 23 are routes that link islands with the mainland and 14 routes that link two islands (intra-island routes). Data refer to the route that has as 9

11 origin the larger airport. Thus, for example, we include the route Madrid-Barcelona-Madrid but not the route Barcelona-Madrid-Barcelona. Since airlines supply is nearly identical in both directions, treating airlines services on a given route as directional would not add more to the analysis. The frequency of the data is semi-annual, as we di erentiate between the summer and the winter seasons in a time period that starts in the winter of 2001 and nishes in the winter of As we use lags of some variables the data for the initial periods are lost. Additionally, our panel is unbalanced, either because data was not available for some periods or because there was no air tra c. As a result, we have 1092 observations in the pricing regression and 1103 observations in the frequency regression. Information on prices (p) have been collected each half of the year in the period using two procedures. For all routes except intra-island routes of Canary Islands, we consider the lowest mean round trip price charged by all airlines operating the route weighted by their corresponding market share. The data have been obtained from airline s web sites using a homogeneous procedure: information has been collected one month before travelling; the price refers to the rst trip of the week, and the return is on Sunday. Prices are corrected for in ation. For intra-islands routes of Canary Islands we don t have information about the prices set by the airlines. However, we use the price cap established by the government, which is published in the O cial Journal of the European Union. This implies that the prices e ectively charged by airlines may be lower that the ones we consider. As we argue later, we consider that this approach is useful to asses if the prices xed by the government di er from those established by airlines in unregulated routes that have similar features. The frequency variable (f) shows the weekly number of ights o ered by airlines for each route. This information has been obtained from the web site of the O cial Airlines Guide (OAG), and it refers to the same sample week as the prices. Data for intra-canary routes come from RDC Aviation Limited (capstats data). Data of demand (X) refers to the number of passengers carried by airlines on the route, including direct and connecting tra c. Information has been obtained from the web site of the Spanish Airports and Air Navigation agency (AENA). 10

12 Population (P op) is constructed as the mean population in the route s origin and destination provinces (NUTS 3). 10 Data have been obtained from the Spanish National Statistics Institute (INE). We use data for population at NUTS 3 level rather than NUTS 2 level, because this variable captures more accurately the size of the urban agglomeration close to the airport. Gross domestic product per capita (GDP c) is the mean GDP per capita in the route s origin and destination regions (NUTS 2). Data have been obtained from the Spanish National Statistics Institute (INE). We use this variable at the region level because the information is not available at the province level for the whole period analyzed. The variable of tourism (T our) refers to the percentage of employment in region s hotels and restaurants. Data have been obtained from Cambridge Econometrics (European regional database publication). Information at the province level is not available. Distance (Dist) is the number of kilometres that separates the route s origin and destination airports. Data has been collected from the WebFlyer site. The Her ndahl-hirschman Index at the route level (HHI) is computed as the sum of the square market share ight frequencies of airlines operating the route. Data on departures of each airline on each route have been obtained from the O cial Airlines Guide (OAG) web site and RDC Aviation limited. The rst variable related to the Spanish universal service policy is island residents price discounts (D discount ). This variable takes the value 0 if the route is not a ected by the discount, and it takes the level of the discount (e.g or 0.50) if it is a ected by the discount. Island residents enjoy a price discount nanced by the government in 23 domestic routes that connect the islands and the mainland and in 14 intra-island routes. As explained in Section 2, in the period the discount was of the 33 %, from it increased progressively from 33 % to 50 %, and since then it has not changed. The dummy for public service obligations (D pso ) re ects the presence of price caps and frequency oors in the route. More speci cally, it takes the value 1 for intra-island routes in 10 According to the Statistical classi cation of Eurostat, a NUTS 2 area should have a range of population between 800,000 and 3,000,000 inhabitants, while NUTS 3 areas should have a range of population 150,000 and 800,000 inhabitants. In practice, the statistical territorial units are de ned in terms of the existing administrative units in the Member States and don t necessarily meet this population range. In Spain, NUTS 2 are Comunidades Autónomas and NUTS 3 are provincias. 11

13 the Canarias since the winter of 2001 and for intra-island routes in Balearics since the summer of 2004, when the obligations were rst imposed. This dummy a ects 14 routes. The dummy variable for summer (D summer ) is constructed taking into account that the summer season starts in the last Sunday of March and nished in the last Saturday of October, both inclusive. Finally, the Time Trend (T imet rend) is constructed as an index that takes the value 1 in 2001, the value 2 in 2002 and so on. Table 1 shows the mean values of the variables used in the empirical analysis for three sub-groups of routes: 1) Routes with no islands as endpoints (unregulated); 2) Routes that link islands with the mainland (routes with price discounts); and 3) Intra-island routes (routes with price discounts and public services obligations). This information reveals that the price per kilometer and the frequencies in routes with no islands as endpoints are similar. However, routes that link islands with the mainland are longer and thicker. Intra-island routes show the lowest mean price and the highest number of frequencies, but are shorter and thinner. As a consequence, these routes have a higher price per kilometer. Note that the number of rms operating in a route is highly dependent on the tra c volume that this route can generate. A number of routes regulated with PSOs have several operators, since these routes are not subjet to any entry regulation. This implies that their tra c is su ciently high and that they can be pro table. On the other hand, the volume of tra c in several unregulated routes is not particularly high. The last table only provides a rough picture of the relation between the variables. In the next section we undertake an econometric multivariate analysis in order to assess more accurately the e ects created by the universal service policy. 5 Estimation and results We estimate the pricing and the frequency equations using the Two-Stage Least Square estimator (2SLS-IV). The reason is that we consider that in the pricing equation the demand and route concentration may be endogenous, and in the frequency equation route concentration can also be endogenous. The instrument suitability tests, the partial R 2 of the rst stage regression and the Hansen s J test of the possible endogeneity of the instruments, show a high correlation between the 12

14 Variable Routes with no islands as endpoints Routes that link islands and mainland Intra island routes All routes Prices (euros) (90.37) (129.80) (28.70) (101.26) Prices per kilometre (euros) 0.39 (0.20) 0.39 (0.30) 0.83 (0.21) 0.46 (0.28) Frequency (weekly flights) (62.10) (43.05) (38.59) (54.22) Demand (number passengers) 191,664 (326,403) 266,168 (262,435) 130,556 (111,379) ( ) Distance (kilometres) (180.04) (723.31) (57.56) (474.56) HHI (Hirschman Herfindahl Index) 0.80 (0.24) 0.53 (0.23) 0.87 (0.19) 0.73 (0.27) Population (inhabitants) 3,024,522 (831,961) 2,534,382 (957,774) 978,640 (46,011) ( ) GDP per capita (euros) 22,417 (3,401) 22,425 (3,060) 19,399 (2,364) ( ) Tourism (percentage of employment in hotels & 0.06 (0.009) 0.12 (0.06) 0.14 (0.02) 0.09 (.05) restaurants in 1/100 ) Number of routes Note: Standard errors are shown in parenthesis Figure 1: Mean values of the variables used in the empirical analysis variables instrumented and the instruments 11, and indicates the exogeneity of the instruments. Furthermore, the LM test shows that the equations are not under-identi ed. Note also that we compute standard errors that are robust to any bias from heteroskedasticity. Additionally, we adjust our estimates by clustering observations from the same route. Our estimation procedure does not take into account the panel data nature of the sample. The use of a xed-e ects model is not appropriate in our context since this technique drops anything that is time-invariant from the model, such as route distance or being an island. Nor is a random-e ects model appropriate because the individual e ects related to routes are probably correlated with the error term, as indicated by the Hausman test. Finally, we can t use the Hausman-Taylor estimator because it assumes that all explanatory variables are exogenous. We report two additional tests in the estimates of the price and frequency equations (Tables 2 and 3). First, the augmented Dickey-Fuller Test whose null hypothesis is that the dependent variable contains a unit root. Second, the Wooldridge test for autocorrelation in panel data whose null hypothesis is no rst-order autocorrelation. The unit root test shows that the dependent variable is stationary so that we do not need to use cointegration techniques. On 11 Instruments are Pop., GDPc, and Tourism for Demand, and HHI at airport level for HHI at route level. 13

15 the other hand, after including the lag of the dependent variable as a regressor we don t nd a problem of autocorrelation. Both the price and the frequency equations are estimated using the whole sample and sub-sets of routes. Recall that we don t have information about the price of intra-canarian routes and that for this reason we consider the price cap set by the government. To check the implications of this approach we report the estimates of the price equation excluding intra- Canaries routes, obtaining similar results than when we use the whole sample. Additionally, we present separately the results of the sub-sample of routes that have an island in one of the endpoints. As a result, we can more accurately di erentiate between the e ects created by price caps and frequency oors and those of residents discounts. Notice that in some routes with islands as endpoints competition from ships is relevant, especially in intra-island routes and in the routes that connect the Mediterranean villages and the Balearic islands. 12 The estimates of the pricing equation are shown in Table 2. We nd that the demand variable is not statistically signi cant. A possible explanation is that the density economies generated by demand are compensated by the airlines market power and by capacity constraints. The distance variable is statistically signi cant and the sign of its coe cient is positive and lower than one, as expected. The coe cient of the variable route concentration is positive and statistically signi cant, indicating that prices are higher when competition is softer. Furthermore, prices are higher in the summer when tourism is more intense, and they have decreased along time during the period studied. The main interest of our result is that airlines set higher prices in routes where the government gives price discounts to island residents than in the rest of domestic routes. Our interpretation of this nding is that discounts reduce the demand elasticity, allowing airlines to set higher prices. This conclusion is supported by our estimates of the price-elasticity of demand that are presented in Appendix B. Indeed, the estimated demand elasticity for all routes in our sample is -1.12, while it is just for routes with islands as endpoints. Remark that 12 To account for intermodal competition, we have made regressions excluding short-haul routes within the mainland but we nd small changes in the results. We have also made additional regressions that include as explanatory variable a dummy for the routes with high-speed train services. This variable is not statistically signi cant and does not alter results for the rest of variables. 14

16 Explanatory Variables All sample All sample excluding Routes with an intra Canarian routes island as endpoint Demand (Q) ( ) ( ) ( ) distance (dist) 0.09 (0.014)*** 0.11 (0.01)*** 0.09 (0.02)*** HHI route (42.68)*** (48.18)*** (152.49)*** D discount (28.13)*** (31.85)*** D pso (13.70)*** (18.32)*** (34.95)*** D Summer (7.46)*** (8.62)*** (18.13)*** Time Trend (T) 4.81 (1.34)*** 5.27 (1.58)*** 1.43 (3.50) lag(prices) 0.18 (0.07)*** 0.14 (0.06)** 0.23 (0.16) Intercept (41.28) (47.16) (140.75) N R 2 F (Joint Significance) Chow Test (H 0 : No structural differences between routes with islands as endpoints and without islands as endpoints) Augmented Dickey Fuller Test (H 0 : the dependent variable contains a unit root) Wooldridge test for autocorrelation in panel data (H 0 : no first order autocorrelation) Tests of instruments: Partial R 2 : X Partial R 2 : HHI airport Underidentification LM test (H 0 : Equation is underidentified) Hansen J test *** *** *** 4.26*** 19.95*** *** *** *** ** *** (H 0 : Instruments Exogenous) Note 1: Standard errors in parenthesis (robust to heteroscedasticity) Note 2: Statistical significance at 1% (***), 5% (**), 10% (*) Note 3: Instruments for the demand and route concentration variables are the following: GDP per capita, population and tourism per capita, and the concentration index at the airport level Figure 2: Pricing equation estimates (2SLS-IV) this result has important implications for the universal service policy applied in the sector, as it indicates that the subsidies given to island residents could partly be pass through to airlines, via the increase of airfares. Unfortunately, we don t have information about the percentage of island residents in each route, which in practice can di er a lot from one route to the other. As a result, we are unable to analyze how this percentage a ects the pricing strategy of the rm. Our conjecture is that the higher the percentage is the higher the price increase in the route should be. A second nding of the price equation is that intra-island routes subject to PSOs exhibit lower prices than other unregulated domestic routes that have similar features. Although we can t verify it, we believe that the sustainability of air services in these routes can be 15

17 attained through the demand increases generated by island residents discounts. That is, the government could be indirectly subsidizing the airlines that operate these routes by fostering the tra c through the price discount. Finally, recall that all routes that have an island as an endpoint also bene t from lower airport fees. Our empirical analysis accounts for domestic routes with no islands as endpoints (not regulated), routes that link the mainland and the islands (routes with price discounts and subsidized airport fees) and intra-island routes (routes with price discounts, public services obligations and subsidized airport fees). We are therefore not able to empirically disentangle the e ect of price discounts and public services obligations from the e ect of lower airport fees. However, we expect that lower airport fees of these routes moderate the price increases generated by price discounts and increase the pro tability of routes regulated with PSOs. The results for the frequency equation are show in Table 3. First note that the lag of the dependent variable shows a strong signi cant e ect, which implies a high inertia in the frequency choices of airlines. In this sense, it is important to mention that the system that currently is used in Spain to allocate the slots o ers very little exibility to airlines to modify their frequencies. The coe cient of the population variable is positive and statistically signi cant when we consider the whole sample. The coe cient of the GDP per capita variable is also positive but not statistically signi cant. As anticipated, we also obtain that airlines reduce frequencies on long and concentrated routes. Furthermore, airlines o er slightly higher frequencies in the summer season (although not in a statistically signi cant way) and, contrary to the pricing equation, we do not nd a clear time trend for the period Importantly, the coe cient of the variable for island residents discounts is not statistically signi cant in the frequency equation. This result is contrary to the prediction of our theoretical analysis that routes bene ting from discounts have a higher demand and consequently o er more frequencies. One explanation could be that the correlation between the variables for discounts and tourism intensity distorts the individual identi cation of the variable of discounts. However, this variable remains highly non-signi cant when we exclude Tourism as a regressor. Another possible explanation is that in Spain airlines have few opportunities to increase route frequencies due to airport congestion and the restrictive system of slot allocation. Additionally, a higher frequency implies additional xed costs (landing fees, gate renting, etc.) 16

18 Explanatory Variables All Sample Routes with an island as endpoint Population (pop) 2.27e 06 (1.14e 06)** 1.80e 06 (1.43e 06) GDP per capita (GDPc) ( ) (0.0004) Tourism (21.74) (28.72) distance (dist) (0.0013)** (0.002) HHI (5.099*** (14.94)*** D discount 1.89 (4.16) D pso 8.81 (3.87)** (6.34)** D Summer 0.12 (0.41) 1.21 (0.85) Time Trend (T) 0.66 (0.54) 0.07 (0.49) lag(frequency) 0.88 (0.009)*** 0.87 (0.05)*** Intercept (5.27)*** (15.88)** N R 2 F (Joint Significance) Chow Test (H 0 : No structural differences between routes with islands as endpoints and without islands as endpoints) Augmented Dickey Fuller Test (H 0 : the dependent variable contains a unit root) Wooldridge test for autocorrelation in panel data (H 0 : no first order autocorrelation) Tests of instruments: Partial R 2 : HHI airport Underidentification LM test *** 18.15*** *** 4.11*** (H 0 : Equation is underidentified) Note 1: Standard errors in parenthesis (robust to heteroscedasticity) Note 2: Stastistical significance at 1% (***), 5% (**), 10% (*) Note 3: Instrument for route concentration is the concentration at the airpòrt level *** 10.54* *** Figure 3: Frequency equation estimates (2SLS-IV) and reduces the opportunity of exploiting density economies through the use of larger aircrafts with higher load factors. In this context, the tra c increase generated by discounts might be adjusted via price increases, which is consistent with our ndings for the price equation. Finally, we obtain that frequencies in intra-island routes protected with frequencies oors are higher than in other routes with similar features, even though the prices in these routes are lower due to the price caps. One intuition for this result is that the use of regional aircrafts in very short-haul routes allows optimizing the use of the planes and the crew. 13 Furthermore, although these routes are protected, they still have quite tra c (most of the travelers are island residents that enjoy of price discounts), which can make them pro table. In fact, in many cases 13 For a more detailed analysis of the e ects of regional aircrafts see Brueckner and Pai (2009). 17

19 the frequency o ered are higher than the ones established by the PSOs. We nish this discussion by showing the magnitude of the e ects of the overall universal service policy applied in Spain in terms of variations in prices and frequencies. Recall that non residents in islands pay the full price set by the airlines while island residents currently only pay the 50 % of the price and the government pays the other 50 % to the airlines. Controlling for several factors, Table 4 shows that prices of round trip ights that bene t from island residents discounts are about 55 euros higher than the prices of the rest of domestic routes. However, discounts don t a ect the frequencies. The prices in intra-island routes protected with price caps are 49 euros lower than those of rest of domestic routes. The e ect of frequency oors is also quite important since airlines o er about 8 more weekly ights in intra-island routes than in the rest of domestic routes. All Sample Routes with an island as endpoint Prices Frequency Prices Frequency Mean euros weekly fligths euros weekly flights Estimated elasticity for D discount Change in values for D discount =1 55 euros 0.91 flights Estimated elasticity for D pso Change in values for D pso = euros 8.27 flights euros flights Overall change in routes with price discounts and public service obligations (D discount =1 + D pso =1) 5.75 euros 7.36 flights Figure 4: Elasticies evaluated at sample means 6 Conclusions We have examined the e ects of the universal service policy applied in the Spanish airline market in the period , showing that prices were higher in the routes where the government gave price discounts to island residents. By contrast, we did not nd a clear e ect of discounts on frequencies. The logic behind this result is that discounts made island residents demand less elastic, allowing airlines to increase the price. On the other hand, the increase of tra c of island residents generated by the discount was not enough to induce airlines to increase their frequencies. Indeed, airlines costs increase when they o er more frequencies and 18

20 in addition they must acquire more slots, which is not always an option. These results suggest that in Spain part of the bene ts of island residents price discounts are transferred to the airlines via price increases, which harms both island residents and the rest of travellers. Therefore, the discounts seem to be working as a subsidy to airlines, in addition to their role of improving citizen mobility, social cohesion, and regional development. We are also able to show that price caps and frequency oors established in intra-island routes in the Canary and Balearic Islands led to lower prices and higher frequencies in these routes than in the rest of unregulated routes that have similar features. Interestingly, airlines can meet these public service obligations without receiving direct subsidies from the government, a part from lower airport fees. We argue that a key element for the viability of this regulatory system is the demand increases generated by island residents discounts. We conclude by noting that the analysis of the universal service policies in air transportation have received very little attention in spite of its relevance in the EU and in other countries. Further research should focus on issues like the competition e ects of these measures and on the mechanism used to nance them. 7 Appendix A This appendix extends the theoretical model of Billette de Villemeur (2004) to analyze the impact of tair-transportatition universal policies. Consider that the demand of a monopolist airline depends on the price p and on the ight frequency f. Consumers preferences on departure times are uniformly distributed over the time and their expected schedule delay cost is v > 0. Therefore, consumers average waiting-time cost is v=2f. 14 Imagine that 0 d 1 is the price discount granted by the government to all travelers and that it is nanced with public funds. Moreover, we call X the total tra c and S(X) the consumers gross surplus generated by the ight. Taking this into account, passengers demand takes the following form: X(p; f) = argmax X fs(x) (p(1 d) + v )Xg: (3) 2f The airline faces a cost C(K) for each ight, where K is the capacity of the aircraft (number 14 The schedule delay cost is the di erence between the preferred and the actual time of the ight. Frequency is a key variable to reduce it. 19

21 of seats). Indeed, the size of the aircraft determines the querosene costs. In the long run the aircraft s capacity is adjusted to the total number of passengers and therefore it is satis ed that X = Kf. The airline s costs can be reduced with a subsidy on the airport fees s 0, which is also nanced with public funds. Moreover, the airline has a xed cost F, that is independent of aircraft capacity. Using this information, the airline pro t function is: (p; f) = px(p; f) f[c(x=f) s] F: (4) In addition to the price discount and the costs subsidy, the government can protect the routes with thin demand by setting public service obligations (PSOs). 15 impose a price cap, p p, and a frequency oor, v 2f v.16 In particular, it can In this case, the airline maximizes the pro t in (4) with respect to p and f, subject to the price cap and the frequency oor restrictions. The rst order conditions of the Lagrangian maximized by the = X(p; f) + (p C = (p C 1 = 0; (5) (C(K) s) + X v f C0 (K) + 2 = 0: (6) 2f 2 where 1 and 2 are the Lagrange multipliers associated to the price and frequency constraints, v = 2f 2 (1 Calling " = From (3) these conditions can be simpli ed by using the fact that the price elasticity of demand yields: p C 0 (K) p [( X 1 X )( 1 1 d ) + 2 X ] v 2f = ( X 1 X ) 1 " ; (7) = C(K) s K C 0 (K): (8) In order to interpret these results, consider rst that the two PSOs are not binding ( 1 = 0 and 2 = 0). In this case, the price and the frequency are determined as in the standard case de ned by Billette the Villemeur (2004): the airline establishes a mark-up over the price 15 Billette de Villemeur (2004) shows that a conveniently designed "price-and-frequency" cap constraint of the form p + v p can implement the second-best allocations for p and f. In our model, we consider separate 2f constraints for the price and the frequency to identify the main e ects of the Spanish regime. 16 This expression can be written as f v : That is, setting a frequency oor is equivalent to limiting 2v consumers average waiting time. 20