Analyzing competition between the high speed train and alternative modes: The case of the Madrid-Zaragoza-Barcelona corridor

Román, Concepción; Espino, Raquel; Martin, Juan Carlos Article Analyzing competition between the high speed train and alternative modes: The case of the Madrid-Zaragoza- Barcelona corridor Journal of Choice Modelling Provided in Cooperation with: Journal of Choice Modelling Suggested Citation: Román, Concepción; Espino, Raquel; Martin, Juan Carlos (2010) : Analyzing competition between the high speed train and alternative modes: The case of the Madrid-Zaragoza- Barcelona corridor, Journal of Choice Modelling, ISSN 1755-5345, University of Leeds, Institute for Transport Studies, Leeds, Vol. 3, Iss. 1, pp. 84-108 This Version is available at: https://hdl.handle.net/10419/66806 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by-nc/2.0/uk/ * Corresponding author, T: +34 928451796, F: + 34 928458183, [email protected] Ŧ T: +34 928458207, F: + 34 928458183, respin[email protected] § T: +34 928458189, F: + 34 928458183, [email protected]pgc.es Analyzing Competition between the High Speed Train and Alternative Modes. The Case of the Madrid-Zaragoza-Barcelona Corridor Concepción Román1,* Raquel Espino1,† Juan Carlos Martín1,§ 1Departamento de Análisis Económico Aplicado. Universidad de Las Palmas de Gran Canaria, 35017 Las Palmas G.C., Spain Received 14 February 2008, revised version received 2 October 2008, accepted 4 December 2009 Abstract While there have been many studies of urban travel demand, little attention has been paid to the analysis of inter-urban rail travel demand. Studies of inter-urban rail demand usually focus on assessment through the conventional cost benefit analysis of this type of investments, in which the emphasis is on the cost side. However, the analysis of the potential benefits, bearing in mind the intermodal competition, is usually neglected. This paper analyzes the potential competition of the high speed train (HST) with the main competing modes on the Madrid- Barcelona route, where a new HST infrastructure has been recently built. The analysis is based on the estimation of disaggregated Nested Logit models using information provided by travellers in the main corridors: Madrid-Zaragoza and Madrid-Barcelona. The utility specification considers the effect of the main level- of-service attributes as well as some latent variables on modal choice. We analyze demand response to various policy scenarios that consider the potential competition between HST and other modes as well as the willingness to pay for improved levels of service. The results highlight the low level of competition that the HST could exert over air transport services in Madrid-Barcelona corridor, showing that policy makers may have been very optimistic about the figures of traffic diversion from air that could be attained. Keywords: Intermodal Competition Stated Preference (SP), Mixed RP/SP data, HST Elasticities, Willingness to pay Journal of Choice Modelling, 3(1), pp. 84-108 www.jocm.org.uk Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 85 1 Introduction Trans-European Transport Networks constitute one of the basic policy instruments applied in the EU to achieve growth, competitiveness and employment. European transport policy has always called for an integrated approach combining, inter alia, measures to revitalize the rail sector, and special emphasis has been placed on the development of high speed train (HST) corridors. These corridors are characterized by dense flows of rail passengers at speeds of 300 km/hr between the principal cities of the EU. European policy makers have tried to revitalize railways, their aim being for HST routes to take virtually all traffic away from air sectors, with more convenient travel times between the core central business districts of the cities. A good example of this type of infrastructure is the new Madrid-Barcelona HST line. The line began operations in the Madrid-Zaragoza-Lleida corridor in 2003 and the complete Madrid-Barcelona line came into commercial service at the beginning of 2008, covering the 625 km between these two cities in under three hours,. Nationally, this line will connect with the Madrid-Seville HST, which entered service in 1992, and, internationally, will produce substantial reductions in travel time between the main cities of the Iberian Peninsula and the principal European cities once the extension to the French border is completed. Madrid and Barcelona are the two largest Spanish cities, with more than 5 and 3 million inhabitants, respectively. From a demographic perspective, Zaragoza (700,000 inhabitants) is the main city located on the corridor linking those two cities. Madrid and Barcelona are also important economic centres of Spain, and the air shuttle between these two cities constitutes one of the most important domestic markets in the world (4.7 million passengers in 2006). The main airlines operating in this market offer a total of more than sixty flights per day, making air transport an attractive alternative, especially in the business-trip segment. In addition, the airport facilities in both cities are well connected to public transport services. The impacts of investments in HSTs can be analyzed in a number of ways. However, at present, the majority of the projects are, in the best of the cases, only assessed at national level, and existing contributions differ regarding coverage and perspective. The papers fall into the following groups: general assessments (Laird et al. 2005; Martín 1997; Nash 1991; Sichelschmidt 1999; Short and Kopp 2005; van Exel et al. 2002; Vickerman 1997); evaluations of the economic profitability of particular corridors or areas (de Rus and Inglada (1993, 1997), for the HST Madrid- Seville; Levinson et al. (1997) for Los Angeles-San Francisco; de Rus and Román (2006) for the HST Madrid-Zaragoza-Barcelona; Steer Davies Gleave (2004) and Atkins (2004) for the case of the UK; de Rus and Nombela (2004) for the European Union; and Martín and Nombela (2007) for the case of Spain); assessments of the regional effects (Blum et al. 1997; Haynes 1997; Plassard 1994; Vickerman 1995); studies of the impacts on accessibility (Fröidh 2005; Gutiérrez et al. 1996; Gutiérrez 2001; Martín et al. 2004; Vickerman et al. 1999); and, finally, regarding intermodal competition, Combes and Linnemer (2000) study the impacts of the creation of a new infrastructure connecting two points and coexisting with old network infrastructure (like roads) using a game-theoretic approach. The analysis of the passengers’ perceptions of and preferences in interurban transport is not new in the literature. Discrete choice analyses, based on SP (Stated Preference), RP (Revealed Preference) or mixed data, are usually advocated by researchers as a proper methodology to assess and compare the preferences of passengers in the context of modal competition. Ahern and Tapley (2008) evaluated Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 86 the preferences and perceptions of bus and train passengers in Ireland using both SP and RP techniques. González-Savignat (2004) analyzed the potential of the high speed train to compete with the airline market. That author used SP techniques proposing a hypothetical market with relevant information, given that the high speed train alternative was not then available on the Madrid-Barcelona route. Ortúzar and Simonetti (2008 -. They used a factorial fractional design of four variables - travel time, fare, comfort, and service delay. They incorporated RP data, obtained in a previous study, including bus, train and airplane travellers. Mixed RP/SP models were estimated and compared with those obtained from the stated preference data alone. Rigas (2009) studied the characteristics and perceptions of leisure passengers identifying the effects on modal choice between boats and air. The analysis was conducted by constructing a Multinomial Logit model, in which the dependent variable was the probability of a passenger choosing to travel by boat. This paper contributes to the empirical literature on HST effects by analyzing the potential competition between the high speed train (HST) and the alternative modes in the Madrid-Zaragoza-Barcelona corridor, focusing both on modelling issues and on policy analysis. Effort is concentrated on the Madrid-Zaragoza and Madrid-Barcelona routes, where the HST could attract more traffic. In contrast with the existing literature, private transport modes (car as driver and car as passenger) are also included as competing alternatives. This is especially relevant in the case of the first route where traffic diversion basically comes from car and conventional trains1 while, in the second, the HST enters the market with the objective of diverting traffic from air transport. The inclusion of the different components of the total journey length (invehicle, waiting, and access+egress time) also represents a novel aspect of the model specification. The analysis is based on the estimation of disaggregated mode choice models using RP and mixed RP/SP datasets collected during 2004. At that time, the HST was already operating between Madrid and Zaragoza (where we collected only RP data), but rail services between Madrid and Barcelona were still provided by conventional trains; thus, in addition to RP data, we obtained SP information about the new alternative. In general, the joint use of RP/SP datasets exploits the advantages and overcomes the limitations of each type of data. RP data are based on individuals’ choices and allow researchers to characterize actual travel behaviour while SP data are based on individuals’ stated choice behaviour in hypothetical scenarios and are useful when the problem is to analyze the demand for new alternatives and/or measure the effect of latent variables and their interactions with other attributes. Our model specification considers how the main level-of-service attributes (such as travel cost, travel time components and service frequency) and latent variables2 (such as comfort and reliability) affect modal choice; the income variable (which is frequently not included in the specification of mode choice models) is included in the utility specification following the recommendations of the literature when the share of income spent in transport is no negligible. Substitution patterns among groups of alternatives are also analyzed through the estimation of Nested Logit models. Measures for the willingness to pay for improving the level of service, elasticity values and demand response to different policy scenarios are obtained. The rest of the paper is organized as follows. Section 2 presents the theoretical framework on which the research is based. Section 3 describes the main characteristics Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 87 of the databases used in the analysis and Section 4 provides the steps followed in the modelling process as well as estimation results. Model applications are shown in Section 5 and, finally, the main conclusions are presented in Section 6. 2 Theoretical Framework Discrete choice models represent the behaviour of individuals when they choose from a finite set of alternatives (e.g. mode choice for interurban trips) and are derived under the assumption of utility maximization by the decision maker. The theoretical basis for their econometric specification is the random utility theory (McFadden 1974), which states that the utility of alternative j for individual q has the expression: jq jq jq UV   (1) Where Vjq is the representative or systematic utility (observed by the analyst) of individual q for alternative Aj and jq  is a random term that includes unobserved effects. jq V depends on the observable attributes jq X  of alternative j as well as on the socio-economic characteristics of individual q. In fact, Vjq is the conditional indirect utility function that includes the role of the constraints faced by the individual when he chooses both the amount of continuous goods and one of the discrete alternatives which, following Lancaster (1966), is represented by a vector of characteristics (see Jara-Díaz 1998, for more details about the consumer theory of discrete choice). The dependent variable represents individual behaviour and is a discrete variable. We have, therefore, a probabilistic model. From the model estimation we can obtain the probability and distribution of the dependent variable for each individual observation. Hence the probability that individual q chooses alternative j is given by the expression: ( ) ( ( ) ) jq jq jq iq iq iq jq jq iq P P V V i j P V V i j                 (2) The derivation of choice probabilities will depend on the different assumptions formulated about the distribution of the unobserved portion of utility jq  . In this regard, the famous Multinomial Logit (MNL) Nested Logit (NL) models are obtained when jq  are iid extreme value and a type of generalized extreme value, respectively (see Train 2003 and Ortúzar and Willumsen 2001 for more details about the derivation of choice probabilities in random utility models). The use of RP/SP data to estimate choice models requires that the variances of the error terms in RP and SP satisfy the following expression (Ben-Akiva and Morikawa 1990): 2 2 2      (3) where  is an unknown parameter, and  and  are the error terms of the RP and SP utilities respectively. Hence, in order to mix the data the following utility functions for a given alternative j are postulated: Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 88 ( ) ( ) RP RP RP RP j j j j j j SP SP SP SP j j j j j j U V X Y U V X Z                      (4) where ,  and  are parameters to be estimated; RP j X and SP j X are common attributes to the RP and SP data sets, respectively; and RP j Y and SP j Z are attributes that only belong to the designated data set. Bradley and Daly (1997) proposed an estimation method based on the construction of an artificial NL structure where RP alternatives are placed just below the root and each SP alternative is placed in a single-alternative nest with a common scale parameter  . 3 The Data 3.1 Madrid-Zaragoza Corridor The analysis of demand in the Madrid-Zaragoza corridor is based on a revealed preference (RP) survey that gathered information about travel behaviour in the principal modes: car as driver, car as passenger, bus, high speed train (HST) and plane. The main interest was to analyze individual preferences in the market situation created after the introduction of the new HST line in this corridor. The survey was conducted during April and May 2004. Bus users were interviewed in the Avenida de America bus station while air transport users were approached at the corresponding boarding gates at Barajas Airport. People travelling by HST were interviewed on board the train, and finally car users were interviewed in the petrol stations strategically located on the national road A-II. The survey was administered to bus, HST and plane users by means of personal interviews, while car users were asked to complete a questionnaire and mail it back. In all cases, the questionnaire was divided into four sections of questions: identification data, trip information, household information and personal information. Trip information includes questions related not only to chosen alternative for the reference trip but also to available modes not chosen by the individual. We obtained a total of 226 valid observations. Table 1 shows the modal split in the sample for this corridor. Table 1. Modal split in the sample. Madrid-Zaragoza Mode Travellers % Car-driver 59 26.11 Car-passenger 17 7.52 Bus 57 25.22 HST 75 33.19 Plane 18 7.96 The dominant modes are car and HST with market shares of around 33%, followed by bus (25.22%) and plane (7.96%). Modal shares in the sample were determined by trying to replicate modal shares in the population, given the available information at the time the surveys were carried out and considering a maximum error of 10% Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 89 (Ortúzar and Willumsen 2001). The market share in the sample for the HST was increased (and the share of car reduced) under the assumption that the new alternative will capture traffic from the car alternatives. Table 2 shows the level-of-service attributes as well as the socioeconomic characteristics of the sample. Total travel time has been divided into its main components: access time, waiting time, in-vehicle-time, and egress time. It is worth highlighting that the total duration of the trip is similar for the HST and the car (212 minutes approximately) although there may be differences in the perception of these modes. While travelling by car provides a higher accessibility to travellers, it has the inconvenience of driving (for the car driver) during the whole trip. Total time by plane is about 15 minutes less than by HST, but access and waiting time (which are usually more negatively perceived by travellers) in this mode comprise nearly 72% of total travel time. 46% of the trips were made for work or education purposes and almost 56% of the individuals were men. We also observe differences in per capita weekly income, ranging from 208 € for bus users to 318 € for HST users; and expenditure rate3, which ranges from 1.56 for bus users to 2.46 for HST/plane users. 3.2 Madrid-Barcelona Corridor Demand analysis in the Madrid-Barcelona corridor is based on a mixed RP/SP database. RP data were obtained from a survey that gathered information about travel behaviour in the principal modes: car as driver, car as passenger, bus, conventional train and plane. The main interest was focused on analyzing individuals’ preferences in the future market situation that would be created after the introduction of the new HST line in this corridor. At the moment of data collection, the cities connected by the HST line were Madrid, Zaragoza and Lleida. At that time, the line connecting Barcelona and other cities in Catalonia was expected to be finished by the end of 2007. Finally, the HST line to Barcelona was inaugurated in February 2008. Our study also aimed to analyze the effect of the latent variables on mode choice decisions. Although latent variables are not usually included in the utility specification (because, in practice, their measurement is not straightforward), they may play an important role in individuals’ choices. SP experiments represent the appropriate tool to analyze these situations because the analyst has the opportunity to present a detailed description of these variables in the experiment. Therefore, plane users, who represented the main source of traffic diversion, were faced with a stated choice experiment between the plane (the dominant mode) and the new HST alternative. As all the participants in the exercise were actually travelling in the corridor, the notravelling option was not considered. The mixed or joint estimation method proposed by Bradley and Daly (1997), combining RP and SP data, enabled us to estimate the utility of the new alternative as well as the utility of the already existing options. Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 90 Table 2. Descriptive analysis of the sample. Madrid-Zaragoza Chosen mode Car driver Car passenger Bus HST Plane Total Choice 59 17 57 75 18 226 Availability 164 92 189 218 171 - LEVEL-OF-SERVICE ATTRIBUTES (Average per available alternative) Access time (minutes) - - 29 27 37 - Waiting time (minutes) - - 30 23 60 - In-vehicle-time (minutes) 213 208 256 129 57 - Egress time (minutes) - - 29 34 42 - Travel cost/Fuel (€) 26.80 15.04 12.84 43.81 69.62 - Toll (€) 2.90 1.60 - - - - Access cost (€) - - 3.29 3.33 6.91 - Egress cost (€) - - 3.80 5.23 8.20 - Headway (minutes) - - 60 75 658 - SOCIOECONOMIC CHARACTERISTICS (Classification per chosen mode) Trip purpose: work or education 20 (34%) 8 (47%) 14 (25%) 52 (69%) 10 (56%) 104 (46%) Trip purpose: Other 39 (66%) 9 (53%) 43 (75%) 23 (31%) 8 (44%) 122 (54%) Access+egress time <60' - - 47 (82%) 55 (73%) 16 (89%) 118 (52%) Access+egress time >60' - - 10 (18%) 20 (27%) 2 (11%) 32 (14%) Men 44 (75%) 7 (41%) 17 (30%) 51 (68%) 13 (72%) 132 (58%) Women 15 (25%) 10 (59%) 40 (70%) 24 (32%) 5 (28%) 94 (42%) Age (average) 37 35 32 38 34 36 Per capita weekly income (average €) 263.98 298.71 207.66 318.36 314.53 274.46 Expenditure rate (average) 2.10 2.25 1.56 2.46 2.46 2.13 Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 91 With regard to data gathering, we used the same procedure as in the corridor Madrid- Zaragoza with the exception of the SP experiment, which was conducted at the corresponding boarding gates at the airport. We used computers to interview plane users, which allowed us to gain realism and adapt the SP experiment to the individuals’ experience. Table 3 shows the sample’s modal split corresponding to the RP survey, where the plane was the dominant mode with a share of close to 67%. The descriptive analysis of the sample is shown in Table 4. In this corridor, total travel time by plane is substantially less than in the rest of the modes but again the proportion of access and waiting time is very high (nearly 70%). Car is the second fastest mode with a total travel time of 70 minutes less than the train. Almost 56% of trips were mandatory (work or education) and 54% of travellers were men. We also observed differences in per capita weekly income, ranging from 167 € for car passengers to 351 € for plane users. And finally, the expenditure rate ranges from 1.23 for car passengers to 2.81 for plane users. The SP survey included two latent variables: reliability and comfort. The former was included to account for the negative effect of delay over scheduled departure time and the latter to analyze the effect of having more space in plane seats. The experiment also included other typical level-of-service attributes, such as travel time, access time, travel cost and headway (time between two consecutive services), which helped us define the overall quality of the alternative. In order to gain realism, the levels assigned to some attributes in the SP exercise were customized to each respondent experience by pivoting the information provided by the RP questions around the reference alternative (the plane, in this case). Thus, the levels of travel cost (cv) and access+egress time (ta) were defined in terms (as plausible percentage variations according to the available information about future fares and access time for the HST) of the values experienced by the respondents, and the levels of the service headway varied with the departure time. In a recent research Rose et al. (2008) suggest the construction of D-efficient designs pivoting attribute levels around a reference alternative. Train and Wilson (2008) also analyze the dependence between the SP attributes and unobserved factors. Table 5 shows a detailed definition of the attribute levels included in the experiment. A main effect factorial fractional design consisting of six attributes (four defined at three levels and two at two levels) and nine scenarios for each alternative was created using the WINMINT4 software. A special code was created for that purpose, which allowed us to obtain RP and SP data in the same survey. Each respondent (the 295 plane users) was faced with nine choice sets that were created automatically by the program, obtaining a total of 2,655 SP observations. After removing 179 inconsistent observations5, we obtained a mixed RP/SP data base of 2,917 observations. Table 3. Modal split in the sample. Madrid-Barcelona Mode Travellers % Car-driver 38 8.62 Car-passenger 18 4.08 Bus 39 8.84 Train (conventional) 51 11.56 Plane 295 66.89 Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 98 5 Applications 5.1 Derivation of the Willingness To Pay Measures Measures of willingness to pay (WTP) express, in monetary terms, changes in the utility that are caused by changes in attributes. They are also referred to as the subjective value of a given attribute kj q (e.g. the subjective value of time), and are derived from the estimation of discrete choice models as the ratio between the marginal utility of this attribute and the marginal utility of travel cost, which coincides with minus the marginal utility of income / / j kj jj Vq Vc       . Specifications of the representative utility, introducing income (e.g. dividing travel cost by the expenditure rate), interactions and quadratic terms yield more complex specifications of the marginal utilities that could take different values for every individual in the sample. This kind of specification also made the computation of confidence intervals for the WTP measures more complex since their distribution is generally unknown and the use of simulation techniques is normally required (Espino et al. 2006b). Aggregate WTP were computed using the sample enumeration method, obtained as the average WTP for the individuals in the sample (see Ortúzar and Willumsen 2001 for more details about the application of this method). It should be pointed out that more complex specifications of the utility (e.g. random parameters) would need to simulate the distribution of the coefficients before applying the sample enumeration method. Table 8 shows the WTP measures obtained in the Madrid-Zaragoza corridor. In general, WTP for travel time savings is greater for mandatory trips (work or education) than for other trip purposes, and takes the highest value for plane users, followed by HST, car and bus. Access+egress time is less valued by individuals for whom this figure is below 60 minutes. These are individuals living in the capital city or surrounding vicinities, however this difference should be interpreted with caution because the parameter corresponding to the incremental term (interaction) was not significant in the estimation process. In this corridor, the relationship SVAT>SVTT>SVWT9 is satisfied for all modes. This means that individuals are more Table 8. Willingness to pay measures. Madrid-Zaragoza Subjective value of Car driver Car passenger Bus HST Plane travel time (€/hour) 25.28 20.54 19.10 25.68 34.22 - Work/education motive 36.13 33.24 29.81 38.89 51.18 - Other motive 13.31 11.19 10.53 14.19 20.66 access+egress time (€/hour) - - 22.76 30.50 41.14 - access+egress <60' - - 22.53 29.05 37.84 - access+egress>60' - - 23.32 33.48 46.06 waiting time (€/hour) - - - 9.14 20.24 headway (€/hour) - - 2.17 2.88 6.39 Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 99 willing to pay for saving access+egress time than for saving travel time, as happens in the majority of both interurban and urban trips. However, waiting time is less valued than travel time, in contrast to what happens in most urban trips, where waiting time (e.g. at the bus stop) is more related to the frequency of the service. For scheduled interurban trips, there is normally a minimum period of waiting time that is subject to the regulation imposed by the operation of the transport system (e.g. passengers must check in at the airport 40 minutes before departure of the flight), but in these cases individuals have the opportunity to undertake many different activities at the transport terminal (e.g. shopping, use a laptop, take meals or drinks, etc.) and do not have such a negative perception of the waiting time. In the Madrid- Barcelona corridor, WTP measures were obtained from a hybrid utility containing common and non-common RP/SP parameters as in equation (4) (see Louviere et al. 2000 for more details). If attributes are defined only for the SP case (i.e. comfort, and reliability), their parameters must be scaled by  . However, those corresponding to attributes measured in the RP data base do not need to be scaled even if they only appear in the SP utility (Cherchi and Ortúzar 2004). In this case, the sample enumeration method was only applied to individuals in the RP data base. Table 9 presents the WTP measures for the Madrid-Barcelona corridor. Once again, the WTP for travel time savings is higher for mandatory trips. We also found that the value increases as the level of comfort is lower. When the level of plane comfort is low, the subjective value of time is similar to that obtained for the HST users. However, if the comfort for plane travellers is increased, their WTP for reduced travel time is substantially lower. The relationship SVAT>SVTT>SVWT is also maintained for travellers in this corridor. As the duration of the journey in this corridor is higher, the perception of waiting time is less negative, and consequently the WTP for waiting time savings is lower. However, it is important to note that this attribute presented a very low significance and these figures must be interpreted with caution. We also obtained a high WTP for reductions in delay time, it being higher in the case of the HST than in that of the plane. In trips where departure times are scheduled and known in advance, delay time produces more disutility. In many of these cases passengers receive no compensation when delays occur, although it is becoming common practice for transport companies wishing to mitigate the inconveniences associated with delays to offer compensation programs as competitive strategies. Although it is beyond the scope of this paper, it may also be possible to analyze whether there are behavioural differences among travellers regarding the departure time (early versus late services). Thus, the importance of properly quantifying the WTP for more reliable transport services is highlighted. The omission of factors, such as delay, in the specification of the utility may bias other WTP measures related to travel time. In this regard, it is possible that the high figures for the value of time obtained in González-Savignat (2004) could be influenced by this factor. We also obtained the WTP for improvements in comfort in the plane alternative. In the SP experiment we aimed to define a level of comfort for the plane similar to that for HST travellers. In our model, comfort was specified as interacting with travel time, thus the WTP for improved comfort varies with the duration of the trip. This was 8.45 euros for trips of approximately one hour in-vehicle. Although this is not a very high figure compared with other WTP, the impact of comfort attributes on the perception of time is not negligible. Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 100 Table 9. Willingness to pay measures. Madrid-Barcelona Corridor Subjective value of Car driver Car passenger Bus Train Plane HST Comfort high Comfort low travel time (€/hour) 17.59 12.37 12.39 14.97 10.55 19.29 19.33 - Work/education motive 18.91 15.48 15.03 17.77 12.96 22.50 22.41 - Other motive 15.24 10.46 10.72 12.18 7.52 15.27 14.00 access+egress time (€/hour) - - 30.34 37.14 46.44 46.45 - access+egress <60' - - 25.96 30.69 40.13 39.50 - access+egress>60' - - 42.78 51.79 61.31 61.60 waiting time (€/hour) - - 6.75 7.98 10.17 - headway (€/hour) - - 2.64 3.12 3.98 3.92 delay (€/hour) - - - - 59.34 64.83 improving comfort from low to high (€) - - - - 8.54 - For mandatory trips, the WTP for travel time savings was substantially higher in the Madrid-Zaragoza corridor. In this corridor, business travellers presented an average weekly income 75€ higher than those in the Madrid-Barcelona corridor. Thus, for these individuals the marginal utility of income is lower and the WTP is consequently higher. 5.2 Elasticity Values The sample enumeration method was used to obtain the aggregated elasticities of the new HST alternative (Ortúzar and Willumsen 2001). Table 10 presents direct and cross elasticities of the probability of choosing the HST. These are computed as the arc elasticity considering 1% variation in the corresponding attribute. The same method could be applied for more complex models, like Mixed Logit, taking into consideration that parameters could be random variables and that probabilities are now integrals that must be approached using simulation. In the Madrid-Zaragoza corridor, cross elasticities were computed with respect to car attributes (travel time and travel cost) while in the Madrid-Barcelona corridor they referred to plane attributes. In all the cases analyzed, we obtained figures below 1, i.e. demand for the HST is inelastic. This means that a one percent increase, for example, in travel cost will reduce demand for the HST in a lower proportion. We have shown that demand for the HST is, in general, more elastic in the Madrid-Barcelona corridor the only exception being travel time. Cross elasticities with respect to car attributes (for Madrid-Zaragoza) are very low, and these figures are consistent with the policy analysis presented in the next section. Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 101 Table 10. Elasticity values for HST Direct elasticities of the probability of choosing HST Attribute Madrid-Zaragoza Madrid-Barcelona Travel cost -0.55 -0.72 Travel time -0.59 -0.38 Access+egress time -0.36 -0.44 Headway -0.05 -0.07 Cross elasticities of the probability of choosing HST Attribute Madrid-Zaragoza (with respect to car attributes) Madrid-Barcelona (with respect to plane attributes) Travel cost 0.12 0.7 Travel time 0.04 0.11 Access+egress time - 0.51 Headway - 0.01 Demand Response and Policy Analysis Demand response to the application of different policies is represented by the percentage change in the aggregate market share of alternative j with respect to the initial situation: 10 0100 jj j j PP PP     (7) Where 1 j P is the aggregate share of alternative j once the policy is applied, and 0 j P is the initial (base year) aggregate share of alternative j. Aggregate market shares are obtained as the average probabilities across the individuals in the sample (i.e. applying the sample enumeration method). We are mainly interested in analyzing the new HST’s potential competition with the alternative modes (especially car and plane). Table 11 shows the policy scenarios analyzed in the two corridors. All the policies refer to a base scenario, which represents the current situation, i.e. considering the new HST for Madrid-Zaragoza and the conventional train for Madrid-Barcelona. In the Madrid-Zaragoza corridor, these policies consider increments in travel cost for the car alternatives (scenarios 1, 2 and 3), and the combination of a 100% increase in car cost with a 25% reduction in HST travel time (scenario 4) and a 10% reduction in HST fares (scenario 5). Figure 3 shows the percentage variation in the predicted market shares with respect to the base situation resulting from the application of the different policies. We find that policies consisting only of penalizing the car alternatives (increasing their travel costs) do not produce substantial increases (in all the cases these figures were below 5%) in the market shares of the alternative public transport competitors. However, a 25% reduction in travel time for the HST in Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 102 Table 11. Policy scenarios ATTRIBUTE SCENARIOS BASE 1 2 3 4 5 6 Madrid-Zaragoza Car cost Actual +10% +50% +100% +100% +100% - HST cost Actual Actual Actual Actual Actual -10% - Delay (train/HST) 0 min 0 min 0 min 0 min 0 min 0 min - Access time (plane) Actual Actual Actual Actual Actual Actual - Waiting time (plane) Actual Actual Actual Actual Actual Actual - Travel time (HST) Actual (HST) Actual (HST) Actual (HST) Actual (HST) -25% Actual (HST) - Madrid-Barcelona Comfort Actual (level 0) Actual (level 0) More space and legroom (level 1) Actual (level 0) Actual (level 0) Actual (level 0) Actual (level 0) Delay (plane) 0 min 0 min 0 min 30 min 0 min 0 min 0 min Delay (train/HST) 0 min 0 min 0 min 0 min 10 min 0 min 0 min Access time (plane) Actual Actual Actual Actual Actual +10% Actual Waiting time (plane) Actual Actual Actual Actual Actual Actual +50% Travel time (train/HST) Actual (Conventional train) -50% (HST) -50% (HST) -50% (HST) -50% (HST) -50% (HST) -50% (HST) combination with a 100% increment in car costs will produce the highest gains for the HST. All the scenarios analyzed in the Madrid-Barcelona corridor considered a 50% reduction in travel time for the new HST. Since the plane is the principal competitor to this new alternative, the different policies are focused on plane attributes. Thus, scenario 1 considers only the reductions in travel time for the HST. Scenarios 2 to 5 consider, ceteris paribus, improvements in plane comfort, increases in delay for plane and train; and increases in access and waiting time for plane, respectively. Demand response to the different scenarios is presented in Figure 4. Reductions in the plane’s market share due to the introduction of the new HST do not exceed 15%, in any of the cases analyzed; the plane being the dominant mode in this corridor with predicted market share close to 65%. Demand for the HST was more sensitive to those policies that penalize time attributes of the plane (delay, access time and waiting time in scenarios 3, 5 and 6). Thus, airlines and airports must operate efficiently in order to maintain air transport as a competitive alternative in medium distance corridors. Although, comfort is an important indicator of service quality for air passengers, it is strongly related to the duration of the trip. In this kind of corridor, improvements in the level of comfort, providing planes with more space between seats, do not produce significant variations in market shares. Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 103 Figure 3. Demand response. Madrid-Zaragoza Figure 4. Demand response. Madrid-Barcelona -25.00% -20.00% -15.00% -10.00% -5.00% 0.00% 5.00% 10.00% 15.00% 20.00% 25.00% CC CA BUS HST PLANE Mode Percentage variation in market share SCEN 1 SCEN 2 SCEN 3 SCEN 4 SCEN 5 -20.00% 0.00% 20.00% 40.00% 60.00% 80.00% 100.00% CC CA BUS TRAIN/HST PLANE Mode Percentage variation in market share SCEN 1 SCEN 2 SCEN 3 SCEN 4 SCEN 5 SCEN 6 Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 104 6 Conclusions In this paper, we analyzed competition of the HST with the main competing modes in the Madrid-Zaragoza-Barcelona corridor. The analysis was based on the estimation of disaggregate demand models using both RP and mixed RP/SP databases. Modal utilities for the RP and SP alternatives were defined in terms of the main level-of- service attributes and various socioeconomic characteristics of the individuals. The model specification aims to explain the changes in the demand for HST as a result of changes in travel times, travel costs, access+egress times, headway, reliability and comfort across all the modes that compete in this corridor. In particular we found the interaction of travel time with the travel purpose and comfort (only in the case Madrid-Barcelona) very interesting. Regarding the structure, our models also capture the existence of correlations between bus, train and plane (in Madrid-Zaragoza) and between train and plane in the case of the RP alternatives in the Madrid-Barcelona corridor. In this case, the correlation is expected to be higher when the new mode starts to operate in the corridor because the HST is a closer substitute to the plane than the conventional train. We obtained different measures of willingness to pay for improved service quality. In general, WTP for travel time savings is higher for mandatory trips and the specification of the latent variable comfort allowed us to confirm the hypothesis that it increases as the level of comfort is lower. In the Madrid-Barcelona corridor, we also obtained a high WTP for reductions in delay time. Finally, we obtained the WTP for improved comfort in the plane (more space and legroom) demonstrating that the impact of comfort attributes on the perception of time is not negligible. We have demonstrated that HST demand is inelastic to price, time and especially to headway. However, it is necessary to recognize that in the short-distance Madrid- Zaragoza corridor, the demand is more sensitive to travel time than to price or accessegress time. Regarding cross-elasticities of HST demand to changes in car price and time, it can be seen that policies that penalize the travel cost are more effective than pure congestion of highways. In the case of competition between HST and air transport in the Madrid-Barcelona corridor, we showed that HST demand is more sensitive to air travel cost and access-egress times. We also analyzed demand response to different policy scenarios that consider variations in some level-of-service attributes. In the case of Madrid-Zaragoza, we obtained a low response to policies that only penalize car alternatives by increasing their travel costs. However, substantial gains for the market share of the HST were obtained when these policies are combined with reductions in HST travel time. In the Madrid-Barcelona corridor, demand for the new HST was more sensitive to those policies that penalize time attributes of the plane. The results of our analysis, together with the low rate of return of HST projects, cast some doubts on the potential competition that HSTs can exert in air markets that have been characterized by a high frequency of air services in the past. However, the HST could be a more competitive alternative in the short distance segments (Madrid-Zaragoza and Zaragoza-Barcelona) by trying to capture traffic from car and bus users. Although demand response is a key element of cost-benefit analysis, other aspects such as the impacts of the new infrastructure on regional development and welfare should also be considered prior to the decision to build new transport infrastructures. Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 105 Acknowledgements The authors wish to thank the financial support obtained from the the Spanish Ministry of Transport under its "Programa de Ayudas a la Investigación 2002 en el área de Transporte”. We would also like to express our gratitude to the director of Madrid Airport, RENFE and the managers of the Bus Terminal (Intercambiador de Avenida de América) for their authorization to conduct the surveys in the corresponding passenger terminals. Finally, we would like to thank the two anonymous referees for their helpful comments. References Ahern, A. A. and Tapley, N., 2008. The use of stated preference techniques to model modal choices on interurban trips in Ireland. Transportation Research Part A, 42(1), 15-27. Atkins, 2004. High-Speed Line Study. Department of Environment, Transport and the Regions, London. Ben-Akiva, M. E. and Morikawa, T., 1990. Estimation of travel demand models from multiple data sources. Proceedings 11th International Symposium on Transportation and Traffic Theory, Yokohama, Japan. Bliemer, M. C. J. and Rose, J. M., 2009. Designing Stated Choice Experiments: State- of-the-art. in Kitamura, R., Yoshi, T. and Yamamoto, T., The Expanding Sphere of Travel Behaviour Research, Selected Papers from the 11th International Conference on Travel Behaviour Research, Ch25, 495-498. Blum, U., Haynes, K. E. and Karlsson, C., 1997. Introduction to the special issue: The regional and urban effects of high-speed trains. The Annals of Regional Science 31(1), 1-20. Bradley, M. A. and Daly, A. J., 1997. Estimation of logit choice models using mixed stated preference and revealed preference information, in Stopher P.R. and Lee- Gosselin M. (Eds.) Understanding Travel Behavior in an Era of Change, 209-232. Pergamon, Oxford. Combes, P. P. and Linnemer, L., 2000. Intermodal competition and regional inequalities. Regional Science and Urban Economics, 30(2), 131-184. Cherchi, E. and Ortúzar, J. de D., 2004. On fitting mode-specific constants in the presence of new options in RP/SP models, Transportation Research Part A, 40(1), 1-18. Espino, R., Ortúzar, J. de D. and Román, C., 2006a. Analysing demand for suburban trips: a mixed RP/SP model with latent variables and interaction effects. Transportation, 33(3), 241–261. Espino, R., Ortúzar, J. de D. and Román, C., 2006b. Confidence interval for willingness to pay measures in mode choice models. Networks and Spatial Economics, 6(2), 77-92. Fröidh, O., 2005. Market effects of regional high-speed trains on the Svealand line. Journal of Transport Geography, 13(4), 352-361. González-Savignat, M., 2004. Competition in air transport. The case of the high speed train. Journal of Transport Economics and Policy, 38(1), 77-108. Gutiérrez, J., 2001. Location, economic potential and daily accessibility: an analysis of the accessibility impact of the high-speed line. Journal of Transport Geography, 9(4), 229-242. Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 106 Gutiérrez, J., González, R. and Gómez, G., 1996. The European high-speed train network: predicted effects on accessibility patterns. Journal of Transport Geography 4(4), 227-238. Haynes, K. E., 1997. Labor markets and regional transportation improvements: the case of high-speed trains: An introduction and review. The Annals of Regional Science, 3191), 57-76. Jara-Díaz, S. R., 1998. Time and income in travel choice: towards a microeconomic activity-based theoretical framework, in Garling T., Laitila T. and Westin K. (Eds) Theoretical Foundations of Travel Choice Modeling, 51-74. Elsevier Science, Nueva York. Jara-Díaz, S. R. and Farah, M., 1987. Transport demand and users’ benefits with fixed income: the goods/leisure trade-off revisited. Transportation Research Part B, 21(2), 165-170. Laird, J. J., Nellthorp, J. and Mackie, P. J., 2005. Network effects and total economic impact in transport appraisal. Transport Policy, 12(6), 537-544. Lancaster, K. J., 1966. A new approach to consumer theory. Journal of Political Economy, 14, 132-157. Levinson, D., Mathieu, J. M., Gillen, D. and Kanafani, A., 1997. The full cost of high speed rail: an engineering approach. The Annals of Regional Science, 31(2), 189- 215. Louviere, J. J., Hensher, D. A. and Swait, J. D., 2000. Stated Choice Methods: Analysis and Application. Cambridge University Press, Cambridge. Martin, F., 1997. Justifying a high-speed rail project: social value vs. regional growth. The Annals of Regional Science, 31(2), 155-174. Martín, J. C. and Nombela, G., 2007. Microeconomic impacts of investments in high speed trains in Spain. The Annals of Regional Science, 41(3), 715-733. , J. , J. n, C., 2004. Data Envelopment Analysis (DEA) index to measure the accessibility impacts of new infrastructure investments: The case of the high-speed train corridor Madrid-Barcelona-French border, Regional Studies, 38(6), 697-712. McFadden, D. 1974. The measurement of urban travel demand, Journal of Public Economics, 3(4), 303-328. Nash, C. A., 1991. The case for high speed rail Institute for Transport Studies. The University of Leeds Working Paper 323, Leeds. , J. de .D. and Simonetti, C., 2008. Modelling the demand for medium distance air travel with the mixed data estimation method. Journal of Air Transport Management, 14(6), 297-303. Ortúzar, J. de D. and Willumsen, L. G., 2001. Modelling Transport. 3rd edition. John Wiley & Sons, Chichester. Plassard, F., 1994. High speed transport and regional development, in the European Conference of Ministers of Transport (Eds) Regional Policy Transport Network. OECD, Paris. Rigas, K., 2009. Boat or airplane? Passengers' perceptions of transport services to islands. The example of the Greek domestic leisure market. Journal of Transport Geography, 17(5), 396-401. Rizzi, L.I. and Ortúzar, J de D., 2003. Stated preference in the valuation of interurban road safety. Accident Analysis and Prevention 35(1), 9-22. Rose, J. M., Bliemer, M. C. J., Hensher, D. A. and Collins, A. C., 2008. Designing efficient stated choice experiments involving respondent based reference alternatives. Transportation Research Part B, 42(4), 395-406 Roman, Espino and Martin, Journal of Choice Modelling, 3(1), pp. 84-108 107 Rus, G. de and Inglada, V., 1993. Análisis coste-beneficio del tren de alta velocidad en España. Revista de Economía Aplicada 3, 27-48. Rus, G. de and Inglada, V., 1997. Cost-benefit analysis of the high-speed train in Spain. The Annals of Regional Science, 31, 175-188. Rus, G. de and Nombela, G., 2004. Is the investment in High Speed Rail socially profitable? Documento de trabajo EIT. Universidad de Las Palmas. Rus, G. de and Román, C., 2006. Análisis económico de la línea de alta velocidad Madrid-Barcelona. Revista de Economía Aplicada XIV(42), 35-79. Short, J. and Kopp, A., 2005. Transport infrastructure: Investment and planning. Policy and research aspects, Transport Policy 12(4), 360-367. Sichelschmidt, H., 1999. The EU programme “trans-European networks”- critical assessment. Transport Policy, 6(3), 169-181. Steer Davies Gleave, 2004. High speed rail: International comparisons. Commission for Integrated Transport, London. Train, K., 2003. Discrete Choice Methods with Simulation. Cambridge University Press, Cambridge. Train, K., Wilson, W. W., 2008. Estimation on stated-preference experiments constructed from revealed-preference choices. Transportation Research Part B, 42(3), 191-203. van Exel, J., Rienstra, S., Gommers, M., Pearman, A. and Tsamboulas D., 2002. EU involvement in TEN development: network effects and European value added, Transport Policy, 9(4), 299-311. Vickerman, R. W., 1995. The regional impacts of Trans-European networks. The Annals of Regional Science, 29(2), 237-254. Vickerman, R. W., 1997. High-speed rail in Europe: experience and issues for future development. The Annals of Regional Science, 31(1), 21-38. Vickerman, R. W., Spiekermann, K. and Wegener, M., 1999. Accessibility and economic development in Europe. Regional Studies, 33(1), 1-15.